SlideShare a Scribd company logo
Deliverable 1.5 LinkedTV annotation tool, final release
Names of authors
Jaap Blom
Date of Deliverable
October 29th
2014
Work Package 1: Intelligent hypervideo analysis
LinkedTV
Television Linked To The Web
Integrated Project (IP)
FP7-ICT-2011-7. Information and Communication Technologies
Grant Agreement Number 287911
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 2/53
Dissemination level1 PU
Contractual date of
delivery
30th
September 2014
Actual date of delivery 29th
October 2014
Deliverable number D1.5
Deliverable name LinkedTV annotation tool, final release
File LinkedTV_D1.5.docx
Nature Report
Status & version Final, v1.0
Number of pages 53
WP contributing to the
deliverable
WP1, WP5, WP2
Task responsible The Netherlands Institute for Sound and Vision
Other contributors RBB
Author(s) Jaap Blom
The Netherlands Institute for Sound and Vision
Reviewer Nicolas Patz
EC Project Officer Thomas Küpper
Keywords video annotations, broadcast editors, interactive television,
data curation, user interface
Abstract (for
dissemination)
Editor tool v2.0; curation of automatically generated
annotations; software.
1 • PU = Public
• PP = Restricted to other program participants (including the Commission Services)
• RE = Restricted to a group specified by the consortium (including the Commission Services)
• CO = Confidential, only for members of the consortium (including the Commission Services))
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 3/53
Table of contents
1 Introduction ...................................................................................... 5
1.1 Document overview............................................................................................... 6
1.2 Exclusions............................................................................................................. 6
1.3 Acknowledgements............................................................................................... 6
1.4 History of the document ........................................................................................ 7
1.5 Purpose of the Document...................................................................................... 7
2 User requirements............................................................................ 8
2.1 Involving the end-users ......................................................................................... 8
2.1.1 Tussen Kunst & Kitsch (AVROTROS) ..................................................... 8
2.1.2 RBB....................................................................................................... 11
2.2 Prioritized user requirements............................................................................... 13
3 Updated data model ....................................................................... 18
3.1 Chapter dimensions ............................................................................................ 18
3.2 Information cards ................................................................................................ 19
4 System architecture ....................................................................... 22
4.1 Changed dependencies with other WPs.............................................................. 22
4.1.1 WP2 new components........................................................................... 22
4.1.2 WP3 new component ............................................................................ 23
4.2 Changes to front-end........................................................................................... 24
4.3 Changes to the back-end .................................................................................... 24
5 Functionalities ................................................................................ 25
5.1 Home page & logging in...................................................................................... 25
5.2 Choosing a video ................................................................................................ 28
5.3 Curating chapters................................................................................................ 29
5.4 Chapter dimensions ............................................................................................ 33
5.5 Adding information cards..................................................................................... 34
5.5.1 Using templates..................................................................................... 36
5.5.2 Creating custom information cards ........................................................ 37
5.6 Adding enrichments on demand.......................................................................... 40
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 4/53
5.6.1 Searching and adding new enrichments................................................ 41
6 Customizing the ET ........................................................................ 45
6.1 Open source ....................................................................................................... 45
6.2 Configuration....................................................................................................... 45
6.3 Extending code – adding dimensions.................................................................. 46
6.4 Extending code – outlook.................................................................................... 49
7 Conclusions & future work ............................................................ 50
7.1 Future work......................................................................................................... 50
8 Bibliography ................................................................................... 52
9 Glossary.......................................................................................... 53
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 5/53
1 Introduction
This document reports on the design and implementation of the final version of the editor tool
(ET) v2.0, where its purpose is to serve the program editing teams of broadcasters that have
adopted LinkedTV’s interactive television solution into their workflow. Two of these teams are
currently represented in the LinkedTV project, namely the RBB team and the AVROTROS
team (formerly known as AVRO).
The main purpose of the ET is to provide a means to correct and curate automatically
generated annotations and hyperlinks created by the audiovisual and textual analysis
technologies developed in WP 1 and 2 of the LinkedTV project. Without the intervention of
human editors to correct this data, there is a reasonable risk of exposing inappropriate,
incorrect or irrelevant information to the viewers of a LinkedTV interactive broadcast.
Compared to the previous version of the ET, the main new feature of v2.0 is the possibility to
fetch enrichments, i.e., links to possibly relevant external information sources, on demand.
This means that an editor now not only has the opportunity to choose from pregenerated
enrichments, but also has the possibility to query and retrieve additional enrichments while
using the ET.
This new feature emerged from interviews with end-users, and after consultation with the
LinkedTV consortium, new services were developed in WP2 to support this new approach.
Moreover, resulting from user tests, conducted by the LinkedTV partner CWI, and several
consortium discussions, two new concepts, namely information cards and chapter
dimensions, were introduced into the ET. These concepts increase the reusability of the ET
for different use cases and narrow the gap between what editors see in the ET and what the
audience sees in the LinkedTV player.
Finally, a modest highlight of v2.0 of the ET is that it also aims to be a development toolkit
that enables programmers to configure and extend it with functionalities that are tailored for
their player based on e.g., the multiscreen toolkit2. Note that this is currently a strategy that is
not yet perfectly reflected on a technical level, but is already propagated on the ET’s new
home page and GitHub3 page in order to notify interested parties.
This report is a follow up of the deliverable "D1.3: LinkedTV annotation tool, first release" and
mainly focuses on anything that is news compared to v1.0 of the ET.
2 Developed by Noterik in WP3 for the purpose of building interactive audiovisual applications
3 Collaborative working platform for hosting (open source) software: https://github.com/
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 6/53
1.1 Document overview
In order to update the user requirements, the direct end-users RBB and the AVROTROS
participated in user trials and interviews in order to obtain feedback on the first version of the
ET and new insights into the day-to-day workflow of editing teams. The setup and outcome
of these trials as well as the following new user requirements are described in Section 2.The
two newly introduced concepts, namely information cards and chapter dimensions are
accounted for in Section 3.
Due to the development of different new web services in WP2 and WP3, the architecture of
the ET underwent several changes. Section 4 reports on these new services and the
resulting new system architecture.
The current features of the ET as well as updates to the interaction design are described in
detail in Section 5. Section 6 describes how users of the open source ET GitHub repository
can customize and configure the tool for different (interactive) television programs and their
respective editing teams. Section 7 presents a summary of the main points described and
gives a lookout on future plans. Section 8 presents the bibliography with related deliverables.
Finally, Section 9 presents a glossary to enlighten possibly unfamiliar terminology.
1.2 Exclusions
Unlike what was reported in D1.3, v2.0 of the ET does not include any specific feature with
the purpose of influencing the input of WP4’s personalization systems. The main reason for
this (see also Section 2) is that current users at RBB and AVROTROS did not express an
interest in spending time on specifically curating content for different user groups, however
the same users did express an interest in LinkedTV’s capabilities of automatic
personalization based on curated data.
1.3 Acknowledgements
CWI greatly contributed in sharing results of relevant user trials, defining methodologies and
preparing (and attending) focus group sessions held at NISV and RBB. EURECOM and UEP
put serious effort in adapting their enrichments services to support the on demand approach.
Condat helped out in updating the LinkedTV API to ensure a better connection with the ET.
RBB perfectly fulfilled the role of end user and provided NISV with useful comments
throughout the development of the ET. The LinkedTV consortium as a whole made sure the
ET, touching the work of most work packages, was included in the right discussions at the
right time.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 7/53
1.4 History of the document
Table 1: History of the document
Date Version Name Comment
2014/22/04 V0.1 Jaap Blom,
NISV
Initial ToC + updated user requirements
2014/29/04 V0.2 Jaap Blom,
NISV; Nicolas
Patz, RBB
Updated user requirements
2014/22/09 V0.3 Jaap Blom,
NISV
First draft after finishing software
2014/16/10 V0.4 Jaap Blom,
NISV
Further updates
2014/17/10 V0.5 Jaap Blom,
NISV
Further updates
2014/20/10 V0.6 Jaap Blom,
NISV
Further updates
2014/21/10 V0.7 Jaap Blom,
NISV
QA ready version
2014/27/10 V0.8 Jaap Blom,
NISV; Lampis
Apostolidis,
CERTH
Integrated feedback from WP1 leader
2014/27/10 V0.9 Jaap Blom,
NISV; Nicolas
Patz, RBB
Integrated feedback from formal QA
2014/27/10 V0.95 Jaap Blom,
NISV
Final version
2014/29/10 V1.0 Jaap Blom,
NISV; Lyndon
Nixon, MODUL
University
Final version with integrated feedback from
scientific coordinator
1.5 Purpose of the Document
The purpose of this document is to inform the reader about the progress behind the creation
of the final version of the ET.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 8/53
2 User requirements
2.1 Involving the end-users
For this version of the ET, again4 end-users at RBB and AVROTROS were involved to
provide feedback. This time, the aim was to obtain feedback on the following:
• The day-to-day workflow of program editing teams
• The usability of the ET v1.0.
• The quality of the automatically detected entities and hyperlinks
The first and most important point was to find out more about how program teams work and
how feasible it is, or under what circumstances it would be feasible for them to incorporate
the ET in their day-to-day workflow. The second point was addressed to obtain direct
feedback on the usability of the first version of the ET in order to find out which features
should be kept and which features should be dropped or improved. For the third point, CWI
executed a test with a user group representing the general public, which provided useful
insights5. Because of this and the limited availability of both the RBB editing staff and the
staff of Tussen Kunst & Kitch (TKK), it was decided, at that time, not to go ahead with a
practical test specific for program editors. Regardless of their limited time, both groups
showed much interest in learning about LinkedTV and the ET.
For future evaluations of the ET v2.0 (see Section 7.1), the limited availability of staff will be
taken into account when creating the tests.
For all the above, CWI greatly contributed in defining the evaluation methodologies and the
preparation of materials for the focus group sessions, tests and interviews held.
Subsequently RBB and NISV made sure to apply the defined methodologies as much as
possible with their user groups. The following sub sections describe the setup and outcomes
of the evaluation efforts for TKK cultural heritage show and RBB news show.
2.1.1 Tussen Kunst & Kitsch (AVROTROS)
On the 27th
of April, NISV and CWI held a focus group session with the program team of TKK
on the premises of NISV. The session agenda was as follows:
4 See D1.3: LinkedTV annotation tool, first release
5 The results on this will be reported in WP3’s final deliverable D3.8: Design guideline document for
concept-based presentations
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 9/53
• Short presentation of LinkedTV and the processing workflow (the basic steps from
analyzing video content, using the ET, and final interactive broadcast)
• Presentation of the main functionalities of the ET
• Presentation of user tests with the LinkedTV second screen application6
• Presentation of future plans of LinkedTV
Throughout the presentations the members of the TKK team could ask questions and
provide immediate feedback. During the presentations, the NISV and CWI hosts asked the
TKK team questions such as:
• whether the presented aspect/feature would fit their current workflow or could be
otherwise adapted to be made to fit their workflow
• what features are especially appreciated and what features are particularly disliked
• whether (if applicable) a feature would target a specific part of their audience
The following table lists the feedback that is applicable to the development of the ET.
Feedback trigger Feedback Effect on ET/LinkedTV
As a reaction to
showing LinkedTV’s
personalization
approach.
“Personalization should be done
automatically, because there are not
a lot of editors in the team. TKK only
have the general public in mind as
their audience, so all curation is
done for this type of audience.”
• The ET should not incorporate
specific features for
personalization.
• The ET does not need to
distinguish between target
audiences.
In reaction to the
chapter editing
functionality of the ET.
“The start and end times of chapters
are available during montage, so
automatic chapter detection is not
necessary if this information could be
incorporated into the system.”
“The short segments of the show are
also important as it discusses
relevant topics as well.”
• For TKK the ET needs to be
able to load chapter
segmentations provided by the
AVROTROS.
• In case of automatic chapter
detection, for TKK the short
chapters are also relevant.
In reaction to showing “We are wondering how many • The ET should highlight the
6 Reported in D6.3: User trial results
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 10/53
Feedback trigger Feedback Effect on ET/LinkedTV
the way an editor uses
the ET to select useful
automatically detected
entities for creating
annotations.
entities are detected on average and
should thus be inspected by the
editors. We can’t spend much time
on this.”
“Seeing the location where the
episode was filmed is not very
interesting [as an entity], but other
locations, such as where the art
objects were found are.”
most relevant (automatically
detected) entities and/or hide
irrelevant entities.
In reaction to showing
the way an editor uses
the ET to select
automatically detected
enrichments/hyperlinks.
“There are way too many
enrichments showing in the ET.”
“Key frames and a short title or
keywords are not enough to judge
whether a hyperlink is relevant or
not.”
“A great feature would be that the
editor selects a number of entities
and then the system would look for
relevant hyperlinks. In the current
situation there are just too many
(irrelevant) links.”
“Some entities are interesting for
inside links, such as e.g. ‘gold’ [show
other episodes with gold objects], but
not for outside links [a Wikipedia
page on ‘gold’ is not interesting.]”
“It’s important to be able to find links
that are based on current events.”
“It’s important to be able to find links
that are surprising to the editor.”
• The ET should be able to reduce
the amount of (irrelevant)
enrichments, in order to reduce
the information overload.
• The ET should have a feature
that allows a user to fetch
enrichments on demand (based
on a selection of one or more
entities).
In reaction to showing
the way an editor adds
annotations using the
ET.
“For other programs [at AVROTROS]
there are boost factors for certain
concepts to improve
recommendation.”
• The ET could be extended with
a feature for adding boosting
values to annotations in order to
aid the WP4 recommendation.
As general feedback on “Currently the editing team does not • Ideally the ET could fulfill the
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 11/53
Feedback trigger Feedback Effect on ET/LinkedTV
the ET as a whole. add any contextualization to TKK.”
“TKK does not have an interactive
website because there was never a
need for it. But it has a potential. The
internet editors publish the press
release, but do not add more links.
The program and top pieces are
described. On social media
(Facebook) they give some info to
get people’s attention.”
“TKK used to have lists of keywords
for each chapter in the episode.”
“There are high quality production
photo’s available.”
“The editor should be responsible for
finding links in the ET. He/she should
be able to select concepts and the
system should look for (surprising)
links.”
“Overall: there is too much ‘noise’
and too many links.”
role of simply being a
convenient tool for finding
contextual information for
television programs.
• The LinkedTV platform should
be able to deal with different
forms of (meta)data coming from
the production phase of
television programs.
• The ET should focus on helping
editors quickly find relevant (and
surprising) links.
Table 2: Feedback from the AVROTROS
2.1.2 RBB
On the 5th
of June 2014 two editors worked together with a colleague from the RBB LinkedTV
team and discussed the ET in its current version. As a reminder of the motivation for showing
this tool, first the end user GUI demo was presented again, then the news show was briefly
analyzed in the sense that the attendants agreed on the chapters’ headlines and the editors
had time to use the ET and make comments and give direct feedback. The ET itself was
briefly presented and some strategic tips were issued in order to guide the editors, but the
general approach was to let them use the tool and note any issue they may find as any tool
that would need hours or days of training and coaching would not have a chance to be
accepted into everyday workflows.
The following table lists the feedback that is applicable to the development of the ET.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 12/53
Feedback trigger Feedback Effect on ET/LinkedTV
In reaction to the
chapter editing
functionality of the ET.
“The prefabricated chapters were not
correctly marked. Beginnings and
endings did not comply with the
beginnings and endings of the
reports and anchor presentations. So
I had to create new chapters on my
own which took about 25 minutes
and would not have been affordable
in every day’s business.”
“The handling of the editor tool is not
intuitive enough especially when it
comes to naming the chapters. For
example most of the new chapters
didn’t even have a thumbnail. So I
couldn’t be quite sure whether
changes of naming of chapters were
automatically saved or not.”
• Automatic chapter detection
should be improved to
accelerate the editing process.
• Chapters must carry a key frame
thumbnail.
In reaction to showing
the way an editor uses
the ET to select useful
automatically detected
entities for creating
annotations
“The prefabricated entities were not
as useful as hoped. I suppose only a
quarter made sense and was useful.
So I had to view every single news
report again, note catchwords and
time codes to create new additional
links. That process again took some
time, at least 20 to 30 minutes per
report.”
• No Editor Tool issue.
In reaction to showing
the way an editor uses
the ET to select
automatically detected
enrichments/hyperlinks
“Creating additional links means in
my opinion too many working steps
which would even not be affordable
in everyday business. It would be
easier and helpful if mark-in and
mark-out could be set right beneath
the player while watching every
single report to create correct time
codes. It would also be timesaving if
there was an input screen for naming
and additional links right beneath the
• Reduce number of steps for
creating new annotations and/or
links.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 13/53
Feedback trigger Feedback Effect on ET/LinkedTV
player.”
In reaction to showing
the way an editor adds
annotations using the
ET.
“In the beginning I was quite
uncertain where to put a URL while
creating additional links. Also the
usage of the "add“-button seemed
not that self-explaining to me. Trial
and error was the only way to find
out, which took again some time.”
“Searching and naming of DBpedia-
links was also not self-explaining
because the input screen was
grayed out and seemed deactivated
to me. That’s why I also didn’t get
the possibility of creating several
additional links in one entity-screen.”
• Clearer guidance and/or a
manual.
As general feedback on
the ET as a whole.
“As I could see umlauts can not be
used but unfortunately there is no
warning message if you mistakenly
do so. So after having mistakenly
used umlauts the complete input
screen for additional links disappears
after saving. So the editor has to do
start from scratch.”
“A preview would also be very
helpful so that the editor can see the
final product while working on it and
possible errors.”
• Enable use of Umlauts.
• Enable preview templates.
Table 3: Feedback from the RBB editing staff
2.2 Prioritized user requirements
Table 4 lists the user requirements for v2.0 of the ET, while Table 5 reports the unfulfilled
user requirements taken from D1.3 that haven’t made it in v2.0 either.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 14/53
ID Requirement Description Priority In v2.0
AUT1 User login. Users can log into the
system using a user name
and password.
Must-
have
Yes
AUT2 Authorization by user
organization.
Users can only access
content of their own
organization.
Must-
have
Yes
OC1 Content for public users. For public users there it
should be possible to test
the tool with a cc-licensed
video.
Must-
have
Yes
HOM1 There should be an attractive
home page with useful context
information.
The home page should
explain LinkedTV and the
purpose of the tool.
Must-
have
Yes
CH1* The creation and editing of
chapters should be more user
friendly.
Nice-to-
have
Yes
ENT1* Cluster entities of the same
name, to enable quicker visual
inspection of the entities.
Currently, when detected
multiple times, entities of
the same name are
displayed for each
occurrence.
Must-
have
Yes
ENT2* Group entities of the same
(DBpedia) type, to enable
quicker visual inspection of all
the entities.
Currently entities of the
same DBpedia class/type
are not grouped together.
Nice-to-
have
No
ENT3* Remove the key frames shown
for each entity.
Currently each entity is
displayed with a key frame
of the corresponding frame
in the video; this fills up the
screen unnecessarily.
Must-
have
Yes
* These requirements all refer to updates of existing functionalities from v1.0. Requirements without an
asterisk are completely new requirements.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 15/53
ID Requirement Description Priority In v2.0
REL1 The enrichments should include
links to related material.
Currently the enrichments
only consist of links taken
from the white list.
Must-
have
Yes
ENR1 External and internal links should
be clearly distinguishable.
It should be visually clear
which links are internal and
which are external.
Must-
have
Yes
ENR2* More recent links should be
shown on top.
Currently displayed
enrichments are not
ordered by date.
Should-
have
Yes
ENR3 It should be possible to generate
enrichments (on the fly) based
on the selection of one or
multiple entities.
Currently the system only
offers enrichments that are
based on each single
automatically generated
entity. In order to get more
meaningful enrichments,
an editor should be able to
select one or more entities
for generating enrichments
on the fly.
Should-
have
Yes
SAV1* Publish the editor’s choices, tags
in the LinkedTV platform.
In v1.0 the save data was
stored in a separate RDF
graph. In v2.0 this data
should be integrated in the
overall LinkedTV graph7.
Must-
have
Yes
PRV1 Preview play-out of editor’s
choices.
A user can view a
simplified version of the
front-end’s play-out, based
on the editor’s choices, by
utilizing a video preview
mode
Should-
have
No
ST1 Track statistics based on the
user actions.
Statistics related to the
user’s behavior with
respect to the acceptance,
Should-
have
No
7 See D2.2 Specification of lightweight metadata models for multimedia annotation
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 16/53
ID Requirement Description Priority In v2.0
rejection and modification
of (automatically
generated) annotations &
hyperlinks are logged for
various evaluation
purposes.
Table 4: User requirements classified with the MoSCoW method and the availability in v2.0
ID Requirement Description
WL1 Edit items in user
organization’s white list.
A user can edit (create, update or remove) the (seed)
URLs in the user organization’s white list.
TPL1 Edit template based on the
user’s organization.
A user can edit the who-what-where template of the
user’s organization.
REC1 Add user tags to describe
the video as a whole.
A user can select tags from a predefined list to globally
describe the main topics of the selected video.
REC2 Add user tags to describe
individual chapters.
A user can select tags from a predefined list to globally
describe the main topics of the selected chapter.
Table 5: Unfulfilled user requirements from v1.0 that could not be realized it in v2.0 either
Table 5 shows requirements that did not make it in v1.0 and won’t be implemented in v2.0 of
the ET either. The reasons for this vary per requirement:
• WL1: There is currently no functionality for this in the LinkedTV API. Implementing
this in the ET therefore is not possible.
• TPL1: The idea of program templates, e.g., one for RBB News and one for TKK,
could not be practically worked out into actual features of the system. Moreover the
concept of entity meta types, namely ‘who’, ‘what’ or ‘where’ was dropped in the ET
v2.0 as it was no longer considered very relevant in combination with the new
concept of chapter dimensions (see Section 3.1).
• REC1 / REC2: This was implemented in an intermediate version of V1, but because
of other more important issues, no model for storing and utilizing the tags in the
platform has been worked out.
Even though not all of the desired features have been implemented, all must-have
requirements listed in Table 4 have been met. Moreover, because LinkedTV appreciates the
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 17/53
need for specific requirements for different (types of) programs, the ET is available (open
source) for any interested party to be further enhanced or customized (see Section 6).
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 18/53
3 Updated data model
Based on new insights, mainly following from the user trials in year 3, two new concepts
were introduced to the ET data model, namely (chapter) dimensions and information cards.
The following sections describe these concepts, which have considerably influenced the
functionalities of the ET v2.0.
3.1 Chapter dimensions
The idea of having chapter dimensions came from the need to semantically organize the
annotations added to a video (chapter). This need became apparent after the development
and user test with the Linked News demonstrator8 developed by CWI. In this demonstrator
the available context information offered to the user, i.e., the video annotations, were all
distributed along different “axes” or as they are now called “dimensions”.
Each dimension in this demonstrator had a different intension of use e.g.
• in depth: show related in depth articles
• opinion: show related opinionated articles
• timeline: show the related information in a timeline
With these examples in mind, the idea came to include the notion of dimension into the
LinkedTV model9. Using dimensions greatly increased the possibility of expressing use case
specific semantics in a generic way:
• each dimension should, by clearly defining its semantics, reflect how information
added to it should be used in an application
• there is no limit to the amount of dimensions that can be added, thus for each subtle
user feature a new dimension could be added
For the ET having dimensions as configurable entities in the software greatly improves the
reusability and adaptability of the tool for different use cases (see Section 6).
In the ET dimensions are currently configured as follows:
• each program (RBB News, TKK show) needs to define one or more dimensions
o Each dimension has:
 an identifier
 a name (to display in the UI, see Figure 15)
8 See D3.5: Requirements Document LinkedTV User Interfaces (Version 2) for the initial demonstrator
and D3.7: LinkedTV User Interfaces Selected and Refined for the eventually developed companion
application
9 See D2.2: Specification of lightweight metadata models for multimedia annotation
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 19/53
 a web service that handles the fetching of enrichments, currently the
following (web) services are supported (see also Section 4):
• TVEnricher
• TVNewsEnricher
• InformationCards10
For illustration, the actual configuration looks like this (in JavaScript):
Figure 1: JavaScript configuration of the TKK dimensions
In Section 6.3 the possibility to customize dimensions is addressed in more detail.
3.2 Information cards
In the ET v1.0, the editor was not able to control the information that is shown for entities
related to annotations. All an editor could do when creating and annotation was:
10 The informationCards service is not a separate web service, but rather functionality embedded in
the ET (read more in Section 5.5)
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 20/53
• assign a single entity to it (select an auto generated one or add a custom DBpedia
URL)
• assign multiple enrichments (select an auto generated one or add a custom URL)
Figure 2: ET v1.0 showing the annotation edit dialog. The user could only add a single URL
representing the (named) entity and add multiple enrichments URLs
Only able to use the ET this way, the implementation of displaying the actual information
related to the entity URL was up to the LinkedTV player11, which also could only do a
dynamic lookup using the entity URL (usually a DBpedia or Wikipedia URL) to fetch the
actual information.
Besides needing control over the information shown to viewers, some topics simply cannot
be expressed very well using single (named) entities, but actually require the creation of
custom entities that could have multiple links to available named entities (expressed as links
to e.g. DBpedia).
One such example is the art object, which represents the main topic for the TKK videos and
fits the aforementioned type of entity:
• most likely there is no online authoritative resource describing the art object shown in
the program
11 Or rather a LinkedTV player, using the platform API different player can be implemented
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 21/53
• art objects consist of possibly multiple references to entities that do exist in
authoritative resources such as DBpedia or the AAT
So for entities such as art objects, like in TKK shows, or for entities that require a high level
of curation, e.g., to ensure high objectivity, which is the case in RBB News, the concept of
information cards was conceived.
In the ET these information cards are presented as a set of key/value pairs where the key is
the name of the intended property, e.g., “title”, “description”, “creator”, and the value is either
a:
• literal (a text string)
• entity, which has:
o label (a text string)
o URI (in this version of the ET a URL to DBpedia)
Since information cards could in theory consist of an infinite number of key value pairs, the
ET incorporated a templating mechanism to enable configuring templates for editors, aiming
to provide a structured way of adding information that can eventually be published to their
LinkedTV player. Figure 3 shows the configuration of the art object information card
template.
Figure 3: ET configuration – information card template for the TKK art object
How information cards and their templates are used is described in Section 5.5.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 22/53
4 System architecture
The architecture described in D1.3 still applies for the most part in v2.0. The biggest changes
in the architecture are:
• the incorporation of several new web services developed within WP2 to support
enrichments on demand
• the incorporation of the EntityProxy developed in WP3, to support the dynamic
inspection of related properties of automatically detected named entities
• the migration of front-end code from Backbone.js12 to AngularJS13
The following sub sections describe the details of these changes.
4.1 Changed dependencies with other WPs
Several new web services have been developed within LinkedTV, which enable the ET to
allow users to trigger enrichments based on the selection of a number of entities. This
section briefly lists these new services per work package they were developed in. For full
details on these services, please consult the specific deliverables they were reported in
(which are referenced in the following sub sections).
4.1.1 WP2 new components
TVEnricher
This service is developed by the LinkedTV partner EURECOM and exposes an API that
allows the ET, or any other program, to request enrichments on demand, based on a query.
In fact, the TVEnricher is a proxy service that passes received queries on to several other
services developed in LinkedTV, namely:
• IRAPI*14: a search engine on top of an index created from crawled web documents
based on RBB’s and TKK’s white list
• MediaCollector*: a service that can search through the items published during the
past 7 days in 10 different social media platforms
• LinkedTV SOLR index*: a service that stores information of related video fragments
• Europeana API15: an API that enables to retrieve items from the Europeana
collection
The TVEnricher is configured for both RBB and TKK (see Section 6).
12 http://backbonejs.org/
13 https://angularjs.org/
14 * For more information on these LinkedTV services see D2.6: LinkedTV Framework for Generating
Video Enrichments with Annotations.
15 http://pro.europeana.eu/api
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 23/53
TVNewsEnricher
From an ET perspective, the TVNewsEnricher basically fits the same purpose as the
TVEnricher; the difference is that it is optimized for the news domain of the RBB use case.
Instead of the services listed for the TVEnricher, the Google Custom Search Engine16 and
the Twitter API17 are used. The result is an API that offers 5 different dimensions to query
along:
• In depth: news articles that offer in depth information on the specified subject
• Other media: news articles from other media sources on the specified subject
• Opinions: opinion pieces on the specified subject
• Timeline: a timeline comprising of news articles on the specified subject
• Twitter: related tweets of the past 7 days on the specified subject
EntityExpansion
The EntityExpansion service (see D2.6), developed by EURECOM, is in the current version
not yet fully integrated, but is on the list of useful extensions for the RBB News configuration.
Basically this service offers the on demand retrieval of additional (named) entities,
specifically for the news domain. The idea is that rather than just analyzing the metadata
(and subtitles and so on) of the news program itself, it analyses related web documents on
the same news topic. This way there is also an opportunity to find entities that come from a
broader context, e.g. including yesterday’s news, than just the news item itself. The inclusion
of the EntityExpansion service is listed under future work in Section 7.1.
4.1.2 WP3 new component
EntityProxy
The EntityProxy18, developed by CWI, is a service that takes DBpedia and Wikipedia URLs
as input and, based on the supplied DBpedia type, returns a list of related properties from
DBpedia.
In the ET this service is used to retrieve information for professional end users when they
want to inspect the details of certain automatically detected entities (see Figure 22). Without
this functionality the user could only judge the value of an entity by its label. So rather than
only reading: “Jan Steen”, by clicking on this label a user can retrieve the additional
information, that Jan Steen was a Dutch painter, and read a description of his life and
anything else that would be typically displayed in an information box on a related Wikipedia
page.
16 https://www.google.com/cse/
17 https://dev.twitter.com/rest/public
18 http://linkedtv.project.cwi.nl/
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 24/53
4.2 Changes to front-end
As mentioned above it was decided to rebuild the front-end for v2.0. The reason for this is
that with the addition of dimensions and information cards, the client side data model has
changed enough for it to be worthwhile to just rebuild it rather than adapt the old code.
Moreover the incorporation of the new on demand functionalities required a different
approach in (asynchronously) loading data from different services.
Another good reason to rebuild the front-end was to get closer to possible contributors for the
ET GitHub, by choosing AngularJS instead of Backbone.js because of its rising popularity
and inherent capabilities to create reusable components.
Next to switching to AngularJS, using SASS19 (in combination with Compass20) further
improved the possibility to create reusable interface components. SASS is a precompiler for
CSS that enables to modularize style sheet code, making it more readable, reusable and
extensible for different programmers.
4.3 Changes to the back-end
To support the new on demand functionalities, all of the components mentioned in Section
4.1.1 and 4.1.2, namely the TVEnricher, TVNewsEnricher and EntityProxy were integrated
into the existing Python Django21 backend. In this case integration means that the ET back-
end handles the HTTP communication between these components ensuring that the user
input is translated to valid (component specific) queries and response data is mapped to the
ET data model.
Due to ongoing changes and updates of the LinkedTV platform, it was decided to wait with
the storage of the new data model to the platform. Instead, all of the editor’s efforts are
currently stored in the ET itself. For this, a Redis22 store is used to store all data in JSON
format.
Listed under future work (see Section 7.1) is the task for storing all of the curated data in the
LinkedTV platform in the form of RDF triples.
19 http://sass-lang.com/
20 http://compass-style.org/
21 https://www.djangoproject.com/
22 http://redis.io/
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 25/53
5 Functionalities
As a result of the new model (see Section 3) and the new system architecture (see Section
4), and aiming to support the new on demand capabilities, the functionalities of v2.0 have
changed in several ways.The following sub sections describe these new features in detail
and incorporate screenshots for illustration. To increase comprehensibility, the order of the
sub sections is based on the recommended workflow for editing a program, which is:
• visit the home page
• login (currently as a RBB or TKK user)23
• curate chapters
• per chapter: fetch & curate enrichments along different dimensions
For better understanding, it is recommended to try out the described functionalities using the
free trial video in the ET. To do this, open a Chrome24 browser and navigate to
http://editortoolv2.linkedtv.eu/trial. Chrome is needed as it supports the video format currently
used by the HTML5 player.
5.1 Home page & logging in
The ET home page can be reached at http://editortoolv2.linkedtv.eu, and informs the visitor
about the following:
• What is the ET?
• What is LinkedTV?
• Try out the ET yourselves!
• The ET is open source. Please visit us on GitHub!
• Who are the people behind LinkedTV?
• Special thanks to our external partners
The home page is designed to attract interested parties in finding out more about LinkedTV.
Also visitors are given the chance to try out the tool themselves, ultimately hoping that they
will be interested contributing to the ET and/or using LinkedTV technology.
The video used for the trial is creative commons25 licensed and comes from Open Beelden26,
which is a site hosted by NISV.
Figures 4-7 show the different sections of the home page.
23 It is also possible to skip login and go to a free trial page (see Figure 5)
24 http://www.google.nl/intl/nl/chrome/browser/
25 http://en.wikipedia.org/wiki/Creative_Commons
26 http://openbeelden.nl/
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 26/53
Figure 4: Home page – top section informing about the ET
The first section informs the visitor about the overall capabilities and purpose of the ET.
Figure 5: Home page – link to the trial page and link to the ET on GitHub
The second section informs the reader about the trial version of the ET.
The third section directs the visitor to the ET GitHub page and informs the reader about the
possibilities of customizing the ET for specific use cases.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 27/53
Figure 6: Home page – link to project website and partner credits
The fourth section presents a link to the LinkedTV project website.
Figure 7: Home page – special thanks section
The final two sections show the logos of consortium partners and affiliated partners.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 28/53
For accessing the ET, known users, currently being RBB and TKK, have been authorized to
login to the system so they can curate their videos. By clicking “login” in the navigation bar at
the top of the page the user is directed to the login page:
Figure 8: Login screen
After successfully logging in, the user is directed to the video selection page, which is
described in the next section.
5.2 Choosing a video
The following image illustrates the video selection screen, which shows all of the videos that
were successfully processed by the LinkedTV platform for the organization the user is a
member of.
Figure 9: Video selection page showing a thumbnail and title of each video
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 29/53
By clicking on one of the video thumbnails, the user is directed to the editing page of the
selected video. The following sections describe the different functionalities of this page.
5.3 Curating chapters
The LinkedTV platform offers several means of automatically identifying the different
chapters of a video (see Section 3 of D1.4: Visual, text and audio information analysis for
hypervideo, final release). In most conceivable cases, depending on each television
program, a chapter is a part of the program that is about a single topic. For RBB for instance
each news item is a separate chapter; for TKK the main chapters are about a single art
object.
Before a user starts adding annotations, linking different kinds of enrichments (see Sections
5.5 and 5.6), the possibly available automatically generated chapters, need to be edited in
order to make sure they are correct.
Figure 10: Chapter panel – showing autogenerated chapters only
Figure 10 shows the video chapters of a TKK program that has not yet been edited; the cog
icon below each chapter thumbnail indicates that a chapter was automatically generated and
not edited by a human (curated). In this example it is clear that the chapters were created by
the chapter segmentation algorithm, which relies on the TKK bumper re-detection27, as the
TKK logo can be seen in each (except the first) thumbnail, which always reflects the starting
frame of the chapter. For correctly editing chapters, the user is recommended to:
27 CERTH/ITI implemented this algorithm for the TKK chapter segmentation as each main chapter in
TKK starts with showing the TKK logo (see also D1.4 in the Bibliography).
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 30/53
• create proper titles for each chapter
• make sure to check the boundaries of each chapter are correct
• add possibly missing chapters if necessary
• remove possibly unnecessary chapters
In order to edit the predefined chapters a user can double click on one of the thumbnails to
open the following dialog:
Figure 11: Chapter edit dialog – the shot selection is hidden
When clicking the “Select shots” bar the shot selection is opened:
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 31/53
Figure 12: Chapter edit dialog – showing shot selection; shots included in the chapter are highlighted
The edit dialog can be used to change the title as well as the start and end times (in
milliseconds) of the chapter. As directly entering the amount of milliseconds is not very
convenient, the user can select a starting and ending shot by using the shot selection panel.
When doing so, the amount of milliseconds is automatically set in the start and end fields. If
after saving the start and end is not perfect, the user can fine-tune them by adjusting the
values in the corresponding fields.
When a chapter has been edited and saved, the cog icon in the chapter panel is replaced by
a ‘human’ icon, indicating that the chapter has been curated (see Figure 13 below).
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 32/53
Figure 13: Chapter panel – showing two curated chapters
When pressing the delete button (at the bottom, see Figure 12) a chapter can be deleted.
Although removed from the view, this will not affect automatically generated chapters to be
removed from the LinkedTV platform. When reloading the page, all automatically generated
chapters are shown on the page again. To avoid confusing the user, the filter button in the
top right of the chapter panel can be used to only show the chapters that were curated (see
Figure 14)
Figure 14: Chapter panel – showing curated chapters only (filter activated)
Finally by pressing the “New chapter” icon a new chapter can be added. The same screen as
Figure 11 is shown and after saving, the chapter is simply added to the list of chapters.
After a user finishes the editing of the automatically generated chapters, it is time to start
adding annotations to each one of them. In this version of the ET there are currently two
kinds of information that can be attached to annotations
• information cards
• enrichments
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 33/53
Section 5.5 describes everything there is to know about working with information cards and
Section 5.6 delves into how to work with all the other types of enrichments. However before
going into detail about the actual addition of annotations it is necessary to explain about how
these annotations are all going to be related to certain dimensions per chapter.
The following section describes how the user can interact with configured chapter
dimensions.
5.4 Chapter dimensions
As described in Section 3.1, each chapter has one or more dimensions configured for adding
enrichments to. The result of such a configuration, in this case for TKK, is reflected in the
annotation panel when navigating to a chapter (by selecting one in the chapter panel as
shown in Figure 10):
Figure 15: Annotation panel – showing the dimensions configured for TKK
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 34/53
If enrichments already have been added this panel looks like the following:
Figure 16: Annotation panel – showing two dimensions with saved annotations
After curating chapters, adding enrichments, in either the form of information cards or
‘regular’ enrichments, is one of the main tasks of the editor. The following two sections
describe how these different kinds of enrichments can be added to each dimension.
5.5 Adding information cards
Following Section 3.2, Figure 1 shows that the dimension named ‘Art object’ is configured to
be an information card dimension. The effect of this is that when double clicking the ‘New’
button in the ‘Art object’ dimension (see Figure 16) opens an empty information card edit
dialog (see Figure 17). When double clicking an existing information card the same screen is
opened, but with all of the saved properties filled out in the form at the top (see Figure 18).
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 35/53
Figure 17: Information card edit dialog – showing an empty art object template
Figure 18: Information card edit dialog – showing an information card being edited
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 36/53
Figure 18 shows the information card dialog partly filled out and with all possible elements
showing:
• Edit form: here the template can be selected and the form can be filled out
• Shot selection panel: here a shot can be selected to be used as poster image for
the information card
• Entity selection panel: for inspecting the properties of automatically generated
entities
• Entity details panel: showing the properties of the selected entity in the entity
selection panel
Before explaining the functionalities of each of these parts in detail, the editor currently has
two different approaches to choose from:
• use one of the preconfigured templates to fill out
• manually create an information card
Both approaches as well as the purpose of the aforementioned panels are described in the
following two sections.
5.5.1 Using templates
Using a template is the most straightforward way of adding an information card, following the
steps described below.
• The user fills in the pregenerated form
• The user selects a shot using the shot selection panel
• The user saves the card
For choosing a template, the user needs to make sure to select one of the configured
templates from the pull down list as shown in Figure 17. When done so, automatically all of
the key value pairs configured (see Section 3.2) for that template are displayed in a form.
For each configured literal, e.g., ‘label’ and ‘description’, the form provides a single text field.
For each configured entity, e.g., ‘creator’ or ‘period’, the form displays a DBpedia search box
(see Figure 18) and a separate text field to be able to edit the label of retrieved DBpedia
entities.
Filling out the form
The text fields are just plain input fields where the user can enter free text. Using the
DBpedia search boxes, the user must choose a DBpedia concept in order to properly fill out
the field. After selection of an entity, the user is allowed to change the label of the retrieved
entity; the underlying DBpedia URI is not displayed to the user.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 37/53
Selecting a shot
Figure 19: Information card edit dialog - shot selection panel
Using the shot selection panel as shown in Figure 19, the user can select one of the
automatically detected shots to be used as a poster image for the information card. After
filling in the form, and choosing a shot, the user can press ‘Save’ at the bottom of the dialog,
to save the information card.
5.5.2 Creating custom information cards
If there is no suitable template configured, the user can add custom information cards based
on either:
• an automatically detected entity
• an entity found using the DBpedia search box
For this, the user initially makes sure to deselect the “Use template” checkbox at the top of
the dialog:
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 38/53
Figure 20: Information card edit dialog – creating a custom information card
Next, the user can select an entity by either selecting one from the “Detected entities” list or
search for one using the DBpedia search box.
Figure 21: Information card panel – the entity ‘Hilversum’ was found using the DBpedia search box
When a user selects an entity from either the list (see Figure 20) or the one displayed after
using the DBpedia search box (see Figure 21), CWI’s EntityProxy (see Section 4.1.2) is
called and fetches related properties by analyzing (if any) the entity’s related DBpedia URL28.
28 The EntityProxy also accepts Wikipedia links, which are transformed into DBpedia URLs. This
transformation however does not always work, as Wikipedia links do not always have direct DBpedia
counterparts.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 39/53
When successful the following information appears at the bottom of the screen (“Fetched
properties”).
Figure 22: Information card edit dialog - showing the data fetched from the EntityProxy
Using the displayed information the user can decide whether the selected entity is indeed the
one needed for this information card. If so the displayed properties can be attached to the
card by clicking on the “Use as card” button. After this the “Fetched properties” are used to
populate the form, representing the properties of the information card, at the top of the
screen.
Figure 23: Information card edit dialog – after setting an entity as the basis for the information card
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 40/53
Finally the user can edit the copied information to make sure it is appropriate for the eventual
viewers of the program. In case the user does not appreciate the poster image copied from
the selected entity, it is possible to select one by choosing a shot from the shot selection
panel in the same manner as can be done when using the template (see Figure 19). After
finishing the editing the user can save by clicking the save button, after which, the dialog is
closed and the user is returned to the main screen.
5.6 Adding enrichments on demand
In the ET v1.0, the user could only select automatically detected enrichments from a static
precalculated list (see D1.3). Moreover each of these enrichments was found based on a
single automatically generated entity each, often being irrelevant or too unspecific.
After having tested this version of the ET with the TTK and RBB program teams the outcome
was that they all thought that going through long lists of enrichments simply is too time
consuming. Instead it was suggested that it would be very useful to be able to pick one or
more automatically detected entities and use these as input for fetching enrichments on
demand. This way the problem of solely having enrichments based on single entities is
solved. Moreover, since the user requests the enrichments on demand, it is instantly clear on
what basis each enrichment is found.
To support this new on demand approach EURECOM had to work on extending their
enrichment tools, namely the TVEnricher and the TVNewsEnricher (see Section 4.1.1), with
APIs that could be accessed by the ET in the following way:
• Input: a query
• Output: a set of hyperlinks with the following properties
o Source: URL or name describing the source (website or system)
o Label: describing the hyperlink
o Poster: (if available) a relevant thumbnail
Unlike dimensions configured with the information card functionality (see previous section),
dimensions configured with services supporting the on demand approach all use two dialog
screens to fetch and edit enrichments:
• searching enrichments on demand; adding/deleting multiple enrichments
• editing/deleting single enrichments
The reason for this difference in respect to the information card dialog is that for the on
demand approach, many enrichments can be found and added at the same time, resulting in
a different interaction design.
The detailed functionalities of the aforementioned dialogs are described in the following two
sections.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 41/53
5.6.1 Searching and adding new enrichments
When no enrichments have been added yet, the user can double-click on the ‘Search / Add’
(see Figure 16) button to open the search enrichment dialog (see an example in Figure 24
below).
Figure 24: Search enrichment dialog – before searching for enrichments
The idea behind this dialog is the possibility to use either of the following to find related
enrichments:
• Curated enrichments: these are all the information cards and enrichments that are
already saved in the selected chapter
• Expanded entities: (only RBB news) these are entities that were fetched on demand
using the EntityExpansion service. However, this has not been implemented yet.
• Autogenerated entities: these are the automatically detected entities
Each of these three sources can be inspected using the tabs displayed below “Choose
entities”. If the user does not find any useful entities, it is always possible to enter search
keywords in the search field at the bottom (see Figure 25). After choosing entities or entering
a search query, the user can trigger the fetching of enrichments by pressing ‘Find links’.
While searching the text of this button will change into ‘Loading…’
Figure 25: Search enrichment dialog – while searching for enrichments
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 42/53
After searching is done, the enrichments are displayed at the bottom of the page, as
illustrated in Figures 26 and 27 below.
Figure 26: Search enrichment dialog – after finding enrichments in the ‘background information
dimension’
Figure 27: Search enrichment dialog – after finding enrichments in the ‘related fragments’ dimension
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 43/53
Depending on which dimension, the results show different information29, but can all be added
and filtered in the same way. By simply double-clicking on one of the results it is added to the
list of saved enrichments at the top (see Figure 28):
Figure 28: Search enrichment dialog – after selecting an enrichment from the found enrichments.
When selecting one or more enrichments in this way and subsequently pressing the ‘Save’
button at the bottom of the dialog, the enrichments are saved to the selected dimension
within the currently active chapter.
Since this dialog is already crowded with different user interface elements, the enrichment
can be further edited in a separate dialog, which can be opened by double clicking on a
saved enrichment as shown in Figure 16. The dialog can be used as follows:
29 The TVEnricher and TVNewsEnricher themselves support multiple dimensions
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 44/53
Figure 29: Enrichment edit dialog – editing a newly saved enrichment
Next to editing the three most basic properties of an enrichment, namely the URI, label and
poster, a user can also determine the start and end time when this enrichment should be
displayed within the active chapter. Similar to the chapter edit dialog (see Figure 12), the
(rough) boundaries of the selection can be conveniently set by choosing two shots from the
shot selection panel.
Even though none of the LinkedTV players so far utilizes this information, this functionality
adds much to the reusability of the ET as it can support use cases where editors need to be
able to control the exact moment a piece of information is offered to viewers.
Lastly, using the enrichment edit dialog an enrichment can be deleted (from a dimension).
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 45/53
6 Customizing the ET
The current implementation of the ET is geared towards facilitating a usable platform for the
current end users, namely RBB and TKK. In its design and architecture however, although
not perfectly implemented as such yet, the ET is setup to enable the flexibility of extending
and configuring a platform for various possible television programs and specific players
working with LinkedTV technology.
The following sub-sections describe the different aspects of what potentially make the ET a
reusable toolkit for configuring and implementing program specific editor tools. To avoid
going into implementation details and since not everything has been implemented yet, the
following will mostly be on a conceptual level.
6.1 Open source
The basic prerequisite for being able to produce a tool that can be used by many different
users is to make it open source. Nowadays GitHub30 seems the most popular platform and
therefore the ET has been published there as well.
By advertising and offering technical support, LinkedTV aims to attract contributors, which
could both improve the code base and strengthen the relation with potential LinkedTV clients.
Moreover by offering the flexibility to adapt (see Section 6.3) the ET for using non-LinkedTV
services as well, there is a possibility to attract a wider audience of contributors, which can
also further improve the ET for LinkedTV and non-LinkedTV interested parties alike.
6.2 Configuration
Next to being open source, the currently implemented configuration options offer interested
parties by default the option of configuring the following:
- Chapter dimensions: in other words, which of LinkedTV’s enrichment services
should be used for what dimension.
- Information card templates: for particular programs, editors might need the
possibility to provide viewers with reoccurring pieces of information. In such cases it
is convenient for editors to configure templates for this information (see 5.5.1 and 3.2
for more details).
The following sub-section describes the details of how the dimension configuration works
and how developers could extend the ET with new services or interface components to
extend the number of possible configurable dimensions.
30 https://github.com/beeldengeluid/linkedtv-editortool
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 46/53
6.3 Extending code – adding dimensions
Besides being able to completely change the ET code or using the currently available
configuration to fulfill end user specific requirements, the architecture of the ET also
incorporates a plugin structure for extending and/or customizing the following two aspects:
- Dimension service
- Dimension user interface dialog component
Currently, the ET v2.0 has two different “dimension services” implemented, namely the
TVEnricher and the TVNewsEnricher, which can be configured as was shown in Figure 131.
Figure 30 shows how the dimension configuration is handled in the architecture.
Figure 30: Architecture – how the ET handles the dimension configuration
This figure presents an example configuration for a chapter dimension called ‘Opinion’ which
uses the TVEnricher service with the parameters “dimension = opinion”. Moreover the
31 Another available “service”, i.e. the “InformationCard”, is not really a service accessed via the ET
API, but is rather an ingrained functionality of the ET. To keep the dimension configuration consistent,
however, this option is configured via the “service” parameter.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 47/53
parameter ClientDialog32, i.e. the aforementioned user interface dialog component, is set to
“GenericEnrichmentDialog”, which will be explained further on.
First it is important to understand that each dimension service can support multiple
dimensions. The most straightforward example is the TVNewsEnricher that supports 5
dimensions, namely: “in-depth”, “opinion”, “timeline”, “other media” and “tweets”, meaning
that this service can be used for 5 different chapter dimensions. In Figure 30 the dimension
“opinion” is configured via the “params” setting. For additional illustration, Figure 31 shows
the RBB dimension configuration, which incorporates all of these 5 dimensions.
32 This parameter has not been implemented yet
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 48/53
Figure 31: JavaScript configuration of the RBB dimensions
Getting back to Figure 30, when a contributor to the ET is interested in implementing a
different dimension service, e.g., an API interfacing with a particular program database, this
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 49/53
can be done by adding a component to the ET back-end that communicates to this API and
translates its output to be compatible with the LinkedTV model33.
Finally, when eventually implemented, using the “ClientDialog” option, a developer can point
to a custom JavaScript module that takes care of showing an editing dialog in the same
manner as e.g. the dialog that was already implemented for fetching multiple enrichments
(see e.g., Figure 24). This last option offers developers a relatively flexible and convenient
way of adding interface components specifically designed for the end-user.
6.4 Extending code – outlook
After presenting several means to adapt the ET to suit the specific needs of particular end-
users, there are several points on which the ET can be improved as a valuable toolset for
dynamically fetching enrichments and annotating video using these enrichments. To name a
few:
- different configurable panels for editing chapters, including e.g. a timeline view
- overall timeline view showing chapters and annotations
- different configurable entity search boxes, instead of just the DBpedia search box
(see Figure 21)
33 The exact details of this can be found in the ET code.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 50/53
7 Conclusions & future work
This document reports on the work done related to the development and release of v2.0 of
the LinkedTV Editor Tool or ET. Initially, the work done for involving the end users at RBB
and AVROTROS, as well as the result of this work, i.e., the updated user requirements were
described in detail. Most notably, these new requirements include the major distinctive
feature of the ET v2.0, namely the ability to fetch enrichments on demand. The actual
implementation of this feature was achieved as a result of close collaboration with
EURECOM who extended their TVEnricher, TVNewsEnricher and EntityExpansion web
services to support on demand requests.
Following this, two newly conceived concepts, namely the information cards and chapter
dimensions were introduced. Both concepts emerged from user tests, conducted by CWI, on
the Linked News Demonstrator and showed potential in increasing the extensibility of the ET
for different use cases. Specifically: information cards allow editors to have more control on
exactly what information they approve of showing in the player. Subsequently, chapter
dimensions enable selected enrichments, including information cards, to be grouped (by
dimension) possibly narrowing the gap between what editors see in the ET and what is
shown in a LinkedTV player.
Based on the updated user requirements and the newly introduced concepts, the system
architecture was updated. This work included updating the back-end with new functionalities
for interacting with the aforementioned EURECOM services as well as translating the output
of these services to the ET data model. Moreover the back-end was extended to include
communication with the EntityProxy developed by CWI. In order to adapt to the new data
model and to increase the convenience for possible contributors of the ET’s open source
base in GitHub, the front-end of the ET was migrated from Backbone.js to AngularJS.
After describing the context and rationale of everything new in v2.0, the main body of this
document described the resulting functionalities of the ET, which include 1) the new home
page, 2) login mechanism, 3) selection of videos 4) curation of chapters, 5) addition of
information cards and finally 6) the addition of other enrichments
Finally for future purposes this report included a special section on the customizability and
extensibility of the ET with the aim of demonstrating its potential for supporting future use
cases as well as its aims to become a fully-fledged developer toolkit, directly contributing to
LinkedTV’s library of products.
7.1 Future work
Following the efforts of publishing of the ET code in somewhat reusable form in GitHub, the
future work will mostly entail improving the reusability and extensibility of the code in order to
offer interested parties a developer toolkit with which it is possible to install and build editor
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 51/53
tools that are tailored for specific program editing teams. To this direction the Netherlands
Institute for Sound and Vision already expressed interest in further developing the ET for
different projects. All further improvements from NISV will be directly updated in the current
ET GitHub project.
Finally for LinkedTV’s final round of evaluations, the ET is also planned to be evaluated by
end users. Besides evaluating the usability of the tool itself, the ET will also most likely be
used to evaluate the quality of the shot and chapter segmentation produced in WP1.
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 52/53
8 Bibliography
D1.3: LinkedTV annotation tool, first release Jaap Blom (S&V); Nicolas Patz (RBB)
D1.4: Visual, text and audio information
analysis for hypervideo, final release
Evlampios Apostolidis, Nikolaos Gkalelis,
Fotini Markatopoulou, Damianos
Galanopoulos, Eftichia Mavridaki, Christina
Papagiannopoulou, Vasileios Mezaris
(CERTH/ITI); Daniel Stein (FRAUNHOFER);
Milan Dojchinovski, Tomáš Kliegr, Milan
Šimůnek, Vaclav Zeman (UEP)
D2.2: Specification of lightweight metadata
models for multimedia annotation
José Luis Redonda Garcia, Raphaël Troncy,
Miroslav Vacura (EURECOM)
D2.6: LinkedTV Framework for Generating
Video Enrichments with Annotations
Tomáš Kliegr, Jan Bouchner, Barbora
Červenková, Milan Dojchinovski, Jaroslav
Kuchař (UEP);
José Luis Redonda Garcia, Raphaël Troncy
(EURECOM); Jan Thomsen (CONDAT);
Dorothea Tsatsou, Georgios Lazaridis,
Pantelis Ieronimakis, Vasileios Mezaris
(CERTH/ITI)
D3.5: Requirements Document LinkedTV
User Interfaces (Version 2)
Lilia Perez Romero, Lynda Hardman, Michiel
Hildebrand (CWI)
D3.7: LinkedTV User Interfaces Selected and
Refined
Michiel Hildebrand, Lilia Perez Romero
(CWI); Daniel Ockeloen, Pieter van Leeuwen
(Noterik)
D6.3: User trial results Katarina Stanoevska, Frederic Junker
(University of St. Gallen); Mar Rodriguez
(RBB); Lotte Belice Baltussen, Evelien Wolda
(The Netherlands Institute for Sound and
Vision); Mieke Leyssen (CWI)
LinkedTV annotation tool, final release D1.5
© LinkedTV Consortium, 2014 53/53
9 Glossary
Annotation A video annotation. Any piece of information/metadata assigned to a
certain part of a video or in other terms: media fragment.
Chapter A segment of a video encompassing a certain topic or scene (e.g.
such as a film scene) or item (e.g. in a news program)
Chapter dimension An annotation layer that has specific semantics defined in line with its
intended use for a related LinkedTV player.
Entity (Automatically detected) named entity
Information card A construct that allows program editors to conveniently curate
information, possibly derived from autogenerated entities, that is to be
shown to viewers
Named entity As defined in http://en.wikipedia.org/wiki/Named-
entity_recognition#Named_entity_types
Shot A segment of video composed by sequential frames captured
uninterruptedly from a single camera.
Segment A part of a video. Segments in LinkedTV are expressed as an
annotation. Currently there are two types of segmentations, namely
shot- and chapter segmentations.

More Related Content

LinkedTV Deliverable D1.5 The Editor Tool, final release

  • 1. Deliverable 1.5 LinkedTV annotation tool, final release Names of authors Jaap Blom Date of Deliverable October 29th 2014 Work Package 1: Intelligent hypervideo analysis LinkedTV Television Linked To The Web Integrated Project (IP) FP7-ICT-2011-7. Information and Communication Technologies Grant Agreement Number 287911
  • 2. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 2/53 Dissemination level1 PU Contractual date of delivery 30th September 2014 Actual date of delivery 29th October 2014 Deliverable number D1.5 Deliverable name LinkedTV annotation tool, final release File LinkedTV_D1.5.docx Nature Report Status & version Final, v1.0 Number of pages 53 WP contributing to the deliverable WP1, WP5, WP2 Task responsible The Netherlands Institute for Sound and Vision Other contributors RBB Author(s) Jaap Blom The Netherlands Institute for Sound and Vision Reviewer Nicolas Patz EC Project Officer Thomas Küpper Keywords video annotations, broadcast editors, interactive television, data curation, user interface Abstract (for dissemination) Editor tool v2.0; curation of automatically generated annotations; software. 1 • PU = Public • PP = Restricted to other program participants (including the Commission Services) • RE = Restricted to a group specified by the consortium (including the Commission Services) • CO = Confidential, only for members of the consortium (including the Commission Services))
  • 3. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 3/53 Table of contents 1 Introduction ...................................................................................... 5 1.1 Document overview............................................................................................... 6 1.2 Exclusions............................................................................................................. 6 1.3 Acknowledgements............................................................................................... 6 1.4 History of the document ........................................................................................ 7 1.5 Purpose of the Document...................................................................................... 7 2 User requirements............................................................................ 8 2.1 Involving the end-users ......................................................................................... 8 2.1.1 Tussen Kunst & Kitsch (AVROTROS) ..................................................... 8 2.1.2 RBB....................................................................................................... 11 2.2 Prioritized user requirements............................................................................... 13 3 Updated data model ....................................................................... 18 3.1 Chapter dimensions ............................................................................................ 18 3.2 Information cards ................................................................................................ 19 4 System architecture ....................................................................... 22 4.1 Changed dependencies with other WPs.............................................................. 22 4.1.1 WP2 new components........................................................................... 22 4.1.2 WP3 new component ............................................................................ 23 4.2 Changes to front-end........................................................................................... 24 4.3 Changes to the back-end .................................................................................... 24 5 Functionalities ................................................................................ 25 5.1 Home page & logging in...................................................................................... 25 5.2 Choosing a video ................................................................................................ 28 5.3 Curating chapters................................................................................................ 29 5.4 Chapter dimensions ............................................................................................ 33 5.5 Adding information cards..................................................................................... 34 5.5.1 Using templates..................................................................................... 36 5.5.2 Creating custom information cards ........................................................ 37 5.6 Adding enrichments on demand.......................................................................... 40
  • 4. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 4/53 5.6.1 Searching and adding new enrichments................................................ 41 6 Customizing the ET ........................................................................ 45 6.1 Open source ....................................................................................................... 45 6.2 Configuration....................................................................................................... 45 6.3 Extending code – adding dimensions.................................................................. 46 6.4 Extending code – outlook.................................................................................... 49 7 Conclusions & future work ............................................................ 50 7.1 Future work......................................................................................................... 50 8 Bibliography ................................................................................... 52 9 Glossary.......................................................................................... 53
  • 5. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 5/53 1 Introduction This document reports on the design and implementation of the final version of the editor tool (ET) v2.0, where its purpose is to serve the program editing teams of broadcasters that have adopted LinkedTV’s interactive television solution into their workflow. Two of these teams are currently represented in the LinkedTV project, namely the RBB team and the AVROTROS team (formerly known as AVRO). The main purpose of the ET is to provide a means to correct and curate automatically generated annotations and hyperlinks created by the audiovisual and textual analysis technologies developed in WP 1 and 2 of the LinkedTV project. Without the intervention of human editors to correct this data, there is a reasonable risk of exposing inappropriate, incorrect or irrelevant information to the viewers of a LinkedTV interactive broadcast. Compared to the previous version of the ET, the main new feature of v2.0 is the possibility to fetch enrichments, i.e., links to possibly relevant external information sources, on demand. This means that an editor now not only has the opportunity to choose from pregenerated enrichments, but also has the possibility to query and retrieve additional enrichments while using the ET. This new feature emerged from interviews with end-users, and after consultation with the LinkedTV consortium, new services were developed in WP2 to support this new approach. Moreover, resulting from user tests, conducted by the LinkedTV partner CWI, and several consortium discussions, two new concepts, namely information cards and chapter dimensions, were introduced into the ET. These concepts increase the reusability of the ET for different use cases and narrow the gap between what editors see in the ET and what the audience sees in the LinkedTV player. Finally, a modest highlight of v2.0 of the ET is that it also aims to be a development toolkit that enables programmers to configure and extend it with functionalities that are tailored for their player based on e.g., the multiscreen toolkit2. Note that this is currently a strategy that is not yet perfectly reflected on a technical level, but is already propagated on the ET’s new home page and GitHub3 page in order to notify interested parties. This report is a follow up of the deliverable "D1.3: LinkedTV annotation tool, first release" and mainly focuses on anything that is news compared to v1.0 of the ET. 2 Developed by Noterik in WP3 for the purpose of building interactive audiovisual applications 3 Collaborative working platform for hosting (open source) software: https://github.com/
  • 6. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 6/53 1.1 Document overview In order to update the user requirements, the direct end-users RBB and the AVROTROS participated in user trials and interviews in order to obtain feedback on the first version of the ET and new insights into the day-to-day workflow of editing teams. The setup and outcome of these trials as well as the following new user requirements are described in Section 2.The two newly introduced concepts, namely information cards and chapter dimensions are accounted for in Section 3. Due to the development of different new web services in WP2 and WP3, the architecture of the ET underwent several changes. Section 4 reports on these new services and the resulting new system architecture. The current features of the ET as well as updates to the interaction design are described in detail in Section 5. Section 6 describes how users of the open source ET GitHub repository can customize and configure the tool for different (interactive) television programs and their respective editing teams. Section 7 presents a summary of the main points described and gives a lookout on future plans. Section 8 presents the bibliography with related deliverables. Finally, Section 9 presents a glossary to enlighten possibly unfamiliar terminology. 1.2 Exclusions Unlike what was reported in D1.3, v2.0 of the ET does not include any specific feature with the purpose of influencing the input of WP4’s personalization systems. The main reason for this (see also Section 2) is that current users at RBB and AVROTROS did not express an interest in spending time on specifically curating content for different user groups, however the same users did express an interest in LinkedTV’s capabilities of automatic personalization based on curated data. 1.3 Acknowledgements CWI greatly contributed in sharing results of relevant user trials, defining methodologies and preparing (and attending) focus group sessions held at NISV and RBB. EURECOM and UEP put serious effort in adapting their enrichments services to support the on demand approach. Condat helped out in updating the LinkedTV API to ensure a better connection with the ET. RBB perfectly fulfilled the role of end user and provided NISV with useful comments throughout the development of the ET. The LinkedTV consortium as a whole made sure the ET, touching the work of most work packages, was included in the right discussions at the right time.
  • 7. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 7/53 1.4 History of the document Table 1: History of the document Date Version Name Comment 2014/22/04 V0.1 Jaap Blom, NISV Initial ToC + updated user requirements 2014/29/04 V0.2 Jaap Blom, NISV; Nicolas Patz, RBB Updated user requirements 2014/22/09 V0.3 Jaap Blom, NISV First draft after finishing software 2014/16/10 V0.4 Jaap Blom, NISV Further updates 2014/17/10 V0.5 Jaap Blom, NISV Further updates 2014/20/10 V0.6 Jaap Blom, NISV Further updates 2014/21/10 V0.7 Jaap Blom, NISV QA ready version 2014/27/10 V0.8 Jaap Blom, NISV; Lampis Apostolidis, CERTH Integrated feedback from WP1 leader 2014/27/10 V0.9 Jaap Blom, NISV; Nicolas Patz, RBB Integrated feedback from formal QA 2014/27/10 V0.95 Jaap Blom, NISV Final version 2014/29/10 V1.0 Jaap Blom, NISV; Lyndon Nixon, MODUL University Final version with integrated feedback from scientific coordinator 1.5 Purpose of the Document The purpose of this document is to inform the reader about the progress behind the creation of the final version of the ET.
  • 8. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 8/53 2 User requirements 2.1 Involving the end-users For this version of the ET, again4 end-users at RBB and AVROTROS were involved to provide feedback. This time, the aim was to obtain feedback on the following: • The day-to-day workflow of program editing teams • The usability of the ET v1.0. • The quality of the automatically detected entities and hyperlinks The first and most important point was to find out more about how program teams work and how feasible it is, or under what circumstances it would be feasible for them to incorporate the ET in their day-to-day workflow. The second point was addressed to obtain direct feedback on the usability of the first version of the ET in order to find out which features should be kept and which features should be dropped or improved. For the third point, CWI executed a test with a user group representing the general public, which provided useful insights5. Because of this and the limited availability of both the RBB editing staff and the staff of Tussen Kunst & Kitch (TKK), it was decided, at that time, not to go ahead with a practical test specific for program editors. Regardless of their limited time, both groups showed much interest in learning about LinkedTV and the ET. For future evaluations of the ET v2.0 (see Section 7.1), the limited availability of staff will be taken into account when creating the tests. For all the above, CWI greatly contributed in defining the evaluation methodologies and the preparation of materials for the focus group sessions, tests and interviews held. Subsequently RBB and NISV made sure to apply the defined methodologies as much as possible with their user groups. The following sub sections describe the setup and outcomes of the evaluation efforts for TKK cultural heritage show and RBB news show. 2.1.1 Tussen Kunst & Kitsch (AVROTROS) On the 27th of April, NISV and CWI held a focus group session with the program team of TKK on the premises of NISV. The session agenda was as follows: 4 See D1.3: LinkedTV annotation tool, first release 5 The results on this will be reported in WP3’s final deliverable D3.8: Design guideline document for concept-based presentations
  • 9. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 9/53 • Short presentation of LinkedTV and the processing workflow (the basic steps from analyzing video content, using the ET, and final interactive broadcast) • Presentation of the main functionalities of the ET • Presentation of user tests with the LinkedTV second screen application6 • Presentation of future plans of LinkedTV Throughout the presentations the members of the TKK team could ask questions and provide immediate feedback. During the presentations, the NISV and CWI hosts asked the TKK team questions such as: • whether the presented aspect/feature would fit their current workflow or could be otherwise adapted to be made to fit their workflow • what features are especially appreciated and what features are particularly disliked • whether (if applicable) a feature would target a specific part of their audience The following table lists the feedback that is applicable to the development of the ET. Feedback trigger Feedback Effect on ET/LinkedTV As a reaction to showing LinkedTV’s personalization approach. “Personalization should be done automatically, because there are not a lot of editors in the team. TKK only have the general public in mind as their audience, so all curation is done for this type of audience.” • The ET should not incorporate specific features for personalization. • The ET does not need to distinguish between target audiences. In reaction to the chapter editing functionality of the ET. “The start and end times of chapters are available during montage, so automatic chapter detection is not necessary if this information could be incorporated into the system.” “The short segments of the show are also important as it discusses relevant topics as well.” • For TKK the ET needs to be able to load chapter segmentations provided by the AVROTROS. • In case of automatic chapter detection, for TKK the short chapters are also relevant. In reaction to showing “We are wondering how many • The ET should highlight the 6 Reported in D6.3: User trial results
  • 10. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 10/53 Feedback trigger Feedback Effect on ET/LinkedTV the way an editor uses the ET to select useful automatically detected entities for creating annotations. entities are detected on average and should thus be inspected by the editors. We can’t spend much time on this.” “Seeing the location where the episode was filmed is not very interesting [as an entity], but other locations, such as where the art objects were found are.” most relevant (automatically detected) entities and/or hide irrelevant entities. In reaction to showing the way an editor uses the ET to select automatically detected enrichments/hyperlinks. “There are way too many enrichments showing in the ET.” “Key frames and a short title or keywords are not enough to judge whether a hyperlink is relevant or not.” “A great feature would be that the editor selects a number of entities and then the system would look for relevant hyperlinks. In the current situation there are just too many (irrelevant) links.” “Some entities are interesting for inside links, such as e.g. ‘gold’ [show other episodes with gold objects], but not for outside links [a Wikipedia page on ‘gold’ is not interesting.]” “It’s important to be able to find links that are based on current events.” “It’s important to be able to find links that are surprising to the editor.” • The ET should be able to reduce the amount of (irrelevant) enrichments, in order to reduce the information overload. • The ET should have a feature that allows a user to fetch enrichments on demand (based on a selection of one or more entities). In reaction to showing the way an editor adds annotations using the ET. “For other programs [at AVROTROS] there are boost factors for certain concepts to improve recommendation.” • The ET could be extended with a feature for adding boosting values to annotations in order to aid the WP4 recommendation. As general feedback on “Currently the editing team does not • Ideally the ET could fulfill the
  • 11. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 11/53 Feedback trigger Feedback Effect on ET/LinkedTV the ET as a whole. add any contextualization to TKK.” “TKK does not have an interactive website because there was never a need for it. But it has a potential. The internet editors publish the press release, but do not add more links. The program and top pieces are described. On social media (Facebook) they give some info to get people’s attention.” “TKK used to have lists of keywords for each chapter in the episode.” “There are high quality production photo’s available.” “The editor should be responsible for finding links in the ET. He/she should be able to select concepts and the system should look for (surprising) links.” “Overall: there is too much ‘noise’ and too many links.” role of simply being a convenient tool for finding contextual information for television programs. • The LinkedTV platform should be able to deal with different forms of (meta)data coming from the production phase of television programs. • The ET should focus on helping editors quickly find relevant (and surprising) links. Table 2: Feedback from the AVROTROS 2.1.2 RBB On the 5th of June 2014 two editors worked together with a colleague from the RBB LinkedTV team and discussed the ET in its current version. As a reminder of the motivation for showing this tool, first the end user GUI demo was presented again, then the news show was briefly analyzed in the sense that the attendants agreed on the chapters’ headlines and the editors had time to use the ET and make comments and give direct feedback. The ET itself was briefly presented and some strategic tips were issued in order to guide the editors, but the general approach was to let them use the tool and note any issue they may find as any tool that would need hours or days of training and coaching would not have a chance to be accepted into everyday workflows. The following table lists the feedback that is applicable to the development of the ET.
  • 12. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 12/53 Feedback trigger Feedback Effect on ET/LinkedTV In reaction to the chapter editing functionality of the ET. “The prefabricated chapters were not correctly marked. Beginnings and endings did not comply with the beginnings and endings of the reports and anchor presentations. So I had to create new chapters on my own which took about 25 minutes and would not have been affordable in every day’s business.” “The handling of the editor tool is not intuitive enough especially when it comes to naming the chapters. For example most of the new chapters didn’t even have a thumbnail. So I couldn’t be quite sure whether changes of naming of chapters were automatically saved or not.” • Automatic chapter detection should be improved to accelerate the editing process. • Chapters must carry a key frame thumbnail. In reaction to showing the way an editor uses the ET to select useful automatically detected entities for creating annotations “The prefabricated entities were not as useful as hoped. I suppose only a quarter made sense and was useful. So I had to view every single news report again, note catchwords and time codes to create new additional links. That process again took some time, at least 20 to 30 minutes per report.” • No Editor Tool issue. In reaction to showing the way an editor uses the ET to select automatically detected enrichments/hyperlinks “Creating additional links means in my opinion too many working steps which would even not be affordable in everyday business. It would be easier and helpful if mark-in and mark-out could be set right beneath the player while watching every single report to create correct time codes. It would also be timesaving if there was an input screen for naming and additional links right beneath the • Reduce number of steps for creating new annotations and/or links.
  • 13. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 13/53 Feedback trigger Feedback Effect on ET/LinkedTV player.” In reaction to showing the way an editor adds annotations using the ET. “In the beginning I was quite uncertain where to put a URL while creating additional links. Also the usage of the "add“-button seemed not that self-explaining to me. Trial and error was the only way to find out, which took again some time.” “Searching and naming of DBpedia- links was also not self-explaining because the input screen was grayed out and seemed deactivated to me. That’s why I also didn’t get the possibility of creating several additional links in one entity-screen.” • Clearer guidance and/or a manual. As general feedback on the ET as a whole. “As I could see umlauts can not be used but unfortunately there is no warning message if you mistakenly do so. So after having mistakenly used umlauts the complete input screen for additional links disappears after saving. So the editor has to do start from scratch.” “A preview would also be very helpful so that the editor can see the final product while working on it and possible errors.” • Enable use of Umlauts. • Enable preview templates. Table 3: Feedback from the RBB editing staff 2.2 Prioritized user requirements Table 4 lists the user requirements for v2.0 of the ET, while Table 5 reports the unfulfilled user requirements taken from D1.3 that haven’t made it in v2.0 either.
  • 14. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 14/53 ID Requirement Description Priority In v2.0 AUT1 User login. Users can log into the system using a user name and password. Must- have Yes AUT2 Authorization by user organization. Users can only access content of their own organization. Must- have Yes OC1 Content for public users. For public users there it should be possible to test the tool with a cc-licensed video. Must- have Yes HOM1 There should be an attractive home page with useful context information. The home page should explain LinkedTV and the purpose of the tool. Must- have Yes CH1* The creation and editing of chapters should be more user friendly. Nice-to- have Yes ENT1* Cluster entities of the same name, to enable quicker visual inspection of the entities. Currently, when detected multiple times, entities of the same name are displayed for each occurrence. Must- have Yes ENT2* Group entities of the same (DBpedia) type, to enable quicker visual inspection of all the entities. Currently entities of the same DBpedia class/type are not grouped together. Nice-to- have No ENT3* Remove the key frames shown for each entity. Currently each entity is displayed with a key frame of the corresponding frame in the video; this fills up the screen unnecessarily. Must- have Yes * These requirements all refer to updates of existing functionalities from v1.0. Requirements without an asterisk are completely new requirements.
  • 15. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 15/53 ID Requirement Description Priority In v2.0 REL1 The enrichments should include links to related material. Currently the enrichments only consist of links taken from the white list. Must- have Yes ENR1 External and internal links should be clearly distinguishable. It should be visually clear which links are internal and which are external. Must- have Yes ENR2* More recent links should be shown on top. Currently displayed enrichments are not ordered by date. Should- have Yes ENR3 It should be possible to generate enrichments (on the fly) based on the selection of one or multiple entities. Currently the system only offers enrichments that are based on each single automatically generated entity. In order to get more meaningful enrichments, an editor should be able to select one or more entities for generating enrichments on the fly. Should- have Yes SAV1* Publish the editor’s choices, tags in the LinkedTV platform. In v1.0 the save data was stored in a separate RDF graph. In v2.0 this data should be integrated in the overall LinkedTV graph7. Must- have Yes PRV1 Preview play-out of editor’s choices. A user can view a simplified version of the front-end’s play-out, based on the editor’s choices, by utilizing a video preview mode Should- have No ST1 Track statistics based on the user actions. Statistics related to the user’s behavior with respect to the acceptance, Should- have No 7 See D2.2 Specification of lightweight metadata models for multimedia annotation
  • 16. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 16/53 ID Requirement Description Priority In v2.0 rejection and modification of (automatically generated) annotations & hyperlinks are logged for various evaluation purposes. Table 4: User requirements classified with the MoSCoW method and the availability in v2.0 ID Requirement Description WL1 Edit items in user organization’s white list. A user can edit (create, update or remove) the (seed) URLs in the user organization’s white list. TPL1 Edit template based on the user’s organization. A user can edit the who-what-where template of the user’s organization. REC1 Add user tags to describe the video as a whole. A user can select tags from a predefined list to globally describe the main topics of the selected video. REC2 Add user tags to describe individual chapters. A user can select tags from a predefined list to globally describe the main topics of the selected chapter. Table 5: Unfulfilled user requirements from v1.0 that could not be realized it in v2.0 either Table 5 shows requirements that did not make it in v1.0 and won’t be implemented in v2.0 of the ET either. The reasons for this vary per requirement: • WL1: There is currently no functionality for this in the LinkedTV API. Implementing this in the ET therefore is not possible. • TPL1: The idea of program templates, e.g., one for RBB News and one for TKK, could not be practically worked out into actual features of the system. Moreover the concept of entity meta types, namely ‘who’, ‘what’ or ‘where’ was dropped in the ET v2.0 as it was no longer considered very relevant in combination with the new concept of chapter dimensions (see Section 3.1). • REC1 / REC2: This was implemented in an intermediate version of V1, but because of other more important issues, no model for storing and utilizing the tags in the platform has been worked out. Even though not all of the desired features have been implemented, all must-have requirements listed in Table 4 have been met. Moreover, because LinkedTV appreciates the
  • 17. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 17/53 need for specific requirements for different (types of) programs, the ET is available (open source) for any interested party to be further enhanced or customized (see Section 6).
  • 18. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 18/53 3 Updated data model Based on new insights, mainly following from the user trials in year 3, two new concepts were introduced to the ET data model, namely (chapter) dimensions and information cards. The following sections describe these concepts, which have considerably influenced the functionalities of the ET v2.0. 3.1 Chapter dimensions The idea of having chapter dimensions came from the need to semantically organize the annotations added to a video (chapter). This need became apparent after the development and user test with the Linked News demonstrator8 developed by CWI. In this demonstrator the available context information offered to the user, i.e., the video annotations, were all distributed along different “axes” or as they are now called “dimensions”. Each dimension in this demonstrator had a different intension of use e.g. • in depth: show related in depth articles • opinion: show related opinionated articles • timeline: show the related information in a timeline With these examples in mind, the idea came to include the notion of dimension into the LinkedTV model9. Using dimensions greatly increased the possibility of expressing use case specific semantics in a generic way: • each dimension should, by clearly defining its semantics, reflect how information added to it should be used in an application • there is no limit to the amount of dimensions that can be added, thus for each subtle user feature a new dimension could be added For the ET having dimensions as configurable entities in the software greatly improves the reusability and adaptability of the tool for different use cases (see Section 6). In the ET dimensions are currently configured as follows: • each program (RBB News, TKK show) needs to define one or more dimensions o Each dimension has:  an identifier  a name (to display in the UI, see Figure 15) 8 See D3.5: Requirements Document LinkedTV User Interfaces (Version 2) for the initial demonstrator and D3.7: LinkedTV User Interfaces Selected and Refined for the eventually developed companion application 9 See D2.2: Specification of lightweight metadata models for multimedia annotation
  • 19. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 19/53  a web service that handles the fetching of enrichments, currently the following (web) services are supported (see also Section 4): • TVEnricher • TVNewsEnricher • InformationCards10 For illustration, the actual configuration looks like this (in JavaScript): Figure 1: JavaScript configuration of the TKK dimensions In Section 6.3 the possibility to customize dimensions is addressed in more detail. 3.2 Information cards In the ET v1.0, the editor was not able to control the information that is shown for entities related to annotations. All an editor could do when creating and annotation was: 10 The informationCards service is not a separate web service, but rather functionality embedded in the ET (read more in Section 5.5)
  • 20. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 20/53 • assign a single entity to it (select an auto generated one or add a custom DBpedia URL) • assign multiple enrichments (select an auto generated one or add a custom URL) Figure 2: ET v1.0 showing the annotation edit dialog. The user could only add a single URL representing the (named) entity and add multiple enrichments URLs Only able to use the ET this way, the implementation of displaying the actual information related to the entity URL was up to the LinkedTV player11, which also could only do a dynamic lookup using the entity URL (usually a DBpedia or Wikipedia URL) to fetch the actual information. Besides needing control over the information shown to viewers, some topics simply cannot be expressed very well using single (named) entities, but actually require the creation of custom entities that could have multiple links to available named entities (expressed as links to e.g. DBpedia). One such example is the art object, which represents the main topic for the TKK videos and fits the aforementioned type of entity: • most likely there is no online authoritative resource describing the art object shown in the program 11 Or rather a LinkedTV player, using the platform API different player can be implemented
  • 21. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 21/53 • art objects consist of possibly multiple references to entities that do exist in authoritative resources such as DBpedia or the AAT So for entities such as art objects, like in TKK shows, or for entities that require a high level of curation, e.g., to ensure high objectivity, which is the case in RBB News, the concept of information cards was conceived. In the ET these information cards are presented as a set of key/value pairs where the key is the name of the intended property, e.g., “title”, “description”, “creator”, and the value is either a: • literal (a text string) • entity, which has: o label (a text string) o URI (in this version of the ET a URL to DBpedia) Since information cards could in theory consist of an infinite number of key value pairs, the ET incorporated a templating mechanism to enable configuring templates for editors, aiming to provide a structured way of adding information that can eventually be published to their LinkedTV player. Figure 3 shows the configuration of the art object information card template. Figure 3: ET configuration – information card template for the TKK art object How information cards and their templates are used is described in Section 5.5.
  • 22. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 22/53 4 System architecture The architecture described in D1.3 still applies for the most part in v2.0. The biggest changes in the architecture are: • the incorporation of several new web services developed within WP2 to support enrichments on demand • the incorporation of the EntityProxy developed in WP3, to support the dynamic inspection of related properties of automatically detected named entities • the migration of front-end code from Backbone.js12 to AngularJS13 The following sub sections describe the details of these changes. 4.1 Changed dependencies with other WPs Several new web services have been developed within LinkedTV, which enable the ET to allow users to trigger enrichments based on the selection of a number of entities. This section briefly lists these new services per work package they were developed in. For full details on these services, please consult the specific deliverables they were reported in (which are referenced in the following sub sections). 4.1.1 WP2 new components TVEnricher This service is developed by the LinkedTV partner EURECOM and exposes an API that allows the ET, or any other program, to request enrichments on demand, based on a query. In fact, the TVEnricher is a proxy service that passes received queries on to several other services developed in LinkedTV, namely: • IRAPI*14: a search engine on top of an index created from crawled web documents based on RBB’s and TKK’s white list • MediaCollector*: a service that can search through the items published during the past 7 days in 10 different social media platforms • LinkedTV SOLR index*: a service that stores information of related video fragments • Europeana API15: an API that enables to retrieve items from the Europeana collection The TVEnricher is configured for both RBB and TKK (see Section 6). 12 http://backbonejs.org/ 13 https://angularjs.org/ 14 * For more information on these LinkedTV services see D2.6: LinkedTV Framework for Generating Video Enrichments with Annotations. 15 http://pro.europeana.eu/api
  • 23. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 23/53 TVNewsEnricher From an ET perspective, the TVNewsEnricher basically fits the same purpose as the TVEnricher; the difference is that it is optimized for the news domain of the RBB use case. Instead of the services listed for the TVEnricher, the Google Custom Search Engine16 and the Twitter API17 are used. The result is an API that offers 5 different dimensions to query along: • In depth: news articles that offer in depth information on the specified subject • Other media: news articles from other media sources on the specified subject • Opinions: opinion pieces on the specified subject • Timeline: a timeline comprising of news articles on the specified subject • Twitter: related tweets of the past 7 days on the specified subject EntityExpansion The EntityExpansion service (see D2.6), developed by EURECOM, is in the current version not yet fully integrated, but is on the list of useful extensions for the RBB News configuration. Basically this service offers the on demand retrieval of additional (named) entities, specifically for the news domain. The idea is that rather than just analyzing the metadata (and subtitles and so on) of the news program itself, it analyses related web documents on the same news topic. This way there is also an opportunity to find entities that come from a broader context, e.g. including yesterday’s news, than just the news item itself. The inclusion of the EntityExpansion service is listed under future work in Section 7.1. 4.1.2 WP3 new component EntityProxy The EntityProxy18, developed by CWI, is a service that takes DBpedia and Wikipedia URLs as input and, based on the supplied DBpedia type, returns a list of related properties from DBpedia. In the ET this service is used to retrieve information for professional end users when they want to inspect the details of certain automatically detected entities (see Figure 22). Without this functionality the user could only judge the value of an entity by its label. So rather than only reading: “Jan Steen”, by clicking on this label a user can retrieve the additional information, that Jan Steen was a Dutch painter, and read a description of his life and anything else that would be typically displayed in an information box on a related Wikipedia page. 16 https://www.google.com/cse/ 17 https://dev.twitter.com/rest/public 18 http://linkedtv.project.cwi.nl/
  • 24. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 24/53 4.2 Changes to front-end As mentioned above it was decided to rebuild the front-end for v2.0. The reason for this is that with the addition of dimensions and information cards, the client side data model has changed enough for it to be worthwhile to just rebuild it rather than adapt the old code. Moreover the incorporation of the new on demand functionalities required a different approach in (asynchronously) loading data from different services. Another good reason to rebuild the front-end was to get closer to possible contributors for the ET GitHub, by choosing AngularJS instead of Backbone.js because of its rising popularity and inherent capabilities to create reusable components. Next to switching to AngularJS, using SASS19 (in combination with Compass20) further improved the possibility to create reusable interface components. SASS is a precompiler for CSS that enables to modularize style sheet code, making it more readable, reusable and extensible for different programmers. 4.3 Changes to the back-end To support the new on demand functionalities, all of the components mentioned in Section 4.1.1 and 4.1.2, namely the TVEnricher, TVNewsEnricher and EntityProxy were integrated into the existing Python Django21 backend. In this case integration means that the ET back- end handles the HTTP communication between these components ensuring that the user input is translated to valid (component specific) queries and response data is mapped to the ET data model. Due to ongoing changes and updates of the LinkedTV platform, it was decided to wait with the storage of the new data model to the platform. Instead, all of the editor’s efforts are currently stored in the ET itself. For this, a Redis22 store is used to store all data in JSON format. Listed under future work (see Section 7.1) is the task for storing all of the curated data in the LinkedTV platform in the form of RDF triples. 19 http://sass-lang.com/ 20 http://compass-style.org/ 21 https://www.djangoproject.com/ 22 http://redis.io/
  • 25. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 25/53 5 Functionalities As a result of the new model (see Section 3) and the new system architecture (see Section 4), and aiming to support the new on demand capabilities, the functionalities of v2.0 have changed in several ways.The following sub sections describe these new features in detail and incorporate screenshots for illustration. To increase comprehensibility, the order of the sub sections is based on the recommended workflow for editing a program, which is: • visit the home page • login (currently as a RBB or TKK user)23 • curate chapters • per chapter: fetch & curate enrichments along different dimensions For better understanding, it is recommended to try out the described functionalities using the free trial video in the ET. To do this, open a Chrome24 browser and navigate to http://editortoolv2.linkedtv.eu/trial. Chrome is needed as it supports the video format currently used by the HTML5 player. 5.1 Home page & logging in The ET home page can be reached at http://editortoolv2.linkedtv.eu, and informs the visitor about the following: • What is the ET? • What is LinkedTV? • Try out the ET yourselves! • The ET is open source. Please visit us on GitHub! • Who are the people behind LinkedTV? • Special thanks to our external partners The home page is designed to attract interested parties in finding out more about LinkedTV. Also visitors are given the chance to try out the tool themselves, ultimately hoping that they will be interested contributing to the ET and/or using LinkedTV technology. The video used for the trial is creative commons25 licensed and comes from Open Beelden26, which is a site hosted by NISV. Figures 4-7 show the different sections of the home page. 23 It is also possible to skip login and go to a free trial page (see Figure 5) 24 http://www.google.nl/intl/nl/chrome/browser/ 25 http://en.wikipedia.org/wiki/Creative_Commons 26 http://openbeelden.nl/
  • 26. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 26/53 Figure 4: Home page – top section informing about the ET The first section informs the visitor about the overall capabilities and purpose of the ET. Figure 5: Home page – link to the trial page and link to the ET on GitHub The second section informs the reader about the trial version of the ET. The third section directs the visitor to the ET GitHub page and informs the reader about the possibilities of customizing the ET for specific use cases.
  • 27. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 27/53 Figure 6: Home page – link to project website and partner credits The fourth section presents a link to the LinkedTV project website. Figure 7: Home page – special thanks section The final two sections show the logos of consortium partners and affiliated partners.
  • 28. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 28/53 For accessing the ET, known users, currently being RBB and TKK, have been authorized to login to the system so they can curate their videos. By clicking “login” in the navigation bar at the top of the page the user is directed to the login page: Figure 8: Login screen After successfully logging in, the user is directed to the video selection page, which is described in the next section. 5.2 Choosing a video The following image illustrates the video selection screen, which shows all of the videos that were successfully processed by the LinkedTV platform for the organization the user is a member of. Figure 9: Video selection page showing a thumbnail and title of each video
  • 29. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 29/53 By clicking on one of the video thumbnails, the user is directed to the editing page of the selected video. The following sections describe the different functionalities of this page. 5.3 Curating chapters The LinkedTV platform offers several means of automatically identifying the different chapters of a video (see Section 3 of D1.4: Visual, text and audio information analysis for hypervideo, final release). In most conceivable cases, depending on each television program, a chapter is a part of the program that is about a single topic. For RBB for instance each news item is a separate chapter; for TKK the main chapters are about a single art object. Before a user starts adding annotations, linking different kinds of enrichments (see Sections 5.5 and 5.6), the possibly available automatically generated chapters, need to be edited in order to make sure they are correct. Figure 10: Chapter panel – showing autogenerated chapters only Figure 10 shows the video chapters of a TKK program that has not yet been edited; the cog icon below each chapter thumbnail indicates that a chapter was automatically generated and not edited by a human (curated). In this example it is clear that the chapters were created by the chapter segmentation algorithm, which relies on the TKK bumper re-detection27, as the TKK logo can be seen in each (except the first) thumbnail, which always reflects the starting frame of the chapter. For correctly editing chapters, the user is recommended to: 27 CERTH/ITI implemented this algorithm for the TKK chapter segmentation as each main chapter in TKK starts with showing the TKK logo (see also D1.4 in the Bibliography).
  • 30. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 30/53 • create proper titles for each chapter • make sure to check the boundaries of each chapter are correct • add possibly missing chapters if necessary • remove possibly unnecessary chapters In order to edit the predefined chapters a user can double click on one of the thumbnails to open the following dialog: Figure 11: Chapter edit dialog – the shot selection is hidden When clicking the “Select shots” bar the shot selection is opened:
  • 31. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 31/53 Figure 12: Chapter edit dialog – showing shot selection; shots included in the chapter are highlighted The edit dialog can be used to change the title as well as the start and end times (in milliseconds) of the chapter. As directly entering the amount of milliseconds is not very convenient, the user can select a starting and ending shot by using the shot selection panel. When doing so, the amount of milliseconds is automatically set in the start and end fields. If after saving the start and end is not perfect, the user can fine-tune them by adjusting the values in the corresponding fields. When a chapter has been edited and saved, the cog icon in the chapter panel is replaced by a ‘human’ icon, indicating that the chapter has been curated (see Figure 13 below).
  • 32. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 32/53 Figure 13: Chapter panel – showing two curated chapters When pressing the delete button (at the bottom, see Figure 12) a chapter can be deleted. Although removed from the view, this will not affect automatically generated chapters to be removed from the LinkedTV platform. When reloading the page, all automatically generated chapters are shown on the page again. To avoid confusing the user, the filter button in the top right of the chapter panel can be used to only show the chapters that were curated (see Figure 14) Figure 14: Chapter panel – showing curated chapters only (filter activated) Finally by pressing the “New chapter” icon a new chapter can be added. The same screen as Figure 11 is shown and after saving, the chapter is simply added to the list of chapters. After a user finishes the editing of the automatically generated chapters, it is time to start adding annotations to each one of them. In this version of the ET there are currently two kinds of information that can be attached to annotations • information cards • enrichments
  • 33. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 33/53 Section 5.5 describes everything there is to know about working with information cards and Section 5.6 delves into how to work with all the other types of enrichments. However before going into detail about the actual addition of annotations it is necessary to explain about how these annotations are all going to be related to certain dimensions per chapter. The following section describes how the user can interact with configured chapter dimensions. 5.4 Chapter dimensions As described in Section 3.1, each chapter has one or more dimensions configured for adding enrichments to. The result of such a configuration, in this case for TKK, is reflected in the annotation panel when navigating to a chapter (by selecting one in the chapter panel as shown in Figure 10): Figure 15: Annotation panel – showing the dimensions configured for TKK
  • 34. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 34/53 If enrichments already have been added this panel looks like the following: Figure 16: Annotation panel – showing two dimensions with saved annotations After curating chapters, adding enrichments, in either the form of information cards or ‘regular’ enrichments, is one of the main tasks of the editor. The following two sections describe how these different kinds of enrichments can be added to each dimension. 5.5 Adding information cards Following Section 3.2, Figure 1 shows that the dimension named ‘Art object’ is configured to be an information card dimension. The effect of this is that when double clicking the ‘New’ button in the ‘Art object’ dimension (see Figure 16) opens an empty information card edit dialog (see Figure 17). When double clicking an existing information card the same screen is opened, but with all of the saved properties filled out in the form at the top (see Figure 18).
  • 35. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 35/53 Figure 17: Information card edit dialog – showing an empty art object template Figure 18: Information card edit dialog – showing an information card being edited
  • 36. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 36/53 Figure 18 shows the information card dialog partly filled out and with all possible elements showing: • Edit form: here the template can be selected and the form can be filled out • Shot selection panel: here a shot can be selected to be used as poster image for the information card • Entity selection panel: for inspecting the properties of automatically generated entities • Entity details panel: showing the properties of the selected entity in the entity selection panel Before explaining the functionalities of each of these parts in detail, the editor currently has two different approaches to choose from: • use one of the preconfigured templates to fill out • manually create an information card Both approaches as well as the purpose of the aforementioned panels are described in the following two sections. 5.5.1 Using templates Using a template is the most straightforward way of adding an information card, following the steps described below. • The user fills in the pregenerated form • The user selects a shot using the shot selection panel • The user saves the card For choosing a template, the user needs to make sure to select one of the configured templates from the pull down list as shown in Figure 17. When done so, automatically all of the key value pairs configured (see Section 3.2) for that template are displayed in a form. For each configured literal, e.g., ‘label’ and ‘description’, the form provides a single text field. For each configured entity, e.g., ‘creator’ or ‘period’, the form displays a DBpedia search box (see Figure 18) and a separate text field to be able to edit the label of retrieved DBpedia entities. Filling out the form The text fields are just plain input fields where the user can enter free text. Using the DBpedia search boxes, the user must choose a DBpedia concept in order to properly fill out the field. After selection of an entity, the user is allowed to change the label of the retrieved entity; the underlying DBpedia URI is not displayed to the user.
  • 37. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 37/53 Selecting a shot Figure 19: Information card edit dialog - shot selection panel Using the shot selection panel as shown in Figure 19, the user can select one of the automatically detected shots to be used as a poster image for the information card. After filling in the form, and choosing a shot, the user can press ‘Save’ at the bottom of the dialog, to save the information card. 5.5.2 Creating custom information cards If there is no suitable template configured, the user can add custom information cards based on either: • an automatically detected entity • an entity found using the DBpedia search box For this, the user initially makes sure to deselect the “Use template” checkbox at the top of the dialog:
  • 38. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 38/53 Figure 20: Information card edit dialog – creating a custom information card Next, the user can select an entity by either selecting one from the “Detected entities” list or search for one using the DBpedia search box. Figure 21: Information card panel – the entity ‘Hilversum’ was found using the DBpedia search box When a user selects an entity from either the list (see Figure 20) or the one displayed after using the DBpedia search box (see Figure 21), CWI’s EntityProxy (see Section 4.1.2) is called and fetches related properties by analyzing (if any) the entity’s related DBpedia URL28. 28 The EntityProxy also accepts Wikipedia links, which are transformed into DBpedia URLs. This transformation however does not always work, as Wikipedia links do not always have direct DBpedia counterparts.
  • 39. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 39/53 When successful the following information appears at the bottom of the screen (“Fetched properties”). Figure 22: Information card edit dialog - showing the data fetched from the EntityProxy Using the displayed information the user can decide whether the selected entity is indeed the one needed for this information card. If so the displayed properties can be attached to the card by clicking on the “Use as card” button. After this the “Fetched properties” are used to populate the form, representing the properties of the information card, at the top of the screen. Figure 23: Information card edit dialog – after setting an entity as the basis for the information card
  • 40. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 40/53 Finally the user can edit the copied information to make sure it is appropriate for the eventual viewers of the program. In case the user does not appreciate the poster image copied from the selected entity, it is possible to select one by choosing a shot from the shot selection panel in the same manner as can be done when using the template (see Figure 19). After finishing the editing the user can save by clicking the save button, after which, the dialog is closed and the user is returned to the main screen. 5.6 Adding enrichments on demand In the ET v1.0, the user could only select automatically detected enrichments from a static precalculated list (see D1.3). Moreover each of these enrichments was found based on a single automatically generated entity each, often being irrelevant or too unspecific. After having tested this version of the ET with the TTK and RBB program teams the outcome was that they all thought that going through long lists of enrichments simply is too time consuming. Instead it was suggested that it would be very useful to be able to pick one or more automatically detected entities and use these as input for fetching enrichments on demand. This way the problem of solely having enrichments based on single entities is solved. Moreover, since the user requests the enrichments on demand, it is instantly clear on what basis each enrichment is found. To support this new on demand approach EURECOM had to work on extending their enrichment tools, namely the TVEnricher and the TVNewsEnricher (see Section 4.1.1), with APIs that could be accessed by the ET in the following way: • Input: a query • Output: a set of hyperlinks with the following properties o Source: URL or name describing the source (website or system) o Label: describing the hyperlink o Poster: (if available) a relevant thumbnail Unlike dimensions configured with the information card functionality (see previous section), dimensions configured with services supporting the on demand approach all use two dialog screens to fetch and edit enrichments: • searching enrichments on demand; adding/deleting multiple enrichments • editing/deleting single enrichments The reason for this difference in respect to the information card dialog is that for the on demand approach, many enrichments can be found and added at the same time, resulting in a different interaction design. The detailed functionalities of the aforementioned dialogs are described in the following two sections.
  • 41. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 41/53 5.6.1 Searching and adding new enrichments When no enrichments have been added yet, the user can double-click on the ‘Search / Add’ (see Figure 16) button to open the search enrichment dialog (see an example in Figure 24 below). Figure 24: Search enrichment dialog – before searching for enrichments The idea behind this dialog is the possibility to use either of the following to find related enrichments: • Curated enrichments: these are all the information cards and enrichments that are already saved in the selected chapter • Expanded entities: (only RBB news) these are entities that were fetched on demand using the EntityExpansion service. However, this has not been implemented yet. • Autogenerated entities: these are the automatically detected entities Each of these three sources can be inspected using the tabs displayed below “Choose entities”. If the user does not find any useful entities, it is always possible to enter search keywords in the search field at the bottom (see Figure 25). After choosing entities or entering a search query, the user can trigger the fetching of enrichments by pressing ‘Find links’. While searching the text of this button will change into ‘Loading…’ Figure 25: Search enrichment dialog – while searching for enrichments
  • 42. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 42/53 After searching is done, the enrichments are displayed at the bottom of the page, as illustrated in Figures 26 and 27 below. Figure 26: Search enrichment dialog – after finding enrichments in the ‘background information dimension’ Figure 27: Search enrichment dialog – after finding enrichments in the ‘related fragments’ dimension
  • 43. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 43/53 Depending on which dimension, the results show different information29, but can all be added and filtered in the same way. By simply double-clicking on one of the results it is added to the list of saved enrichments at the top (see Figure 28): Figure 28: Search enrichment dialog – after selecting an enrichment from the found enrichments. When selecting one or more enrichments in this way and subsequently pressing the ‘Save’ button at the bottom of the dialog, the enrichments are saved to the selected dimension within the currently active chapter. Since this dialog is already crowded with different user interface elements, the enrichment can be further edited in a separate dialog, which can be opened by double clicking on a saved enrichment as shown in Figure 16. The dialog can be used as follows: 29 The TVEnricher and TVNewsEnricher themselves support multiple dimensions
  • 44. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 44/53 Figure 29: Enrichment edit dialog – editing a newly saved enrichment Next to editing the three most basic properties of an enrichment, namely the URI, label and poster, a user can also determine the start and end time when this enrichment should be displayed within the active chapter. Similar to the chapter edit dialog (see Figure 12), the (rough) boundaries of the selection can be conveniently set by choosing two shots from the shot selection panel. Even though none of the LinkedTV players so far utilizes this information, this functionality adds much to the reusability of the ET as it can support use cases where editors need to be able to control the exact moment a piece of information is offered to viewers. Lastly, using the enrichment edit dialog an enrichment can be deleted (from a dimension).
  • 45. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 45/53 6 Customizing the ET The current implementation of the ET is geared towards facilitating a usable platform for the current end users, namely RBB and TKK. In its design and architecture however, although not perfectly implemented as such yet, the ET is setup to enable the flexibility of extending and configuring a platform for various possible television programs and specific players working with LinkedTV technology. The following sub-sections describe the different aspects of what potentially make the ET a reusable toolkit for configuring and implementing program specific editor tools. To avoid going into implementation details and since not everything has been implemented yet, the following will mostly be on a conceptual level. 6.1 Open source The basic prerequisite for being able to produce a tool that can be used by many different users is to make it open source. Nowadays GitHub30 seems the most popular platform and therefore the ET has been published there as well. By advertising and offering technical support, LinkedTV aims to attract contributors, which could both improve the code base and strengthen the relation with potential LinkedTV clients. Moreover by offering the flexibility to adapt (see Section 6.3) the ET for using non-LinkedTV services as well, there is a possibility to attract a wider audience of contributors, which can also further improve the ET for LinkedTV and non-LinkedTV interested parties alike. 6.2 Configuration Next to being open source, the currently implemented configuration options offer interested parties by default the option of configuring the following: - Chapter dimensions: in other words, which of LinkedTV’s enrichment services should be used for what dimension. - Information card templates: for particular programs, editors might need the possibility to provide viewers with reoccurring pieces of information. In such cases it is convenient for editors to configure templates for this information (see 5.5.1 and 3.2 for more details). The following sub-section describes the details of how the dimension configuration works and how developers could extend the ET with new services or interface components to extend the number of possible configurable dimensions. 30 https://github.com/beeldengeluid/linkedtv-editortool
  • 46. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 46/53 6.3 Extending code – adding dimensions Besides being able to completely change the ET code or using the currently available configuration to fulfill end user specific requirements, the architecture of the ET also incorporates a plugin structure for extending and/or customizing the following two aspects: - Dimension service - Dimension user interface dialog component Currently, the ET v2.0 has two different “dimension services” implemented, namely the TVEnricher and the TVNewsEnricher, which can be configured as was shown in Figure 131. Figure 30 shows how the dimension configuration is handled in the architecture. Figure 30: Architecture – how the ET handles the dimension configuration This figure presents an example configuration for a chapter dimension called ‘Opinion’ which uses the TVEnricher service with the parameters “dimension = opinion”. Moreover the 31 Another available “service”, i.e. the “InformationCard”, is not really a service accessed via the ET API, but is rather an ingrained functionality of the ET. To keep the dimension configuration consistent, however, this option is configured via the “service” parameter.
  • 47. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 47/53 parameter ClientDialog32, i.e. the aforementioned user interface dialog component, is set to “GenericEnrichmentDialog”, which will be explained further on. First it is important to understand that each dimension service can support multiple dimensions. The most straightforward example is the TVNewsEnricher that supports 5 dimensions, namely: “in-depth”, “opinion”, “timeline”, “other media” and “tweets”, meaning that this service can be used for 5 different chapter dimensions. In Figure 30 the dimension “opinion” is configured via the “params” setting. For additional illustration, Figure 31 shows the RBB dimension configuration, which incorporates all of these 5 dimensions. 32 This parameter has not been implemented yet
  • 48. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 48/53 Figure 31: JavaScript configuration of the RBB dimensions Getting back to Figure 30, when a contributor to the ET is interested in implementing a different dimension service, e.g., an API interfacing with a particular program database, this
  • 49. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 49/53 can be done by adding a component to the ET back-end that communicates to this API and translates its output to be compatible with the LinkedTV model33. Finally, when eventually implemented, using the “ClientDialog” option, a developer can point to a custom JavaScript module that takes care of showing an editing dialog in the same manner as e.g. the dialog that was already implemented for fetching multiple enrichments (see e.g., Figure 24). This last option offers developers a relatively flexible and convenient way of adding interface components specifically designed for the end-user. 6.4 Extending code – outlook After presenting several means to adapt the ET to suit the specific needs of particular end- users, there are several points on which the ET can be improved as a valuable toolset for dynamically fetching enrichments and annotating video using these enrichments. To name a few: - different configurable panels for editing chapters, including e.g. a timeline view - overall timeline view showing chapters and annotations - different configurable entity search boxes, instead of just the DBpedia search box (see Figure 21) 33 The exact details of this can be found in the ET code.
  • 50. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 50/53 7 Conclusions & future work This document reports on the work done related to the development and release of v2.0 of the LinkedTV Editor Tool or ET. Initially, the work done for involving the end users at RBB and AVROTROS, as well as the result of this work, i.e., the updated user requirements were described in detail. Most notably, these new requirements include the major distinctive feature of the ET v2.0, namely the ability to fetch enrichments on demand. The actual implementation of this feature was achieved as a result of close collaboration with EURECOM who extended their TVEnricher, TVNewsEnricher and EntityExpansion web services to support on demand requests. Following this, two newly conceived concepts, namely the information cards and chapter dimensions were introduced. Both concepts emerged from user tests, conducted by CWI, on the Linked News Demonstrator and showed potential in increasing the extensibility of the ET for different use cases. Specifically: information cards allow editors to have more control on exactly what information they approve of showing in the player. Subsequently, chapter dimensions enable selected enrichments, including information cards, to be grouped (by dimension) possibly narrowing the gap between what editors see in the ET and what is shown in a LinkedTV player. Based on the updated user requirements and the newly introduced concepts, the system architecture was updated. This work included updating the back-end with new functionalities for interacting with the aforementioned EURECOM services as well as translating the output of these services to the ET data model. Moreover the back-end was extended to include communication with the EntityProxy developed by CWI. In order to adapt to the new data model and to increase the convenience for possible contributors of the ET’s open source base in GitHub, the front-end of the ET was migrated from Backbone.js to AngularJS. After describing the context and rationale of everything new in v2.0, the main body of this document described the resulting functionalities of the ET, which include 1) the new home page, 2) login mechanism, 3) selection of videos 4) curation of chapters, 5) addition of information cards and finally 6) the addition of other enrichments Finally for future purposes this report included a special section on the customizability and extensibility of the ET with the aim of demonstrating its potential for supporting future use cases as well as its aims to become a fully-fledged developer toolkit, directly contributing to LinkedTV’s library of products. 7.1 Future work Following the efforts of publishing of the ET code in somewhat reusable form in GitHub, the future work will mostly entail improving the reusability and extensibility of the code in order to offer interested parties a developer toolkit with which it is possible to install and build editor
  • 51. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 51/53 tools that are tailored for specific program editing teams. To this direction the Netherlands Institute for Sound and Vision already expressed interest in further developing the ET for different projects. All further improvements from NISV will be directly updated in the current ET GitHub project. Finally for LinkedTV’s final round of evaluations, the ET is also planned to be evaluated by end users. Besides evaluating the usability of the tool itself, the ET will also most likely be used to evaluate the quality of the shot and chapter segmentation produced in WP1.
  • 52. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 52/53 8 Bibliography D1.3: LinkedTV annotation tool, first release Jaap Blom (S&V); Nicolas Patz (RBB) D1.4: Visual, text and audio information analysis for hypervideo, final release Evlampios Apostolidis, Nikolaos Gkalelis, Fotini Markatopoulou, Damianos Galanopoulos, Eftichia Mavridaki, Christina Papagiannopoulou, Vasileios Mezaris (CERTH/ITI); Daniel Stein (FRAUNHOFER); Milan Dojchinovski, Tomáš Kliegr, Milan Šimůnek, Vaclav Zeman (UEP) D2.2: Specification of lightweight metadata models for multimedia annotation José Luis Redonda Garcia, Raphaël Troncy, Miroslav Vacura (EURECOM) D2.6: LinkedTV Framework for Generating Video Enrichments with Annotations Tomáš Kliegr, Jan Bouchner, Barbora Červenková, Milan Dojchinovski, Jaroslav Kuchař (UEP); José Luis Redonda Garcia, Raphaël Troncy (EURECOM); Jan Thomsen (CONDAT); Dorothea Tsatsou, Georgios Lazaridis, Pantelis Ieronimakis, Vasileios Mezaris (CERTH/ITI) D3.5: Requirements Document LinkedTV User Interfaces (Version 2) Lilia Perez Romero, Lynda Hardman, Michiel Hildebrand (CWI) D3.7: LinkedTV User Interfaces Selected and Refined Michiel Hildebrand, Lilia Perez Romero (CWI); Daniel Ockeloen, Pieter van Leeuwen (Noterik) D6.3: User trial results Katarina Stanoevska, Frederic Junker (University of St. Gallen); Mar Rodriguez (RBB); Lotte Belice Baltussen, Evelien Wolda (The Netherlands Institute for Sound and Vision); Mieke Leyssen (CWI)
  • 53. LinkedTV annotation tool, final release D1.5 © LinkedTV Consortium, 2014 53/53 9 Glossary Annotation A video annotation. Any piece of information/metadata assigned to a certain part of a video or in other terms: media fragment. Chapter A segment of a video encompassing a certain topic or scene (e.g. such as a film scene) or item (e.g. in a news program) Chapter dimension An annotation layer that has specific semantics defined in line with its intended use for a related LinkedTV player. Entity (Automatically detected) named entity Information card A construct that allows program editors to conveniently curate information, possibly derived from autogenerated entities, that is to be shown to viewers Named entity As defined in http://en.wikipedia.org/wiki/Named- entity_recognition#Named_entity_types Shot A segment of video composed by sequential frames captured uninterruptedly from a single camera. Segment A part of a video. Segments in LinkedTV are expressed as an annotation. Currently there are two types of segmentations, namely shot- and chapter segmentations.