Chung-Jui LAI - Polarization of Political Opinion by News Media
- 2. Achievements
Education
• 2-month(2019) research in College of
Communication and Information, Florida State
University,Tallahassee, FL, USA
• A white-hat at NCCST (National Center of
Cyber Security technology) in 2017,2018.
• Sub-Lieutenant, police of information
management,
Wanhua Precinct,Taipei City Police Department
Department of Information
Management, College of Police
Science and Technology, Central
Police University(CPU),Taiwan
Work experience
• Paper presentation: 'A Framework for SQL Injection Investigations:
Detection, Investigation, and Forensics,’ The 2018 IEEE International
Conference on Systems, Man, and Cybernetics (SMC 2018),Miyazaki,
Japan, pp. 2834-2839,Oct. 7-10, 2018. (EI) (ISBN:978-1-5386-6649-4)
• Paper Acceptance: ' “On the left side, there’s nothing right. On the right
side, there’s nothing left:” Polarization of Political Opinion by News
Media, ’ iConference 2020.
- 13. Soften the
opposite
thought
▪ Dawn of the ordinary person
▪ One plays multiple roles. (e.g. students, victims, retired teacher, famous teacher)
▪ No ossified thinking. → close to people’ thoughts (sense of participation)
▪ Make good use of public opinions and make them good for specific group.
▪ Don’t emphasize the profession.
Pretend
Source: Google Trends
The search percentage of each mayors
(2018/1~2019/12)
Han Kuo-yu
Ko Wen-je
- 14. ▪ Fake accounts
▪ Transfer the battlefield of public opinions
▪ Push articles routinely by program
▪ The economy of digital advertisement (1/10)
Assemble!
Traditional
media
Social
media
Source: DMA (Taiwan Digital Media and Marketing Association)
- 18. ▪ The legalization of same-sex marriage
▪ 2014 Taiwan mayor election(Computational Propaganda in Taiwan: Where Digital
Democracy Meets Automated Autocracy)
▪ AI technology
▪ Text mining
▪ Sentimental analysis
▪ Personal texts
▪ The smuggling of cigarette of National Security Bureau in 2019.7
▪ The protest in Hong Kong now
▪ The spy case of Wang Li-qiang
▪ The murder case of Chen Tong-jia
▪ The mansion case of Han Kuo-yu (claimed for the civilian status)
▪ Recheck the old cases and find something illegal to against opposite
- 19. ▪Super Typhoon Jebi destroyed the Kansai airport, Japan
▪Some information claimed that the other government sent the bus to
pick up its national but the Taiwan Representative Office at Osaka did
nothing.
▪The account, “idcc,” emerged at the website, “PTT,” and start to attack
the representative.
▪The representative couldn’t afford the pressure and suicided after the
tremendous public opinions targeting him.
▪The account was manipulated by someone and the prosecutor found
that one of the IP Address was located at the Legislature Yuan.
- 20. The owner of BBS announced that they stopped to receive any account apply after 2018.9.15
The suicide case of Taiwan (2018.9.14)
Taiwan election(2018.11.14)
Freedom of
speech
Information
warfare
Cyber
warrior
Media
literacy
Media
framing
Filter
bubble
- 22. ▪ Introduction
▪ Political Opinion Mining on Presidential Election
▪ Sentimental Analysis
▪ Study Framework
▪ Data Collection
▪ Data Filtering & Select Criteria
▪ Data Analysis (4 models)
▪ Conclusion
- 24. The evaluation of
social media.
(e.g. Facebook, WhatsApp,
Instagram,Weibo)
Popularity
Massive usage
(posting, sharing)
The advantage of
social media.
Immediately
Widespread
The disadvantage of
social media.
Preconception
Disinformation content
(partial opinion)
- 26. LIWC (Linguistic Inquiry Word Count) is a
software that can analyze the percentage of
variables and reflect different emotions,
thinking styles etc. within the text.
Why using LIWC?
• The power of text analysis.
• The dictionary of LIWC contains almost 6,400 words.
• The clear classification of LIWC dimensions.(for example)
1.Affect Words (e.g. positive emotion, negative emotion)
2.Social Words (e.g. family, friends, male and female
referents)
3.Time Orientation (e.g. post, present, future focus)
4.Informal Speech (e.g. swear words)
- 29. DATA COLLECTION
with PYTHON CODE
▪ Twitter Developer for analyzing
tweets.
▪ Using Twitter Stream API,
“tweepy,” to collect the tweets
within Twitter.
▪ Stream API can catch the
prompt tweets that contain the
keyword we select.Keyword:Trump
- 30. ▪ Date : 2019-09-11 14:10:10
▪ User_name : @BBCWorld
▪ User_followers : 25751623
▪ User_location : London, UK
▪ Text :“Today our nation honours and
mourns the nearly 3,000 lives that
were stolen from us” President
Trump pays tribute to the victims
of the 9/11 attacks in US
Tweets Retweets Total
Total 211,249
(25.42%)
619,764
(74.58%)
831,013
(100%)
The example of the tweet & the collection amount
- 31. Unformatted data
sorted and classified
was removed
Original tweets
were separated
from retweets
News media’s
account names
were sorted to
extract the tweets
The dataset was
divided into three
categories of news
media agencies
The remaining tweets and
retweets of non-relevant
accounts were removed
Left wing
Right wing
Central
- 32. Category Tweets Example
Left-Wing 850 (63.24%)
Alternet, CNN Opinion, Democracy Now,The Daily Beast,The Huffington Post,The
Intercept, Jacobin,Mother Jones, MSNBC,The NewYorker,The NewYork Times
Opinion,The Nation, Slate,Vox, Mashable, ABC,The Atlantic, BuzzFeed News, CBC,
CNN Online News,The Economist,The Guardian, NBC,The NewYork Time Online
News, NPR Opinion, Politico,TIME,The Washington Post, CBS,The Daily Show,
Newsweek,VanityFair
Right-Wing 158 (11.76%)
Fox News Online News, Reason, The Wall Street Journal Opinion, Examiner,The
Washington times,The American Spectator, Breitbart,The Blaze, CBN,The Daily
Caller, Daily Mail,The Daily Wire, Fox News Opinion,The Federalist, National
Review, NewYork Post, Newsmax
Central 336 (25%)
AP, Reuters, Bloomberg,The Christian Science Monitor,The Hill, BBC, USA Today,
The Wall Street Journal Online, NPR Online News
Total 1,344 (100%)
- 34. Sentiment analysis by text
mining using LIWC
Logistic Regression
( left-wing is set as 0 ).
Data Visualization with radar
chart.
- 35. Logistic regression of cognitive loads
Coefficients Estimates St. Error Z-value
Intercept -1.649 .124 -13.269***
Cogproc -.102 .028 -3.679**
Cause .129 .048 2.710**
Discrep .169 .054 3.088**
Tentat -.048 .050 -.971
Certain .068 .046 1.483
differ .155 .047 3.313***
Note: ***: p<.001, **: p<0.01, *: p<0.05
- 36. Coefficients Estimates St. Error Z-value
intercept -1.489 .123 -12.089***
Affect -.294 .226 -1.299
Posemo .257 .228 1.129
Negemo .290 .230 1.259
Anx -.197 .111 -1.770
Anger .011 .061 .173
sad -.244 .129 -1.890
Logistic regression of affective processes
Note: ***: p<.001, **: p<0.01, *: p<0.05
- 37. Coefficients Estimates St. Error Z-value
Intercept -1.442 .232 -6.212***
Pronoun -.025 .019 -1.303
Article -.066 .024 -2.781**
Prep -.030 .017 -1.708
Auxverb -.006 .033 -.177
Conj .018 .030 .602
negate .042 .052 .807
Verb .046 .022 2.067*
Logistic regression of analytical thinking styles
Note: ***: p<.001, **: p<0.01, *: p<0.05
- 38. Coefficients Estimates St. Error Z-value
Intercept -1.442 .176 -8.213***
Focuspast .068 .028 2.442*
Focusfuture -.033 .050 -.658
Posemo -.024 .032 -.736
Negemo -.012 .047 -.260
Sad -.281 .133 -2.116*
Anx -.150 .112 -1.337
Anger .013 .062 .206
Certain .031 .045 .695
Work -.05 .022 -2.192*
Money .002 .046 .050
Logistic regression of profiles of political sentiment
Note: ***: p<.001, **: p<0.01, *: p<0.05
- 41. ▪ Allsides.org media bias rating: https://www.allsides.com/media-bias/media-bias-ratings
▪ Tumasjan, A.,T.O. Sprenger, P.G. Sandner, and I.M.Welpe. Predicting elections with Twitter:What 140
characters reveal about political sentiment. in Proceedings of the Fourth International AAAI
Conference on Weblogs and Social Media (ICWSM'10). 2010.Washington, DC: Association for the
Advancement of Artificial Intelligence, 178-185.
▪ Tripathi, G. and N. S. Opinion mining: A review. International Journal of Information & Computation
Technology, 2014. 4(16): 1625-1635.
▪ Yu, B., S. Kaufmann, and D. Diermeier. Exploring the characteristics of opinion expressions for political
opinion classification. in Proceedings of the 9th Annual International Digial Government Research
Conference. 2008. Montreal, Canada, 82-91.
▪ Tumasjan, A.,T.O. Sprenger, P.G. Sandner, and I.M.Welpe. Election forecasts with Twitter: How 140
characters reflect the political landscape. Social Science Computer Review, 2011. 29(4): 402-418.
▪ Stieglitz, S. and L. Dang-Xuan. Political communication and influence through Microblogging- An
empirical analysis of sentiment in Twitter messages and retreet behavior. in Proceedings of the 2012
45th Hawaii International Conference on System Sciences (HICSS'45). 2012. Hawaii: IEEE Computer
Society, 3500-3509.
▪ Dang-Xuan, L. and S. Stieglitz. Impact and diffusion of sentiment in political communication- An
empirical analysis of political Weblogs. in Proceedings of the 2012 Sixth International AAAI
Conference on Weblogs and Social Media (ICWSM'12). 2012. Dublin, Ireland: Association for the
Advancement of Artificial Intelligence, 427-430.
- 42. ▪ Stieglitz, S. and L. Dang-Xuan. Impact and diffusion of sentiment in public communication on Facebook.
in Proceedings of the 2012 European Conference on Information Systems (ECIS 2012). 2012. Association
for Information Systems (AIS), 1-13.
▪ Wang, H., D. Can, A. Kazemzadeh, F. Bar, and S. Narayanan. A system for real-time Twitter sentiment
analysis of 2012 U.S. presidential election cycle. in Proceedings of the 50th Annual Meeting of the
Association for Computational Linguistics (ACL'12). 2012. Jeju Island, Republic of Korea: Association for
Computational Linguistics, 115-120.
▪ Nooralahzadeh, F.,V. Arunachalam, and C. Chiru. 2012 Presidential elections on Twitter- An analysis of
how the US and French election were reflected in tweets. in Proceedings of the 2013 19th International
Conference on Control Systems and Computer Science. 2013. Bucharest, Romania: IEEE.
▪ Alashri, S., S.S. Sandala,V. Bajaj, E. Parriott,Y. Awazu, and K.C. Desouza.The 2016 US Presidential election
on Facebook: An exploratory analysis of sentiments. in Proceedings of the 2018 51st Hawaii International
Conference on System Sciences (HICSS'51). 2018.Waikoloa Village, Hawaii Big Island: University of
Hawaii, 1771-1780.
▪ Jordan, K.N., J.W. Pennebaker, and C. Ehrig.The 2016 U.S. Presidential candidates and how people
tweeted about them. Special Collection: SMaPP Global Special Issue, 2018: 1-8.
▪ Pennebaker, J.W., R.L. Boyd, K.N. Jordan, and K. Blackburn.The development and psychometric
properties of LIWC2015, 2015. by University of Texas at Austin.
- 46. The clues of fake news in Taiwan
228messages 152shares 1,846messages 1,085shares 183messages 108shares
- 48. The clues of fake news in
Taiwan
Social websites (e.g. BBS, PTT, )
- 50. The
prevention
methods in
Taiwan
Google forbids any political advertisement
during the 2020 election.
Facebook shows the contributor of the
advertisement.
Taiwan Factcheck Center
1. Reported 2. Media concern 3. misinformation
Line Rumor Beat
- 51. The Dilemma
&
Difficulty
Law (legal the business, news media exclusive of
other people)
Jurisdiction in Taiwan
Technology (VPN, private group)
Some Companies are not cooperative.
Tons of messages(include pictures, videos) have to
be checked.
Limited
resources
- 52. Freedom of speech
Freedom of news
The profits of the public
The prevention of crime
Both ends of the balance
LawDemocracy
- 54. People, Process, Technology
People
Process Technology
1. The ecology of Country
2. Different ages
3. The distribution of messages
4. Popular issues
5. Media literacy(education)
1. ISP or website manager
2. AI on data mining
3. Algorithm
4. Detective methods
1. To make law.
2. SOP (Standard Operation
Procedure)
3. The cooperation between
government and NGO.
4. Make the right recipe.
Sentimental analysis