SlideShare a Scribd company logo

Machine Learning and Search -State of Search 2016

•Download as PPTX, PDF•
•
Eric Enge
Eric Enge

Eric Enge is a notable speaker, author, and entrepreneur in the fields of machine learning, search, and SEO. He is the CEO of Stone Temple Consulting. This document contains a series of slides by Eric Enge discussing machine learning and how it is impacting search engine algorithms like Google's RankBrain. Some key points discussed include how machine learning allows algorithms to better understand language patterns and contextual meanings, detect related concepts, and select the most relevant algorithms for a given query. Eric argues that while machine learning will continue to evolve search, content quality, authority, and user experience will remain important ranking factors for websites to focus on.

1 of 51
Eric Enge @stonetemple Machine Learning, Search & SEO
Eric Enge @stonetemple ERIC ENGE, CEO STONE TEMPLE CONSULTING NOTABLE SPEAKER, BOOK AUTHOR & ENTREPRENEUR 2016 US Search Awards 2016 Search Personality of the Year 2016 Landys Search Marketer of the Year
Eric Enge @stonetemple Nationalselfrelianceassociation.com Eric Enge – Fun Fact 1 October 2010
Eric Enge @stonetemple Eric Enge – Fun Fact 2 ntier.org February 2011
Eric Enge @stonetemple Machine Learning and Deep Learning Our Best Student!
Eric Enge @stonetemple Simple Learning Problem
Eric Enge @stonetemple With a Simple Solution
Eric Enge @stonetemple Much More Complicated Problem This is an Example of “Supervised” Machine Learning
Eric Enge @stonetemple And, Now the Complexity Went WAY up
Eric Enge @stonetemple Machine Learning is a 2 Step Process
Eric Enge @stonetemple Supervised Machine Learning Sample Problem: Recognizing a Face
Eric Enge @stonetemple Related Stories Auto- Discovered by Machine Learning Google News Uses Unsupervised Machine Learning
Eric Enge @stonetemple DeepMind’s AlphaGo Program Beat the World Champion 10 atoms in the known universe 10 possible positions in the game of Go 100 171
Eric Enge @stonetemple Enter RankBrain
Eric Enge @stonetemple Bloomberg Broke the News in October 2015
Eric Enge @stonetemple Basic Concept: Stop Words
Eric Enge @stonetemple Query That Doesn’t Work Well With Stop Words
Eric Enge @stonetemple Top Organic Results for the Query “Office”
Eric Enge @stonetemple “The” and “Office” are capitalized
Eric Enge @stonetemple “Yeah -- being able to represent strings of text in very high- dimensional space and "see" how they relate to one another.” Google Spokesperson’s Answer to a Question I Asked About RankBrain
Eric Enge @stonetemple Notice Patterns in Language Usage
Eric Enge @stonetemple Notice Patterns in Language Usage The • Classic stop word • But in some cases critical to phrase meaning • e.g. “The Office” Graphic • Could relate to violence • Could relate to language • Could relate to design • Meaning determined by context Without • Negatives traditionally ignored by Google • But in some cases critical to meaning of page
Eric Enge @stonetemple What Does RankBrain Do?
Eric Enge @stonetemple What Does RankBrain Do? Can result in selecting different algos
Eric Enge @stonetemple Can You Get 100% Score on Super Mario Without Walkthrough? Negatives
Eric Enge @stonetemple Our RankBrain Study
Eric Enge @stonetemple June/July 2015 Google Result January 2016 Google Result Pulled Results from Google on Two Different Dates Did This for 500K Queries
Eric Enge @stonetemple 28 Aggregate Results
Eric Enge @stonetemple Why Are PDFs So Weak? July 2015 Answer
Eric Enge @stonetemple Now Relates “Weak” to Security January 2016 Answer
Eric Enge @stonetemple Where is Celtics Bench? July 2015 Answer
Eric Enge @stonetemple Now Applies Meaning of “Where is” Better January 2016 Answer
Eric Enge @stonetemple Tourist Map of Mdina Malta Pre-RankBrain – July 2015
Eric Enge @stonetemple Example of How RankBrain Impacted Algo Selection
Eric Enge @stonetemple Machine Learning Summary Thoughts – What Will Such Algorithms Look For?
Eric Enge @stonetemple Quality Content
Eric Enge @stonetemple 37 Ways to Measure Authority (Links)
Eric Enge @stonetemple Source: Chartbeat How About Social Media as a Signal?
Eric Enge @stonetemple No Correlation Between Shares and Links Source: Moz and BuzzSumo Zero (or Near Zero) Means NO Correlation
Eric Enge @stonetemple BUT, They Will Look for Some Form of Social Proof
Eric Enge @stonetemple Need to Have the Right Training Data
Eric Enge @stonetemple Very Hard to Test Edge Cases You Can’t Define
Eric Enge @stonetemple Deriving and Testing Machine Learning Algorithms Takes a Long Time
Eric Enge @stonetemple When Do We Apply Machine Learning? When coming up with human algorithms proves to be impossible
Eric Enge @stonetemple
Eric Enge @stonetemple How to Prepare for This?
Eric Enge @stonetemple Machine Learning is Just a Tool in the Toolkit It’s Not Going to Blow Up SEO as We Know it
Eric Enge @stonetemple Personal Assistants More Likely to Drive Disruptive Change
Eric Enge @stonetemple What to Do? Align With the Search Engines 1. Work on Content Quality 2. Work on Your Authority 3. Increase User Satisfaction With Your Site 4. Create Demand for What You Do
Eric Enge @stonetemple Some Practical Ideas 1. Are you working on building your brand? 2. Do you have recognizable experts working with you? 3. Review engagement data on major pages. 4. Check your site search. 5. Review competitor sites for ideas. 6. Perform advanced content comparison analysis. 7. Design your site for Smartphone users. 8. Make your site as fast as possible.
Eric Enge @stonetemple Eric Enge eenge@stonetemple.com @stonetemple +Eric Enge (508) 962-8474 www.stonetemple.com Thank You!

More Related Content

Machine Learning and Search -State of Search 2016