{"id":120782,"date":"2019-03-05T20:49:32","date_gmt":"2019-03-05T20:49:32","guid":{"rendered":"https:\/\/www.searchenginewatch.com\/?p=120782"},"modified":"2020-02-17T13:02:19","modified_gmt":"2020-02-17T13:02:19","slug":"google-rankbrain-myths-misconceptions","status":"publish","type":"post","link":"https:\/\/searchenginewatch.com\/2019\/03\/05\/google-rankbrain-myths-misconceptions\/","title":{"rendered":"Google RankBrain: Clearing up the myths and misconceptions"},"content":{"rendered":"

It\u2019s been nearly 3\u00bd years since Google<\/a> first announced their usage of RankBrain (October 26th 2015, but it had started being rolled out early 2015, in multiple languages).<\/strong><\/p>\n

In that time, there\u2019s been little in the way of details coming from G about what it is or how it works.<\/span><\/p>\n

The result is that numerous SEOs have stepped up to fill that void with their own speculations and opinions, and in doing that, have caused all sorts of confusion.<\/span><\/p>\n

This is my attempt to correct and clean up some of that mess.<\/span><\/p>\n

(There is a TL:DR<\/a> at the bottom if you want to skip the verbiage :D)<\/span><\/p>\n

What does RankBrain do?<\/b><\/h2>\n

Though there isn\u2019t much publicly available, what we do have is fairly specific:<\/span><\/p>\n

\u201c<\/span>If RankBrain<\/a> sees a word or phrase it isn\u2019t familiar with, the machine can make a guess as to what words or phrases might have a similar meaning and filter the result accordingly, making it more effective at handling never-before-seen search queries.<\/span><\/i>\u201d<\/span><\/p>\n

– Greg Corrado, from Bloomberg’s\u00a0<\/i>Google Turning Its Lucrative Web Search Over to AI Machines<\/i><\/a>}<\/p><\/blockquote>\n

Or, if you want it more succinct than that;<\/span><\/p>\n

\u201c… <\/span>Lemme try one last time: Rankbrain lets us understand queries better.<\/span><\/i> …\u201d<\/span><\/p>\n

– Gary Illyes (@methode), on <\/em><\/span>Twitter<\/span><\/a><\/em><\/p><\/blockquote>\n

Google<\/a> receives a fair percentage of queries per day that it hasn\u2019t seen before:\u00a0<\/span>15% at last check<\/span><\/a>.<\/span><\/p>\n

These may include misspellings and typos, elisions\/omissions, unusual phrasing\/syntactic structures, the wrong word(s) being used, negations (\u201cnot x\u201d), things that have only just happened etc. etc. etc.<\/span><\/p>\n

RB receives these weird, wonderful, and new searches, and attempts to identify existing searches and results that are probably suitable for the searcher\u2019s query.<\/span><\/p>\n

How does RankBrain work?<\/b><\/h2>\n

Again, we aren\u2019t exactly given a guided tour by G on this, but there are a few bits and pieces.<\/span><\/p>\n

\u201c… <\/span>RankBrain uses artificial intelligence to embed vast amounts of written language into mathematical entities — called vectors — that the computer can understand. If RankBrain sees a word or phrase it isn\u2019t familiar with, the machine can make a guess as to what words or phrases might have a similar meaning and filter the result accordingly, making it more effective at handling never-before-seen search queries.<\/span><\/i> \u2026\u201d<\/span><\/p>\n

– Greg Corrado, from Bloomberg\u2019s : <\/span><\/i>Google Turning Its Lucrative Web Search Over to AI Machines<\/span><\/i><\/a><\/p><\/blockquote>\n

So, rather than looking at words and attempting to parse them and understand the semantics (traditional Natural Language Processing [NLP<\/a>]), it converts them into numbers and plots them on a chart (with multiple dimensions, not just X and Y).<\/span><\/p>\n

Items near each other possess some form of relationship. The type of relationship will be reflected by each term\u2019s position and distance from its neighbors.<\/span><\/p>\n

If that sounds vaguely familiar, that\u2019s because it sounds very similar to Word2Vector.<\/span><\/p>\n

So when G receives a query it doesn\u2019t quite recognize, it can find semantically related pieces, and look at the results.<\/span><\/p>\n

But, what if it\u2019s wrong?<\/span><\/p>\n

Well, that\u2019s where Gary Illyes\u2019s answer to a question on his recent Reddit AMA may come in:<\/strong><\/h3>\n

\u201c…<\/span><\/p>\n

RankBrain is a PR-sexy machine learning ranking component that uses historical search data to predict what would a user most likely click on for a previously unseen query. It is a really cool piece of engineering that saved our butts countless times whenever traditional algos were like, e.g. “oh look a “not” in the query string! let’s ignore the hell out of it!”,<\/span><\/i> but it’s generally just relying on (sometimes) months old data about what happened on the results page itself<\/i><\/b>, not on the landing page<\/a>. Dwell time, CTR, … those are generally made up crap. Search is much more simple than people think.<\/span><\/i><\/p>\n

\u2026\u201d<\/span><\/p>\n

– Gary Illyes (@methode), on <\/span>Reddit<\/span><\/a><\/p><\/blockquote>\n

I\u2019ve added the bold to draw your eye to the key part.<\/span><\/p>\n

G may go back and look at what gets clicked for different searches, and check their performance. This can help the system learn what suggestions are suitable, and which ones are fails.<\/span><\/p>\n

If you want something with a bit more meat, you may be wanting some patents?<\/strong><\/h3>\n

If so, I was lucky enough to get some help from Bill Slawski,\u00a0<\/span>who pointed me to two potentially interesting patents:<\/span><\/p>\n