Google’s new RankBrain Adds Machine Learning and AI to its Search Algorithm
In Chapter 13 we learned about the overarching structure of the internet and how it is pretty much a giant tangled “web” of nodes connecting to a bunch of other nodes in a directed graph. We also learned that despite this complexity, on a macroscopic scale, the internet does seem to have a large structure inherent, most notably with its Strongly Connected Component (SCC). How does a giant search engine like Google parse through this massive globule of information, given a search query? The backbone to this answer lies in the PageRank algorithm that we learned and analyzed in chapter 14.
However, Googles complete algorithm, which is called “Hummingbird”, takes in many different “signals”, with the PageRank algorithm results being just one component out of hundreds of others. The contribution of each signal is not the same, however. Traditionally, these signals have been discovered manually by the clever thinking of Google’s search engineers. But Google has recently announced to the public a new signal, named RankBrain, that relies on the power of artificial intelligence and Machine learning, rather than the human brain, to deliver more relevant search results. This algorithm will certainly help Google with unusual and novel search queries that the search engine may not have handled before; up to 15% of google searches have never been searched before! How does a search engine adapt to provide relevant results to new search queries?
RankBrain actually does not seem to be a traditional metric that measures the importance of a website. Rather, it is a method that helps refine the search query entered by connecting it to other, closely related, and perhaps more specific queries. Have you ever typed a poorly worded question into google? RankBrain would use machine learning to help link your query to other, more popular queries that are more likely to have what you are looking for. No matter how good your PageRank algorithm is, if your query is not good (specific), then you cant expect anything good to come out of it. But now, you can!
Although the specific details on how exactly Google’s new algorithm works have not been revealed, and probably wont be considering it is the secret sauce to their search business, its announcement is a good reminder that there are still many problems and much to learn about in the search and information industry, and solutions like RankBrain are only one of many that will optimize our search experiences far into the future.
[Source: http://9to5google.com/2015/10/26/google-rankbrain-artificial-intelligence-search/]
[Source: http://searchengineland.com/meet-rankbrain-google-search-results-234386]