RankBrain: Google’s Solution to PageRank’s Comprehension Problem
With technology advancing at a dramatic pace, companies are more frequently turning to artificial intelligence research. Google has been one of the biggest proponents of such research, specifically focusing on machine learning. Such efforts have resulted in their fairly new program, RankBrain, which is implemented to help produce more accurate results when using the popular search engine. Though this idea is not directly related to the ideas of PageRank that we have discussed in lecture recently, it is connected to the question that drives PageRank as well as other search algorithms that google uses: when a user enters a search, what determines what will be at the top of the page?
In order to answer this question, Google implements many different programs to ensure that the content and results that are returned first from a search are of most value to the user. The earliest of these programs, and the one that is most often cited when determining the results of a search query, is the algorithmic process of PageRank. Though the process is deeply complex, the basic intuition behind it involves giving each page a weighted “rank”, which is determined by factors such as outgoing links, and which constantly changes through repeated improvement. When a user searches a certain topic, the pages with the highest rank in regard to that topic will appear at the top of the page. While PageRank is effective for determining which pages should appear based on a given search, there is still a flaw in the program regarding the search query itself. It should be kept in mind that the user who is doing the search is human, and is not always going to enter their search in a perfect way for the PageRank algorithm to comprehend. When outside words and language are used by the user, or if the user asks a very specific and unique question, PageRank may not be as effective by producing content that may not be relevant to the search topic. That is where RankBrain comes in, Google’s program that uses artificial intelligence to interpret search queries. It is through this program working concurrently with PageRank (among other programs) that Google is able to be most effective in determining what content to display first.
So how does the RankBrain program work? The program stores written language in mathematical form that a computer is able to process, and is used mainly used when a user googles something that the normal PageRank program may not be able to understand. Using artificial intelligence, when RankBrain encounters a word or phrase that it is not familiar with in a search, it can make an accurate “guess” on what other words and phrases may be associated with the unknown. The program then relates this guess back to PageRank, and the program runs as it normally would. As mentioned in the article, over 15% of searches on Google in a day have never been seen by the search engine before, which is why a program like RankBrain is so beneficial. Another interesting aspect of RankBrain is the idea of machine learning that is at its core. With every search that it encounters, RankBrain is constantly improving and adding data to its system that it can utilize in the future. The program is constantly evolving and becoming more accurate every time it is implemented, making Google at the cutting edge of technology by incorporating artificial intelligence into one of its major businesses.
Article Consulted:
https://www.bloomberg.com/news/articles/2015-10-26/google-turning-its-lucrative-web-search-over-to-ai-machines