Skip to main content

Google brings in BERT to improve its search results


In this day and age, where Google receives over 63000 search queries per second, and 35 billion per day, it comes to no surprise that any change in their search algorithm is important news. As the most popular and widely used search engine (over 90% of search queries are Google Searches), a small modification to this algorithm can have significant repercussions.


This article explains how Google recently implemented the BERT (Bidirectional Encoder Representations from Transformers) algorithm to better optimize its search results. BERT uses neural networks and natural language processing to better process data where the order of the elements is important. As mentioned in the article, this change is optimal for “longer, more conversational queries”. Therefore, it is better suited to handling longer phrases and sentences, allowing Google to return more reliable results. This update will affect around 10% of search results, which is the greatest change Google has had in many years. 


This article is relevant to the subjects studied in lecture because we went over how companies like Google use the PageRank algorithm to determine what results to return when a person performs a search over the Internet. While the PageRank algorithm we saw in class would be better at performing searches based on individual keywords, the new algorithm would be better for coherent questions. The implementation of BERT will have a great impact for many of the parties involved. For example, as 10% of the searches are modified, business owners that depend on website traffic could have their revenues greatly affected. In addition, SEO (search engine optimization) services would have to change their methods in order to adapt to the new changes. 


Leave a Reply

Blogging Calendar

October 2019