Google launched the largest search engine update

Google has launched the largest search engine update in five years. The new technology based on artificial intelligence will become more “understanding”, as it will analyze not individual words, but the requests of users as a whole Google began to introduce an update to the algorithm of its search engine, which in its blog calls the largest in at least five years.

The Financial Times notes that this is the first time that the technology of natural language processing has been applied in practice, which has attracted the attention of scientists last year. Understanding the language is one of the biggest challenges for artificial intelligence due to its variability, the newspaper writes. The meaning of words may vary depending on who and in what context uses these or those words.

So far, Google’s search engine algorithm has tried to determine which of the words in a user’s query has the most meaning. He ignored short or frequent words. This allowed the search engine to determine the main subject of the request, but often it was difficult to understand what exactly the user wanted.

Google Predicted Euro vision Winner Based on Search Queries

The new technology, known as BERT, relies on a language model based on an analysis of huge amounts of text on the web. Instead of parsing the words in a query one by one, BERT parses them simultaneously, including previously ignored short words. As an example of a query that the search engine has now learned to deal with, company representatives cited this: “How old was Taylor Swift when Kanye jumped on stage?”

At the same time, the company reports that the changes will remain invisible to most users. The update will affect one out of ten queries, said FT Pandu Nyak, vice president of search. In recent days, Google has already started updating the software of its data centers in order to change the process of processing requests in English, and in the near future it will deal with requests in other languages.

Nyack also emphasized that in some cases the new algorithm finds information worse than the old. He emphasized that the work on the project will be constantly continued. “I don’t think that we are close to solving the language problem, but this is a good step,” said Jeff Dean, head of the Google’s artificial intelligence department, to the newspaper.