Today, it was reported that Google is currently adjusting its core search algorithm. According to the company, this will help the search results increase by 10%. It is also reported that Google will introduce sophisticated natural language processing (NLP) technology in its search results.
Google is claiming that the new search result will improve the search results by a better understanding of words in sentences. Pandu Nayak, a Google researcher and vice president of search business said that the old Google search algorithm used to break down sentences into “a series of words” and analyze the meaning of important words to give simple local search results whereas the new algorithm understands the context and gives more accurate and accurate results.
It is reported that Google’s new search algorithm is based on the language expression model BERT (Bidirectional Encoder Representations from Transformers).
Jeff Dean, senior researcher and senior vice president of research at Google said that over time, this training has become very effective in getting the NLP model to understand the context.
All improvements to the search go through a series of tests to make sure they really improve the results of the query. One of the tests involved the use of Google’s core human reviewers who trained the company’s algorithms by rating the quality of search results. Along with this, Google will also conduct real-time A/B testing.
Google said that the company has made efforts to ensure that the BERT is not added to the search algorithm which is a common problem in machine learning. The training model of machine learning itself is also biased. Since BERT is trained on a huge English sentence corpus, these sentences are also biased in themselves, so this is a problem that needs close attention.
However, Google confirmed that all queries won’t be affected by BERT. How this tool works is still a mystery. Google intends to keep some of these processes secret to prevent spammers from spoofing their systems.