The source – youtube.com
Bidirectional Encoder Representations from Transformers (BERT) is a model that has led to a breakthrough in search history. Google has been using BERT since 2018, but only decided to share the benefits now.
It is an ML model that helps Google process language and understand the context in which it appears. Previously, the search engine extracted the most important words from a query to process it. With BERT, secondary, shorter words are also taken into account and that changes things.
Because it's hard for computers to capture the tone, intent, and context of speech like humans do, BERT helps process just that. Since 2019, it has improved a large number of queries by targeting different meanings of words in different order.