Read the text below.
Google’s search engine can now better understand the meaning of a whole phrase when generating search results.
Pandu Nayak, the vice president of Search, announced the update to the search engine in October. Nayak said that to develop the latest improvements they made in Search, their team focused on the science of language learning.
Previously, Search treated queries as individual words instead of a single phrase with one meaning. The platform had the tendency to consider only words that it regards as important and ignore others. It also did not have the ability to interpret changes in meaning based on the order of words in a phrase. Thus, results would sometimes not be exactly what searchers were looking for.
To improve Search’s ability to understand human language and answer questions, Nayak and his team developed a new system. The company named this new system Bidirectional Encoder Representations from Transformers (BERT). Through BERT, Search can now better comprehend the meaning of an entire query.
For example, in the older version, if somebody typed the phrase “2019 Brazil traveler to USA need a visa,” Search would yield top results for the visa requirements of a US citizen traveling to Brazil. With BERT, Search is now able to return relevant top results for those traveling from Brazil to the United States using the same search entry.
BERT was initially designed to understand English, but Google plans to extend its advanced features to people worldwide. Nayak said that through the AI system they have developed, improving Search for other languages will be easy. So far, Search has seen great improvements in Korean, Hindi, and Portuguese.