The main reason for this update is that 15% of current searches by users on hong kong telegram data Google are about topics that have not been seen or searched for before . The aim is to achieve a mechanism that can respond to queries that cannot be anticipated. To do this, the goal is to understand the daily language used by users, regardless of how they write it or combine words. An algorithm based on neural networks for Natural Language Processing (NLP) .
How Google BERT works
Until now, Google analyzed word by word the expressions that users entered into the search panel, but with this new update, its intention is to relate each and every word, not only to understand its meaning, but also in the context in which the search was made .
This will be possible because, unlike the old algorithm, it will take into account connectors and prepositions and not just keywords. The company presented several examples to demonstrate how it works:
They performed the following search comparing the two algorithms: "2019 brazil traveler to usa need a visa". The old algorithm, which ignored the preposition “a”, did not perform the search correctly, since it changed the meaning of the sentence. However, the new algorithm, thanks to its understanding of this preposition, was able to perform a more careful search and returned the expected result.
