Google Search will soon understand normal questions, not just keywords
Ever had to search for something on Google, but youre not exactly sure what it is, so you just use some language that vaguely implies it?
Google announced today its rolling out a new machine learning-based language understanding technique calledBidirectional Encoder Representations from Transformers, or BERT.
According to Google, when it comes to ranking results, BERT will help Search better understand one in 10 searches in the U.S.
In the announcement, Pandu Nayak, Googles VP of search, called this kind of searching keyword-ese, or typing strings of words that they think well understand, but arent actually how theyd naturally ask a question. Its amusing to see these kinds of searches heck, Wired has make a whole cottage industry out of celebrities reacting to these keyword-ese queries in their Autocomplete video series but Nayaks correct that this is not how most of us would naturally ask a question.
Nayak said this the biggest leap forward in the past five years, and one of the biggest leaps forward in the history of Search. Google offered several examples of this in action, such as Do estheticians stand a lot at work, which apparently returned far more accurate search results.
We use cookies and analyse traffic to this site. By continuing to use this site, closing this banner, or clicking "I Agree", you agree to the use of cookies. Read our privacy poplicy for more information.