قالب وردپرس درنا توس
Home / Business / Google Search launches new technology to better understand questions

Google Search launches new technology to better understand questions



Meet BERT, the machine-driven language technology, which Google executives said would better understand the context of search.

Posted October 25, 2019 at 3:02 PM ET


Tarik Kizilkaya / Getty Images

MOUNTAIN VIEW, California – Google is making a major change to the way it presents search results. Most people won't notice it, but, Google executives say, the new technology represents a leap forward in the company's ability to understand what people are asking for when they apply.

Google's search engine uses natural language processing to analyze questions. For example, for the listing "get medicine at pharmacy", Google will submit the words "medicine" and "pharmacy" and look for pages related to these two keywords.

As of today, the company is launching a new machine learning driven language processing method called Bidirectional Encoder Representations From Transformers, or BERT for short. BERT looks at the sequences of words in search – not just the words themselves – to get more information about the intent behind them.

For example, results for "can you get medicine for any pharmacy" previously earned a link to a MedlinePlus 2017 article about getting a prescription filled, and missed the point that the search was looking for information on how to get prescription to someone else . Using BERT, Google's search engine now displays a 2002 article from the Department of Health and Human Services about how a friend or family member should obtain the medicine on your behalf.

Pandu Nayak, Google's search manager, gave another example of a press event yesterday, using the query "How old was Taylor Swift when Kanye walked on stage?" Before BERT, Google appeared on videos of the event in 2009, in which the rapper interrupted the pop star's acceptance speech during the MTV Video Music Awards. After BERT, Google presents as its first result an excerpt from a BBC article, which states: "A 19-year-old Swift had just defeated Beyoncé to win the best female video for her country-pop teen anthem You Belong With Me." Google's search returns the automatically highlighted "19-year-old" for emphasis.

"On the ranking front, this is the biggest change we've had in the last five years – and one of the biggest from the beginning," Nayak said.

While most people probably won't realize that a new natural processing technology is at work, behind the scenes, the change is likely to affect websites that rely on Google for traffic. The tech giant controls 90.8% of the search market across all products, including YouTube and Google Maps. Google uses the search and news tools to drive most traffic to news site articles, especially those related to politics and technology.

In response to a reporter's question about Google, it saw significant changes in traffic it sent to some specific sites in tests by BERT, said Ben Gomes, vice president of core search, "We didn't see any particular pattern there. But if you have niche issues, a niche publisher might show up for that. "

Google executives did not say whether websites should expect to see more or less traffic, but Gomes added that improving user functionality would result in more searches , which will bring more traffic to all websites. "When we answer more exotic questions, hopefully it will cause people to ask more and more exotic questions," he said.

BERT, Nayak said, was released to improve more complex searches, and would affect only 1 in 10 searches in the United States, yet with billions of searches per day, hundreds of millions of results have changed overnight.

There are areas that BERT is not coping well, Nayak said. In one example, for the search "tartan", BERT promoted dictionary results because it is a technology that focuses solely on text. Pre-BERT searches showed images of tartan, which is a more compelling result.

In another case, when searching for "Which state is south of Nebraska?" BERT surfaces the Wikipedia page for Southern Nebraska, as opposed to the ideal result, which would be the Wikipedia page for Kansas.

People who use Google will not know if their results are driven by BERT and cannot revert to non-BERT results. Going forward, BERT will be used on all US searches. Nayak said he is confident that BERT will generally improve the results for more complicated questions: "We are playing a statistical game here. Together we know that any change will have some gains and some losses. … I think the payout [of BERT] is very much there. ”

Originally, BERT will only be used in English search in the United States.


Source link