Which search engines use artificial intelligence

Artificial intelligence in search engines?

Status: 04/26/2017 5:24 p.m. | archive

What is often quoted and discussed on the Internet lands high on Google. Often, however, the search engine does not really recognize what we actually want to know. In the future, search engines could therefore use artificial intelligence to understand us better.

This is what the beta version of the Yewno search engine looks like.

A comparison: Anyone who asks the search engines Google and Yewno for the term "World Wide Web" receives different results. On page one of the Google results list there are some references to Wikipedia, a lexicon entry, links to two glossaries and some magazine articles. The results at Yewno: web browser, web page, hyperlink, HTML, Tim Berners-Lee - as well as more than 15 other related terms, their relationships with one another and a brief definition of the term "World Wide Web".

"Yewno had to read a lot to deliver this result," says Berthold Gillitzer, department head at the Bavarian State Library in Munich. The library, which in turn has a stock of over ten million books, is the first institution in Europe to cooperate with the search engine provider Yewno. The company is a by-product of a biotechnology start-up. It was there that the idea of ​​automatically scouring scientific literature for unknown contexts germinated. Yewno is still in the experimental stage, but is on the way to the Holy Grail of search engines: the so-called semantic search, a search based on contexts of meaning.

Don't always search for strings accurately

"All search engines work with strings," says Gillitzer. There has always been the problem of ambiguity. "If you enter the term 'bank', the search engine simply doesn't know whether I'm looking for a bank or a piece of seating furniture." Websites or books that contain the word you are looking for sometimes have no relation to each other. Computer scientists and linguists have been racking their brains over this problem for decades.

Some have tried to invisibly integrate the meaning of websites. Behind the scenes of the website, categories from a catalog of meanings should be listed. Basically, however, they would only be keywords. And adding this additional code to all content on the World Wide Web afterwards is impractical. Yewno no longer needs such help. "At Yewno there is no database with technical terms or a lexicon in the background," says Gillitzer. Rather, the search engine learns from texts alone and from a linguistic analysis of the texts.

A search engine that just keeps getting smarter

Yewno actually learns like a child: the more the computer program reads, the more it understands. The search engine is offered in Munich in addition to the library catalog. At the moment, Yewno only knows the English language and is only learning from scientific magazines, but of course that should change. It is already obvious what a huge leap technology represents: If you search with Yewno, you will get tons of useful cross-references, you will come across useful connections that you might not have thought of.

Semantic search engines: the archivists of the future?

Basically, the software is the artificial output of a dying profession. "In the past you could just talk to the librarian," says Gillitzer. You presented your request to him, and he then understood what you were looking for - because he was able to assess you and because he knew the inventory. Today, however, global knowledge is increasing so quickly that no librarian can keep up. The Bavarian State Library alone is growing by around 100,000 books every year.

Yewno gives an idea of ​​the potential of search engine technology that can independently record and classify knowledge with the help of artificial intelligence. It will be interesting to see when this approach will be integrated into the existing Internet search engines.

This topic in the program:

NDR Info | Network world | 04/27/2017 | 8:08 am