April 12, 2006

A Million Portraits at Yandex.News

Internet, April 4. 2006. Yandex provides access to background information about people mentioned by the Russian mass media. Now a search in Yandex.News returns links to so-called press portraits of those whose names correspond best to the query. Press portraits are formed automatically by a special computer program, using mass media materials for sources.

Each press portrait contains a short background description of a person, specifically place of work and position, quotes, and statements of third parties. All the details are supplemented with excerpts from source documents; this helps grasp the context and assess information reliability. In essence, a press portrait is another type of Yandex.News search results page. It is possible to search for people not only by the name (for example, Anatoly Chubais or Ostap Bender), but also by organization (like Soyuzmultfilm, Microsoft, or the House of Lords) — in this case, search results will display everyone who is or was related to the specified organization, according to mass media.

“Last year, the pages of the service offered Names in the News — ratings of the most popular people, automatically formed for each news item, column, or search results. Today we use technologies that allow both to identify the names of people covered by a news item and to provide background information about every person involved,” tells Tatyana Platonova, head of Yandex.News service. “The press portraits are especially handy when the media write about someone unfamiliar to the general audience. Now the users can find out who is who in mere seconds.”

The new service is based on a unique technology owned by Yandex company. Extracting facts from text and grouping them into a press portrait (i.e. determining which of the facts refer to the same person) is performed automatically, without human assistance. “For the first time in Russia, a fact extraction technology is used on a mass web-service and processes large text blocks. The purpose of this technology is to recognize identity of various designations of the same object and differentiate between objects with identical names,” explains Lev Gershenzon, leading developer of Yandex company. “We learned to automatically link two types of objects: people and organizations. However, despite the fact that we have made a major achievement, information about people may turn out to be incomplete or inaccurate, since press portraits are composed by a computer program. We plan to perfect the technology and will be very grateful for any feedback and suggestions on enhancing the quality of the service.”

Starting today, Yandex.News has become an aggregator of both news and background/reference information. The reference section of Yandex.News service has more than one million of press portraits, and 30,000 of them are updated on a daily basis. The News service was launched six years ago, so most of the press portraits are created on the basis of analysis of mass media reports since 2000. The service is now adding news sources’ archives from previous years, improving the completeness and accuracy. Currently, Yandex.News database stores more than 9 million information messages from over 1,000 sources.

For more information about the technology, visit http://news.yandex.ru/people-search-tech.html.

Elena Kolmanovskaya, Chief Editor
Michael Ushakov, Head of Yandex PR
Phone: +7 495 739-70-00
Fax: +7 495 739-70-70
E-mail: pr@yandex-team.ru

Close window | Back to top

Copyright 2018 Yandex