Поиск Почта Карты Маркет Новости Словари Блоги Видео Картинки
компания → интернет-математика

Contest description

The Switching Detection Challenge provides a unique opportunity to consolidate and scrutinize the work from industrial labs on predicting the user search actions using user logged search behavior context. It provides a fully anonymized dataset shared by Yandex which has anonymized users, queries, clicks and search engine switching actions.

Users often use more than one search engine to satisfy their information needs. When a user deliberately switches from one engine to another in order to continue their search, this means they require additional information and perspectives. This may be because their information need requires more than one engine can provide, or because the first engine has failed. For some search sessions, the fact of switching can be easily monitored. For instance, users might click a search engine URL on the search engine result page (Yandex has always had links to all major search engines at the bottom of its results) or the fact of switching might be monitored via the browser toolbar. However, the same users might switch in a way, which cannot be monitored. For instance, they might use their browser bookmarks to switch the engines, or they might use another browser (without the toolbar installed). Detecting the fact of switching, when switching events cannot be monitored directly is difficult, and is not yet a solved problem. It is important for understanding of users' satisfaction with the search engine and the complexity of search. This Challenge and the shared dataset aims to enable a whole new set of researchers to study the problem of search engine switching detection.

The Switching Detection Challenge is a part of series of contests organized by Yandex called "Internet Mathematics". This year’s event is the seventh since 2004. Participants will again compete in finding solutions to a real-life problem based on real-life data. In previous years, participants tried to learn to rank documents, predict traffic jams, find similar images and predict relevance of documents using search logs.

Congratulations to the prize winners:

1st prize: Out of mEmory  Denis Savenkov, Dmitry Lagun, Qiaoling Liu. Emory University [report]

2nd prize: Insight Pavel Kalinin. Voronezh State University [report]

3rd prize: GraphLab Qiang Yan(Institute of Automation, Chinese Academy of Sciences), Xingxing Wang(Computer Network Information Center, Chinese Academy of Sciences), Heng Gao, Dongying Kong (Institute of Computing Technology, Chinese Academy of Sciences), Yangbao Lee(Chinese Academy of Sciences) [report]

Reports of the best teams were presented at the WSCD workshop, collocated with WSDM2013 conference.

The contest organizers can be reached via email challenge@yandex-team.ru.


07.03.2013:  We are pleased to present the report of wangzongzaimeia team, which scored fourth place in the Challenge: [report]