Many scientific disciplines are represented in Yahoo! Labs Search Sciences, and within those disciplines lie a number of critically important application areas. As the web rapidly changes, Search Sciences constantly strives to push the limits of their expertise and continue to innovate in these areas.

Ranking

Ranking is the holy grail of Search Sciences. Techniques from information retrieval and statistical modeling are brought together in interdisciplinary pursuit of continual improvements to precision and recall of the search experience. Of course search results are ranked, but Search Sciences is also pushing the limits in ranking queries, particular genres of content (e.g., reviews), and snippets of text for creating optimal excerpts. Search Sciences is constantly innovating on features for ranking and on modeling algorithms; see their publications list for detailed technical information on their recent work.

Classification

Classifying pages according to a variety of taxonomies is critically important for search and therefore an important activity in the Search Sciences group. On the content analysis side, they use classification in the ranker, for topic identification and topic segmentation, and for fighting spam. Search Sciences builds classifiers for many other purposes as well - to distinguish between query classes, or to detect patterns of behavior in order to learn from our users and to combat adversarial activities.

Information Extraction

Structured data is very effective in many areas of search, from information integration to presenting rich results. The web, however, consists predominantly of unstructured and extremely varied content. The Search Sciences group uses information extraction algorithms to bridge the gap between the unstructured content and the applications that require richer representation of explicit knowledge. Yahoo! Labs Search Sciences is truly pushing the boundaries of information extraction, both in terms of precision and in terms of scale of application.

Summarization

Search engines would not be effective in presenting results to users if they could not summarize documents and websites. Yahoo! Labs Search Sciences builds specialized summarizers that use techniques from information retrieval and natural language processing to construct excerpts of web pages, news articles and sometimes even entire websites.

Back