Towards temporal web search

Abstract

This paper shifts the focus of Web search towards finding and exploiting small text nuggets, rather than full-length documents, assumming that the type of targeted information (e.g., date) is specified in the queries. Each nugget is a document sentence fragment that encodes open-domain factual information associated to some entity. The entities are dates (e.g., 1971) when the events captured by the nuggets must have occurred. The nuggets are extracted from the unstructured text of Web documents, based on lightweight text processing. Since per-document time stamps are available only in restricted collections such as news articles, but not in arbitrary Web documents, the extraction of the relevant dates relies on simple, local text analysis. The resulting time-stamped text nuggets have immediate applications in Web search, as direct results for queries asking about the date of an event explicitly (e.g., "When was the transistor invented") or implicitly (e.g., "Golden Gate Bridge built").