Spotting Facts in the Wild
Tomasz Tylenda, Yafang Wang, Gerhard Weikum · Max Planck Digital Library · 2014
Knowledge bases have become key assets for many tasks related to search and analytics.Retrieving textual evidence for the facts about an entity -in news, social media, or other web contents -is important for assessing the truth of statements, gathering statistics about the saliency of facts, and as a basis for summarizing documents or entire corpora.This paper addresses the problem of spotting occurrences of known facts in textual documents, reporting the presence or absence of facts.Our solution is based on rich dictionaries of paraphrases for entity names and binary relations, and uses heuristic rules for high precision and recall.We evaluated our method by finding facts from Freebase in a corpus of stylistically diverse biographies from the web.