ReNoun Fact Extraction for Nominal Attributes 英文文献资料.pdfVIP

ReNoun: Fact Extraction for Nominal Attributes Mohamed Yahya∗ Steven Euijong Whang, Rahul Gupta, Alon Halevy Max Planck Institute for Informatics Google Research myahya@mpi-inf.mpg.de {swhang,grahul,halevy}@ Abstract gine typically returns (Sawant and Chakrabati, 2013; Singhal, 2012; Yahya et al., 2012). To Search engines are increasingly relying on support such answers, the search engine main- large knowledge bases of facts to provide tains a knowledge base that describes various at- direct answers to users’ queries. How- tributes of an entity (e.g., (Nicolas Sarkozy, ever, the construction of these knowledge wife, Carla Bruni)). Upon receiving a query, bases is largely manual and does not scale the search engine tries to recognize whether the to the long and heavy tail of facts. Open answer is in its knowledge base. information extraction tries to address this For the most part, the aforementioned knowl- challenge, but typically assumes that facts edge bases are constructed using manual tech- are expressed with verb phrases, and there- niques and car