Paper: Boosting Relation Extraction with Limited Closed-World Knowledge

ACL ID C10-2155
Title Boosting Relation Extraction with Limited Closed-World Knowledge
Venue International Conference on Computational Linguistics
Session Poster Session
Year 2010
Authors

This paper presents a new approach to im- proving relation extraction based on min- imally supervised learning. By adding some limited closed-world knowledge for confidence estimation of learned rules to the usual seed data, the precision of re- lation extraction can be considerably im- proved. Starting from an existing base- line system we demonstrate that utilizing limited closed world knowledge can ef- fectively eliminate ”dangerous” or plainly wrong rules during the bootstrapping pro- cess. The new method improves the re- liability of the confidence estimation and the precision value of the extracted in- stances. Although recall suffers to a cer- tain degree depending on the domain and the selected settings, the overall perfor- mance measured by F-score considerably improves. Final...