TitleOpen Information Extraction to KBP Relation in 3 Hours
Publication TypeConference Paper
Year of Publication2013
AuthorsSoderland S, Gilmer J, Bart R, Etzioni O, Weld DS
Conference NameText Analysis Conference - Knowledge Base Propagation (KBP)
Date or Month PublishedNovember
Conference LocationGaithersburg, MD
AbstractWe participated in both the English Slot Filling and Entity Linking in the 2013 TAC-KBP evaluation. Our Slot Filling system provides an answer to the following conjectures: Can Open Information Extraction (Open IE) form the basis of a high precision extractor for a set of target relations in an ontology? And, just as importantly, can this be done with a minimum of human knowledge engineering? We built rules to map Open IE extractions to KBP Slot Filling relations and found that just three hours of rule creation gave extractor precision of 0.79. Spending a total of 12 hours refining the rules increased precision slightly to 0.80 with recall near the median of other KBP systems.
Downloadshttp://homes.cs.washington.edu/~soderlan/Soderland-KBP2013.pdf PDF
Month of PublicationNovember
Citation Key10006