A Machine learning Filter for Relation Extraction

The TAC KBP English slot filling track is an evaluation campaign that targets the extraction of 41 pre-identified relations related to specific named entities. In this work, we present a machine learning filter whose aim is to enhance the precision of relation extractors while minimizing the impact on recall. Our approach aims at filtering relation extractors' output using a binary classifier based on a wide array of features including syntactic, lexical and statistical features. We experimented the classifier on 14 of the 18 participating systems in the TAC KBP English slot filling track 2013. The results show that our filter is able to improve the precision of the best 2013 system by nearly 20\% and improve the F1-score for 17 relations out of 33 considered.