Paper: Unsupervised Entailment Detection between Dependency Graph Fragments

ACL ID W11-0202
Title Unsupervised Entailment Detection between Dependency Graph Fragments
Venue Workshop on Biomedical Natural Language Processing
Session
Year 2011
Authors

Entailment detection systems are generally designed to work either on single words, re- lations or full sentences. We propose a new task – detecting entailment between depen- dency graph fragments of any type – which relaxes these restrictions and leads to much wider entailment discovery. An unsupervised framework is described that uses intrinsic sim- ilarity, multi-level extrinsic similarity and the detection of negation and hedged language to assign a confidence score to entailment rela- tions between two fragments. The final system achieves 84.1% average precision on a data set of entailment examples from the biomedical domain.