Paper: A Study on Dependency Tree Kernels for Automatic Extraction of Protein-Protein Interaction

ACL ID W11-0216
Title A Study on Dependency Tree Kernels for Automatic Extraction of Protein-Protein Interaction
Venue Workshop on Biomedical Natural Language Processing
Session
Year 2011
Authors

Kernel methods are considered the most ef- fective techniques for various relation extrac- tion (RE) tasks as they provide higher accu- racy than other approaches. In this paper, we introduce new dependency tree (DT) ker- nels for RE by improving on previously pro- posed dependency tree structures. These are further enhanced to design more effective ap- proaches that we call mildly extended depen- dency tree (MEDT) kernels. The empirical re- sults on the protein-protein interaction (PPI) extraction task on the AIMed corpus show that tree kernels based on our proposed DT struc- tures achieve higher accuracy than previously proposed DT and phrase structure tree (PST) kernels.