Computational Linguistics Colloquium

Thursday, 15 January, 16:15
Conference Room, Building C7 4

Discovering Verb Classes in Biomedical Texts

Lexical verb classes (e.g. in VerbNet) have been used to support various (multilingual) NLP application tasks, such as computational lexicography, language generation, machine translation, word sense disambiguation and semantic role labeling. However, exploitation of the classes in domain-specific applications has not been possible because existing manually built classifications are incomprehensive and unsuitable for specific domains. Recent research shows that it is possible, using current NLP and machine learning technology, to automatically discover verb classes in cross-domain corpora with promising accuracy. We apply similar technology (without any domain-specific tuning) to the important domain of biomedicine, and show that the resulting classification, acquired from a corpus of biomedical journal articles, is highly accurate and strongly domain specific. We discuss how such an automatically acquired classification can be used to improve the processing, mining and extraction of knowledge from the rapidly growing literature in biomedicine.

If you would like to meet with the speaker, please contact Valia Kordoni.