Paper: Competitive Self-Trained Pronoun Interpretation

ACL ID N04-4009
Title Competitive Self-Trained Pronoun Interpretation
Venue Human Language Technologies
Session Short Paper
Year 2004
Authors

We describe a system for pronoun interpre- tation that is self-trained from raw data, that is, using no annotated training data. The result outperforms a Hobbsian baseline algorithm and is only marginally inferior to an essentially identical, state-of-the-art su- pervised model trained from a substantial manually-annotated coreference corpus.