| A Workshop on Machine Learning in Natural Language Processing |
Organizers: Shalom Lapin and Ido Dagan |
Pattern-based Lexical Acquisition and Self-training of Statistical Parsers: Two Cases in a Construction Grammar Approach
Ari Rappoport
Hebrew University of Jerusalem
I will briefly overview the Construction Grammar approach in Cognitive Linguistics, and will present two lines of research done at the HUJI NLP group that have a shared construction grammar motivation. First, I will describe our algorithm for using patterns in order to acquire lexical categories in a fully unsupervised and language independent manner. Our results are superior to all previous (even POS-supervised) work, while the computation is orders of magnitude more efficient. Second, I will show how self-training can enhance the performance of statistical parsers and adaption of parsers between domains. I will close by speculating on the shared aspects of these results. |