Feature-Rich Log-Linear Lexical Model for Latent Variable PCFG Grammars
Zhongqiang Huang, Mary P. Harper · International Joint Conference on Natural Language Processing · 2011
Context-free grammars with latent annotations (PCFG-LA) have been found to be effective for parsing many languages; however, currently their lexical model may be subject to over-fitting and requires language engineering to handle out-ofvocabulary (OOV) words. Inspired by previous studies that have incorporated rich features into generative models, we propose to use a feature-rich log-linear lexical model to train PCFG-LA grammars that are more robust to rare and OOV words. The proposed lexical model has three advantages: over-fitting is alleviated via regularization, OOV words are modeled using rich features, and lexical features are exploited for grammar induction. Our approach results in significantly more accurate PCFG-LA grammars that are flexible to train for different languages (with test F scores of 90.5, 85.0, and 81.9 on WSJ, CTB6, and ATB, respectively).