Loading...

One step toward a richer model of unsupervised grammar induction

Feili, H ; Sharif University of Technology | 2005

231 Viewed
  1. Type of Document: Article
  2. Publisher: Association for Computational Linguistics (ACL) , 2005
  3. Abstract:
  4. Probabilistic Context-Free Grammars (PCFGs) are useful tools for syntactic analysis of natural languages. Availability of large Treebank has encouraged many researchers to use PCFG in language modeling. Automatic learning of PCFGs is divided into three different categories, based on the needed data set for the training phase: supervised, semi-supervised and unsupervised. Most current inductive methods are supervised, which need a bracketed data set in the training phase. However, lack of this kind of data set in many languages, has encouraged us to pay more attention to unsupervised approaches. So far, unsupervised approaches have achieved little success. By considering a history-based notion, we propose an extension of the inside-outside algorithm introduced by Lari and Young. Our experiments show that inducing more conditioned grammars improves the quality of the output grammar
  5. Keywords:
  6. Context free grammars ; Modeling languages ; Natural language processing systems ; Semi-supervised learning ; Syntactics ; Automatic-learning ; Grammar induction ; Inductive method ; Natural languages ; Probabilistic context free grammars ; Semi-supervised ; Syntactic analysis ; Unsupervised approaches ; Context free languages
  7. Source: International Conference on Recent Advances in Natural Language Processing, RANLP 2005, 21 September 2005 through 23 September 2005 ; Volume 2005-January , 2005 , Pages 197-203 ; 13138502 (ISSN) ; 9549174336 (ISBN)
  8. URL: http://lml.bas.bg/ranlp2005/DOCS/RANLP2005.pdf#page=210