A Separate-and-Learn Approach to EM Learning of PCFGs

WeproposeanewapproachtoEMlearning of PCFGs. We completely separate the process of EM learning from that of parsing, andfor theformer, weintroduce a new EM algorithm called the graphical EM algorithm that runs on a new data structurecalledsupport graphsextracted from WFSTs (well formed substring tables) of various parsers. Learning experiments with PCFGs using two Japanese corpora indicate that our approach can signicantly outperform the existing approaches using the Inside-Outside algorithm (Baker, 1979) and Stolcke's EM algorithm (Stolcke, 1995).

• Type: Conference
• Year: 2001
• Where: NLPRS
• Authors: Taisuke Sato, Shigeru Abe, Yoshitaka Kameya, Kiyoaki Shirai
• Keywords: EM Algorithm, Graphical Em Algorithm, Inside-Outside Algorithm, Natural Language Processing, NLPRS 2001