Probabilistic  CFG with Latent Annotations

DMCA

Probabilistic CFG with Latent Annotations (2005)

Cached

Download Links

by Takuya Matsuzaki , Yusuke Miyao , Jun’ichi Tsujii

Citations:

101 - 1 self

BibTeX

@MISC{Matsuzaki05probabilisticcfg,
    author = {Takuya Matsuzaki and Yusuke Miyao and Jun’ichi Tsujii},
    title = {Probabilistic CFG with Latent Annotations},
    year = {2005}
}

OpenURL

Abstract

This paper defines a generative probabilistic model of parse trees, which we call PCFG-LA. This model is an extension of PCFG in which non-terminal symbols are augmented with latent variables. Finegrained CFG rules are automatically induced from a parsed corpus by training a PCFG-LA model using an EM-algorithm. Because exact parsing with a PCFG-LA is NP-hard, several approximations are described and empirically compared. In experiments using the Penn WSJ corpus, our automatically trained model gave a performance of 86.6 % (F ¥ , sentences ¦ 40 words), which is comparable to that of an unlexicalized PCFG parser created using extensive manual feature selection.

Keyphrases

latent annotation probabilistic cfg pcfg-la model generative probabilistic model non-terminal symbol cfg rule latent variable extensive manual feature selection several approximation parse tree penn wsj corpus unlexicalized pcfg parser

DMCA