Paper: Head-Driven PCFGs With Latent-Head Statistics

ACL ID W05-1512
Title Head-Driven PCFGs With Latent-Head Statistics
Venue Workshop On Parsing Technology
Year 2005

Although state-of-the-art parsers for nat- ural language are lexicalized, it was re- cently shown that an accurate unlexical- ized parser for the Penn tree-bank can be simply read off a manually refined tree- bank. While lexicalized parsers often suf- fer from sparse data, manual mark-up is costly and largely based on individual lin- guistic intuition. Thus, across domains, languages, and tree-bank annotations, a fundamental question arises: Is it possible to automatically induce an accurate parser from a tree-bank without resorting to full lexicalization? In this paper, we show how to induce head-driven probabilistic parsers with latent heads from a tree-bank. Our automatically trained parser has a perfor- mance of 85.7% (LP/LR F1), which is al- ready better than that of early lexicalized...

  author    = {Prescher, Detlef},
  title     = {Head-Driven {PCFGs} with Latent-Head Statistics},
  booktitle = {Proceedings of the Ninth International Workshop on Parsing Technology},
  month     = {October},
  year      = {2005},
  address   = {Vancouver, British Columbia},
  publisher = {Association for Computational Linguistics},
  pages     = {115--124},
  url       = {}