Web-Scale Features for Full-Scale Parsing

TitleWeb-Scale Features for Full-Scale Parsing
Publication TypeConference Paper
Year of Publication2011
AuthorsBansal, M., & Klein D.
Page(s)693-702
Other Numbers3271
Abstract

Counts from large corpora (like the web) canbe powerful syntactic cues. Past work hasused web counts to help resolve isolated ambiguities,such as binary noun-verb PP attachmentsand noun compound bracketings. Inthis work, we first present a method for generatingweb count features that address the fullrange of syntactic attachments. These featuresencode both surface evidence of lexicalaffinities as well as paraphrase-based cuesto syntactic structure. We then integrate ourfeatures into full-scale dependency and constituentparsers. We show relative error reductions

Acknowledgment

We would like to thank the anonymous reviewers for their helpful suggestions. This research is supported by BBN under DARPA contract HR0011-06-C-0022.

URLhttp://www.icsi.berkeley.edu/pubs/speech/webscalefeatures11.pdf
Bibliographic Notes

Proceedings of the 49th annual Meeting of the Association for Computational Linguistics (ACL HLT 2011), pp. 693-702, Portland, Oregon

Abbreviated Authors

M. Bansal and D. Klein

ICSI Research Group

Speech

ICSI Publication Type

Article in conference proceedings