Dependency grammar induction via bitext projection constraints

DMCA

Dependency grammar induction via bitext projection constraints (2009)

Cached

Download Links

by Kuzman Ganchev , Jennifer Gillenwater , Ben Taskar

Venue:	In ACL-IJCNLP
Citations:	35 - 5 self

BibTeX

@INPROCEEDINGS{Ganchev09dependencygrammar,
    author = {Kuzman Ganchev and Jennifer Gillenwater and Ben Taskar},
    title = {Dependency grammar induction via bitext projection constraints},
    booktitle = {In ACL-IJCNLP},
    year = {2009}
}

OpenURL

Abstract

Broad-coverage annotated treebanks necessary to train parsers do not exist for many resource-poor languages. The wide availability of parallel text and accurate parsers in English has opened up the possibility of grammar induction through partial transfer across bitext. We consider generative and discriminative models for dependency grammar induction that use word-level alignments and a source language parser (English) to constrain the space of possible target trees. Unlike previous approaches, our framework does not require full projected parses, allowing partial, approximate transfer through linear expectation constraints on the space of distributions over trees. We consider several types of constraints that range from generic dependency conservation to language-specific annotation rules for auxiliary verb analysis. We evaluate our approach on Bulgarian and Spanish CoNLL shared task data and show that we consistently outperform unsupervised methods and can outperform supervised learning for limited training data. 1

Keyphrases

dependency grammar induction bitext projection constraint source language parser generic dependency conservation spanish conll parallel text partial transfer possible target tree word-level alignment language-specific annotation rule grammar induction limited training data task data approximate transfer several type discriminative model many resource-poor language linear expectation constraint auxiliary verb analysis unsupervised method wide availability previous approach accurate parser supervised learning

DMCA