Non-projective dependency parsing using spanning tree algorithms

DMCA

Non-projective dependency parsing using spanning tree algorithms (2005)

Cached

Download Links

Other Repositories/Bibliography

DBLP

by Ryan Mcdonald , Fernando Pereira , Kiril Ribarov , Jan Hajič

Venue:	In Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing
Citations:	383 - 10 self

BibTeX

@INPROCEEDINGS{Mcdonald05non-projectivedependency,
    author = {Ryan Mcdonald and Fernando Pereira and Kiril Ribarov and Jan Hajič},
    title = {Non-projective dependency parsing using spanning tree algorithms},
    booktitle = {In Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing},
    year = {2005},
    pages = {523--530}
}

OpenURL

Abstract

We formalize weighted dependency parsing as searching for maximum spanning trees (MSTs) in directed graphs. Using this representation, the parsing algorithm of Eisner (1996) is sufficient for searching over all projective trees in O(n 3) time. More surprisingly, the representation is extended naturally to non-projective parsing using Chu-Liu-Edmonds (Chu and Liu, 1965; Edmonds, 1967) MST algorithm, yielding an O(n 2) parsing algorithm. We evaluate these methods on the Prague Dependency Treebank using online large-margin learning techniques (Crammer et al., 2003; McDonald et al., 2005) and show that MST parsing increases efficiency and accuracy for languages with non-projective dependencies. 1

Keyphrases

non-projective dependency tree algorithm weighted dependency mst algorithm online large-margin learning technique increase efficiency projective tree parsing algorithm prague dependency treebank maximum spanning tree

DMCA