Headed-Span-Based Projective Dependency Parsing

Songlin Yang, Kewei Tu


Abstract
We propose a new method for projective dependency parsing based on headed spans. In a projective dependency tree, the largest subtree rooted at each word covers a contiguous sequence (i.e., a span) in the surface order. We call such a span marked by a root word headed span. A projective dependency tree can be represented as a collection of headed spans. We decompose the score of a dependency tree into the scores of the headed spans and design a novel O(n3) dynamic programming algorithm to enable global training and exact inference. Our model achieves state-of-the-art or competitive results on PTB, CTB, and UD
Anthology ID:
2022.acl-long.155
Volume:
Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
May
Year:
2022
Address:
Dublin, Ireland
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
2188–2200
Language:
URL:
https://aclanthology.org/2022.acl-long.155
DOI:
10.18653/v1/2022.acl-long.155
Bibkey:
Cite (ACL):
Songlin Yang and Kewei Tu. 2022. Headed-Span-Based Projective Dependency Parsing. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2188–2200, Dublin, Ireland. Association for Computational Linguistics.
Cite (Informal):
Headed-Span-Based Projective Dependency Parsing (Yang & Tu, ACL 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.acl-long.155.pdf
Code
 sustcsonglin/span-based-dependency-parsing
Data
Penn Treebank