Unsupervised Dependency Parsing: Let’s Use Supervised Parsers

Phong Le; Willem Zuidema

doi:10.3115/v1/N15-1067

Unsupervised Dependency Parsing: Let’s Use Supervised Parsers

Phong Le, Willem Zuidema

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

We present a self-training approach to unsupervised dependency parsing that reuses existing supervised and unsupervised parsing algorithms. Our approach, called ‘iterated reranking’ (IR), starts with dependency trees generated by an unsupervised parser, and iteratively improves these trees using the richer probability models used in supervised parsing that are in turn trained on these trees. Our system achieves 1.8% accuracy higher than the state-of-the-part parser of Spitkovsky et al. (2013) on the WSJ corpus.

Original language	English
Title of host publication	Human Language Technologies: The 2015 Annual Conference of the North American Chapter of the ACL
Place of Publication	Denver, Colorado
Publisher	Association for Computational Linguistics
Pages	651-661
Number of pages	11
DOIs	https://doi.org/10.3115/v1/N15-1067
Publication status	Published - May 2015
Event	2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies - Denver, United States Duration: 31 May 2015 → 5 Jun 2015 http://naacl.org/naacl-hlt-2015/

Conference

Conference	2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
Abbreviated title	NAACL HLT 2015
Country/Territory	United States
City	Denver
Period	31/05/15 → 5/06/15
Internet address	http://naacl.org/naacl-hlt-2015/

Access to Document

10.3115/v1/N15-1067

http://www.aclweb.org/anthology/N15-1067

Cite this

@inproceedings{943830129a90478daa111ed6994e6d84,

title = "Unsupervised Dependency Parsing: Let{\textquoteright}s Use Supervised Parsers",

abstract = "We present a self-training approach to unsupervised dependency parsing that reuses existing supervised and unsupervised parsing algorithms. Our approach, called {\textquoteleft}iterated reranking{\textquoteright} (IR), starts with dependency trees generated by an unsupervised parser, and iteratively improves these trees using the richer probability models used in supervised parsing that are in turn trained on these trees. Our system achieves 1.8% accuracy higher than the state-of-the-part parser of Spitkovsky et al. (2013) on the WSJ corpus.",

author = "Phong Le and Willem Zuidema",

year = "2015",

month = may,

doi = "10.3115/v1/N15-1067",

language = "English",

pages = "651--661",

booktitle = "Human Language Technologies: The 2015 Annual Conference of the North American Chapter of the ACL",

publisher = "Association for Computational Linguistics",

note = "2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , NAACL HLT 2015 ; Conference date: 31-05-2015 Through 05-06-2015",

url = "http://naacl.org/naacl-hlt-2015/",

}

Le, P & Zuidema, W 2015, Unsupervised Dependency Parsing: Let’s Use Supervised Parsers. in Human Language Technologies: The 2015 Annual Conference of the North American Chapter of the ACL. Association for Computational Linguistics, Denver, Colorado, pp. 651-661, 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , Denver, Colorado, United States, 31/05/15. https://doi.org/10.3115/v1/N15-1067

TY - GEN

T1 - Unsupervised Dependency Parsing: Let’s Use Supervised Parsers

AU - Le, Phong

AU - Zuidema, Willem

PY - 2015/5

Y1 - 2015/5

N2 - We present a self-training approach to unsupervised dependency parsing that reuses existing supervised and unsupervised parsing algorithms. Our approach, called ‘iterated reranking’ (IR), starts with dependency trees generated by an unsupervised parser, and iteratively improves these trees using the richer probability models used in supervised parsing that are in turn trained on these trees. Our system achieves 1.8% accuracy higher than the state-of-the-part parser of Spitkovsky et al. (2013) on the WSJ corpus.

AB - We present a self-training approach to unsupervised dependency parsing that reuses existing supervised and unsupervised parsing algorithms. Our approach, called ‘iterated reranking’ (IR), starts with dependency trees generated by an unsupervised parser, and iteratively improves these trees using the richer probability models used in supervised parsing that are in turn trained on these trees. Our system achieves 1.8% accuracy higher than the state-of-the-part parser of Spitkovsky et al. (2013) on the WSJ corpus.

U2 - 10.3115/v1/N15-1067

DO - 10.3115/v1/N15-1067

M3 - Conference contribution

SP - 651

EP - 661

BT - Human Language Technologies: The 2015 Annual Conference of the North American Chapter of the ACL

PB - Association for Computational Linguistics

CY - Denver, Colorado

T2 - 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies

Y2 - 31 May 2015 through 5 June 2015

ER -

Unsupervised Dependency Parsing: Let’s Use Supervised Parsers

Abstract

Conference

Access to Document

Fingerprint

Cite this