Combining strategies for tagging and parsing Arabic

M Alabbas; A M Ramsay

Combining strategies for tagging and parsing Arabic

M Alabbas, A M Ramsay

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

Abstract

We describe a simple method for combiningtaggers which produces substantiallybetter performance than any of thecontributing tools. The method is verysimple, but it leads to considerable improvementsin performance: given threetaggers for Arabic whose individual accuraciesrange from 0.956 to 0.967, thecombined tagger scores 0.995–a sevenfoldreduction in the error rate whencompared to the best of the contributingtools.Given the effectiveness of this approachto combining taggers, we have investigatedits applicability to parsing. Forparsing, it seems better to take pairs ofsimilar parsers and back off to a third ifthey disagree.

Original language	English
Title of host publication	host publication
Publisher	Association for Computational Linguistics
Pages	73-77
Number of pages	5
Publication status	Published - 2014
Event	Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing - Doha Duration: 1 Jan 1824 → … http://www.aclweb.org/anthology/W14-3609

Conference

Conference	Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing
City	Doha
Period	1/01/24 → …
Internet address	http://www.aclweb.org/anthology/W14-3609

Keywords

Natural language processing
dependency parsing
part of speech tagging

Access to Document

Cite this

@inproceedings{ad35754c703845cf889d6e06b322be99,

title = "Combining strategies for tagging and parsing Arabic",

abstract = "We describe a simple method for combiningtaggers which produces substantiallybetter performance than any of thecontributing tools. The method is verysimple, but it leads to considerable improvementsin performance: given threetaggers for Arabic whose individual accuraciesrange from 0.956 to 0.967, thecombined tagger scores 0.995–a sevenfoldreduction in the error rate whencompared to the best of the contributingtools.Given the effectiveness of this approachto combining taggers, we have investigatedits applicability to parsing. Forparsing, it seems better to take pairs ofsimilar parsers and back off to a third ifthey disagree.",

keywords = "Natural language processing, dependency parsing, part of speech tagging",

author = "M Alabbas and Ramsay, {A M}",

year = "2014",

language = "English",

pages = "73--77",

booktitle = "host publication",

publisher = "Association for Computational Linguistics",

address = "United States",

note = "Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing ; Conference date: 01-01-1824",

url = "http://www.aclweb.org/anthology/W14-3609",

}

TY - GEN

T1 - Combining strategies for tagging and parsing Arabic

AU - Alabbas, M

AU - Ramsay, A M

PY - 2014

Y1 - 2014

N2 - We describe a simple method for combiningtaggers which produces substantiallybetter performance than any of thecontributing tools. The method is verysimple, but it leads to considerable improvementsin performance: given threetaggers for Arabic whose individual accuraciesrange from 0.956 to 0.967, thecombined tagger scores 0.995–a sevenfoldreduction in the error rate whencompared to the best of the contributingtools.Given the effectiveness of this approachto combining taggers, we have investigatedits applicability to parsing. Forparsing, it seems better to take pairs ofsimilar parsers and back off to a third ifthey disagree.

AB - We describe a simple method for combiningtaggers which produces substantiallybetter performance than any of thecontributing tools. The method is verysimple, but it leads to considerable improvementsin performance: given threetaggers for Arabic whose individual accuraciesrange from 0.956 to 0.967, thecombined tagger scores 0.995–a sevenfoldreduction in the error rate whencompared to the best of the contributingtools.Given the effectiveness of this approachto combining taggers, we have investigatedits applicability to parsing. Forparsing, it seems better to take pairs ofsimilar parsers and back off to a third ifthey disagree.

KW - Natural language processing

KW - dependency parsing

KW - part of speech tagging

M3 - Conference contribution

SP - 73

EP - 77

BT - host publication

PB - Association for Computational Linguistics

T2 - Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing

Y2 - 1 January 1824

ER -

Combining strategies for tagging and parsing Arabic

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this