Skip to main navigation Skip to search Skip to main content

Enriching SMT Training Data via Paraphrasing

  • Wei He
  • , Shiqi Zhao
  • , Haifeng Wang
  • , Ting Liu
  • Harbin Institute of Technology
  • Baidu Inc

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper proposes a novel method to resolve the coverage problem of SMT system. The method generates paraphrases for source-side sentences of the bilingual parallel data, which are then paired with the target-side sentences to generate new parallel data. Within a statistical paraphrase generation framework, we employ an object function, named Sentence Novelty, to select paraphrases which having the most novel information to the bilingual training corpus of the SMT model. Meanwhile, the context is considered via a language model in the source language to ensure the fluency and accuracy of paraphrase substitution. Compared to a state-of-the-art phrase based SMT system (Moses), our method achieves an improvement of 1.66 points in terms of BLEU on a small training corpus which simulates a resource-poor environment, and 1.06 points on a training corpus of medium size.

Original languageEnglish
Title of host publicationIJCNLP 2011 - Proceedings of the 5th International Joint Conference on Natural Language Processing
EditorsHaifeng Wang, David Yarowsky
PublisherAssociation for Computational Linguistics (ACL)
Pages803-810
Number of pages8
ISBN (Electronic)9789744665645
StatePublished - 2011
Event5th International Joint Conference on Natural Language Processing, IJCNLP 2011 - Chiang Mai, Thailand
Duration: 8 Nov 201113 Nov 2011

Publication series

NameIJCNLP 2011 - Proceedings of the 5th International Joint Conference on Natural Language Processing

Conference

Conference5th International Joint Conference on Natural Language Processing, IJCNLP 2011
Country/TerritoryThailand
CityChiang Mai
Period8/11/1113/11/11

Fingerprint

Dive into the research topics of 'Enriching SMT Training Data via Paraphrasing'. Together they form a unique fingerprint.

Cite this