Context sensitive paraphrasing with a global unsupervised classifier

Michael Connor, Dan Roth

Research output: Chapter in Book/Report/Conference proceedingConference contribution


Lexical paraphrasing is an inherently context sensitive problem because a word's meaning depends on context. Most paraphrasing work finds patterns and templates that can replace other patterns or templates in some context, but we are attempting to make decisions for a specific context. In this paper we develop a global classifier that takes a word v and its context, along with a candidate word u, and determines whether u can replace v in the given context while maintaining the original meaning. We develop an unsupervised, bootstrapped, learning approach to this problem. Key to our approach is the use of a very large amount of unlabeled data to derive a reliable supervision signal that is then used to train a supervised learning algorithm. We demonstrate that our approach performs significantly better than state-of-the-art paraphrasing approaches, and generalizes well to unseen pairs of words.

Original languageEnglish (US)
Title of host publicationMachine Learning
Subtitle of host publicationECML 2007 - 18th European Conference on Machine Learning, Proceedings
Number of pages12
ISBN (Print)9783540749578
StatePublished - 2007
Externally publishedYes
Event18th European Conference on Machine Learning, ECML 2007 - Warsaw, Poland
Duration: Sep 17 2007Sep 21 2007

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4701 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Other18th European Conference on Machine Learning, ECML 2007

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science


Dive into the research topics of 'Context sensitive paraphrasing with a global unsupervised classifier'. Together they form a unique fingerprint.

Cite this