DLS@CU: Sentence Similarity from Word Alignment

Md Arafat Sultan, Steven Bethard, Tamara Sumner

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

59 Scopus citations

Abstract

We present an algorithm for computing the semantic similarity between two sentences. It adopts the hypothesis that semantic similarity is a monotonically increasing function of the degree to which (1) the two sentences contain similar semantic units, and (2) such units occur in similar semantic contexts. With a simplistic operationalization of the notion of semantic units with individual words, we experimentally show that this hypothesis can lead to state-of-the-art results for sentence-level semantic similarity. At the SemEval 2014 STS task (task 10), our system demonstrated the best performance (measured by correlation with human annotations) among 38 system runs.

Original languageEnglish (US)
Title of host publication8th International Workshop on Semantic Evaluation, SemEval 2014 - co-located with the 25th International Conference on Computational Linguistics, COLING 2014, Proceedings
EditorsPreslav Nakov, Torsten Zesch
PublisherAssociation for Computational Linguistics (ACL)
Pages241-246
Number of pages6
ISBN (Electronic)9781941643242
StatePublished - 2014
Externally publishedYes
Event8th International Workshop on Semantic Evaluation, SemEval 2014 - Dublin, Ireland
Duration: Aug 23 2014Aug 24 2014

Publication series

Name8th International Workshop on Semantic Evaluation, SemEval 2014 - co-located with the 25th International Conference on Computational Linguistics, COLING 2014, Proceedings

Conference

Conference8th International Workshop on Semantic Evaluation, SemEval 2014
Country/TerritoryIreland
CityDublin
Period8/23/148/24/14

ASJC Scopus subject areas

  • Computational Theory and Mathematics
  • Language and Linguistics
  • Linguistics and Language

Fingerprint

Dive into the research topics of 'DLS@CU: Sentence Similarity from Word Alignment'. Together they form a unique fingerprint.

Cite this