Skip to main navigation Skip to search Skip to main content

Making semantic topicality robust through term abstraction

  • SUNY Buffalo

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

Abstract

Despite early intuitions, semantic similarity has not proven to be robust for splitting multiparty interactions into separate conversations. We discuss some initial successes with using thesaural headwords to abstract the semantics of an utterance. This simple profiling technique showed improvements over baseline conversation threading models.

Original languageEnglish
Title of host publicationSEW 2009 - Semantic Evaluations
Subtitle of host publicationRecent Achievements and Future Directions at the 2009 North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL HLT 2009 - Proceedings
EditorsMari Ostendorf, Michael Collins, Shri Narayanan, Douglas W. Oard, Lucy Vanderwende
PublisherAssociation for Computational Linguistics (ACL)
Pages46-51
Number of pages6
ISBN (Electronic)9781932432312
StatePublished - 2009
Event2009 Semantic Evaluations: Recent Achievements and Future Directions, SEW 2009 at the 2009 North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL HLT 2009 - Boulder, United States
Duration: Jun 4 2009 → …

Publication series

NameSEW 2009 - Semantic Evaluations: Recent Achievements and Future Directions at the 2009 North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL HLT 2009 - Proceedings

Conference

Conference2009 Semantic Evaluations: Recent Achievements and Future Directions, SEW 2009 at the 2009 North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL HLT 2009
Country/TerritoryUnited States
CityBoulder
Period06/4/09 → …

Fingerprint

Dive into the research topics of 'Making semantic topicality robust through term abstraction'. Together they form a unique fingerprint.

Cite this