Feeling is understanding: From affective to semantic spaces

Elias Iosif, Alexandros Potamianos

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

Abstract

Motivated by theories of language developmentwe investigate the contribution of affect to lexical semantics in the context of distributional semantic models (DSMs). The relationship between semantic and affective spaces is computationally modeled for the task of semantic similarity computation between words. It is shown that affective spaces contain salient information for lexical semantic tasks. We further investigate specific semantic relationships where affective information plays a prominent role. The relations between semantic similarity and opposition are studied in the framework of a binary classification problemapplied for the discrimination of synonyms and antonyms. For the case of antonyms, the use of affective features results in 33% relative improvement in classification accuracy compared to the use of semantic features.

Original languageEnglish
Title of host publicationIWCS 2015 - Proceedings of the 11th International Conference on Computational Semantics
PublisherAssociation for Computational Linguistics (ACL)
Pages162-172
Number of pages11
ISBN (Electronic)9781941643334
Publication statusPublished - 2015
Externally publishedYes
Event11th International Conference on Computational Semantics, IWCS 2015 - London, United Kingdom
Duration: 15 Apr 201517 Apr 2015

Publication series

NameIWCS 2015 - Proceedings of the 11th International Conference on Computational Semantics

Conference

Conference11th International Conference on Computational Semantics, IWCS 2015
Country/TerritoryUnited Kingdom
CityLondon
Period15/04/1517/04/15

Fingerprint

Dive into the research topics of 'Feeling is understanding: From affective to semantic spaces'. Together they form a unique fingerprint.

Cite this