Good, Great, Excellent: Global Inference of Lexical Intensities

TitleGood, Great, Excellent: Global Inference of Lexical Intensities
Publication TypeJournal Article
Year of Publication2013
Authorsde Melo, G., & Bansal M.
Published inTransactions of the Association for Computational Linguistics
Page(s)279-290
Other Numbers3401
Abstract

Adjectives like good, great, and excellent aresimilar in meaning, but differ in intensity. Intensityorder information is very useful forlanguage learners as well as in several NLPtasks, but is missing in most lexical resources(dictionaries, WordNet, and thesauri). In thispaper, we present a primarily unsupervisedapproach that uses semantics from Web-scaledata (e.g., phrases like good but not excellent)to rank words by assigning them positionson a continuous scale. We rely on MixedInteger Linear Programming to jointly determinethe ranks, such that individual decisionsbenefit from global information. When rankingEnglish adjectives, our global algorithmachieves substantial improvements over previouswork on both pairwise and rank correlation

Acknowledgment

This work was partially funded by the Deutscher Akademischer Austausch Dienst (DAAD) through a postdoctoral fellowship.

URLhttps://www.icsi.berkeley.edu/pubs/ai/ICSI_goodgreatexcellent13.pdf
Bibliographic Notes

Transactions of the Association for Computational Linguistics, ed. L. Lee, pp. 279-290.

Abbreviated Authors

G. de Melo and M. Bansal

ICSI Research Group

AI

ICSI Publication Type

Article in journal or magazine

Reprint Editioncode and data freely available.</P>