Similarity of Semantic Relations
.
(2006)cite arxiv:cs.CL/0608100Comment: related work available at http://purl.org/peter.turney/.

There are at least two kinds of similarity. Relational similarity iscorrespondence between relations, in contrast with attributional similarity,which is correspondence between attributes. When two words have a high degreeof attributional similarity, we call them synonyms. When two pairs of wordshave a high degree of relational similarity, we say that their relations areanalogous. For example, the word pair mason:stone is analogous to the paircarpenter:wood. This paper introduces Latent Relational Analysis (LRA), amethod for measuring relational similarity. LRA has potential applications inmany areas, including information extraction, word sense disambiguation, andinformation retrieval. Recently the Vector Space Model (VSM) of informationretrieval has been adapted to measuring relational similarity, achieving ascore of 47% on a collection of 374 college-level multiple-choice word analogyquestions. In the VSM approach, the relation between a pair of words ischaracterized by a vector of frequencies of predefined patterns in a largecorpus. LRA extends the VSM approach in three ways: (1) the patterns arederived automatically from the corpus, (2) the Singular Value Decomposition(SVD) is used to smooth the frequency data, and (3) automatically generatedsynonyms are used to explore variations of the word pairs. LRA achieves 56% onthe 374 analogy questions, statistically equivalent to the average human scoreof 57%. On the related problem of classifying semantic relations, LRA achievessimilar gains over the VSM.
  • @benz
Diese Publikation wurde noch nicht bewertet.

Bewertungsverteilung
Durchschnittliche Benutzerbewertung0,0 von 5.0 auf Grundlage von 0 Rezensionen
    Bitte melden Sie sich an um selbst Rezensionen oder Kommentare zu erstellen.