Avtomatska razširitev in čiščenje sloWNeta - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2014

Automatic extension and cleaning of sloWNet

Avtomatska razširitev in čiščenje sloWNeta

Abstract

In this paper we present a language-independent and automatic approach to extend a wordnet by recycling different types of already existing language resources, such as machine-readable dictionaries, parallel corpora and Wikipedia. The approach, applied to Slovene, takes into account monosemous and polysemous words, general and specialized vocabulary as well as simple and multi-word lexemes. The extracted words are assigned one or several synset ids based on a classifier that relies on several features including distributional similarity. In the next step we also identify and remove highly dubious (literal, synset) pairs, based on simple distributional information extracted from a large corpus in an unsupervised way. Automatic and manual evaluation show that the proposed approach yields very promising results.
V prispevku predstavljamo jezikovno neodvisno in avtomatsko razširitev wordneta z uporabo heterogenih že obstoječih jezikovnih virov, kot so strojno berljivi slovarji, vzporedni korpusi in Wikipedija. Pristop, ki ga preizkusimo na slovenščini, upošteva tako eno- kot večpomenske besede, splošno in specializirano besedišče, pa tudi eno- in večbesedne lekseme. Izluščenim besedam enega ali več pomenov pripišemo s pomočjo klasifikatorja, ki temelji na naboru različnih značilk, predvsem pa na distribucijski podobnosti. V naslednjem koraku s pomočjo distribucijskih informacij, izluščenih iz velikega korpusa, identificiramo in odstranimo zelo dvomljive kandidate. Avtomatska in ročna evalvacija rezultatov pokaže, da uporabljeni pristop daje zelo spodbudne rezultate.
Fichier principal
Vignette du fichier
isjt2014_07.pdf (695.67 Ko) Télécharger le fichier
Origin : Publisher files allowed on an open archive
Loading...

Dates and versions

hal-01078839 , version 1 (30-10-2014)

Identifiers

  • HAL Id : hal-01078839 , version 1

Cite

Darja Fišer, Benoît Sagot. Avtomatska razširitev in čiščenje sloWNeta. Devete konference Jezikovne Tehnologije / Ninth Language Technologies Conference, Oct 2014, Ljubljana, Slovenia. ⟨hal-01078839⟩
131 View
43 Download

Share

Gmail Facebook X LinkedIn More