Word sense disambiguation for Hungarian using transformers

In this paper we investigate the applicability of contextual word embeddings for the task of word sense disambiguation (WSD) in Hungarian. We show that a simple k–nn (k–nearest neighbors) approach which relies on multilingual BERT representations can yield highly accurate results in terms of F-score...

Full description

Saved in:
Bibliographic Details
Main Author: Berend Gábor
Corporate Author: Magyar Számítógépes Nyelvészeti Konferencia (16.) (2020) (Szeged)
Format: Book part
Published: 2020
Series:Magyar Számítógépes Nyelvészeti Konferencia 16
Kulcsszavak:Nyelvészet - számítógép alkalmazása, Szemantika
Online Access:http://acta.bibl.u-szeged.hu/67685
Description
Summary:In this paper we investigate the applicability of contextual word embeddings for the task of word sense disambiguation (WSD) in Hungarian. We show that a simple k–nn (k–nearest neighbors) approach which relies on multilingual BERT representations can yield highly accurate results in terms of F-scores when evaluated for word sense disambiguation.
Physical Description:3-13
ISBN:978-963-306-719-2