Back to search

Institutional Repository of Vilnius University

Effectiveness of pre-trained morphological parsers for text classification /

Abstract

dc:description

State-of-the-art models for natural language processing (NLP), such as transformers, employ positional encodings, which, according to our hypothesis, are not as effective at extracting grammatical functions of words for highly inflected languages. This is due to highly inflected languages employing a free word order, making the relative positions of words less related to the meaning of the words, unlike in non-inflected languages. A submodule for extracting grammatical information has been constructed and pre-trained in order to aid the overall model in natural language processing tasks for highly inflected languages. Text classification tasks were performed in the Lithuanian language to evaluate the effectiveness of such a submodule. It was found that the submodule did not significantly improve the performance in solving text classification tasks, with suggestion that such a submodule may be detrimental, yet it could make the learning process more stable.

Degree

thesis:*
Grantor dc:publisher
Institutional Repository of Vilnius University
Year dc:date
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Šliogeris, Vytenis,

Subjects

dc:subject × 1

Rights

dc:rights
Statement dc:rights
  • info:eu-repo/semantics/openAccess
Language dc:language
eng

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
Vilnius University
Base URL
epublications.vu.lt/oai
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Šliogeris, Vytenis,. Effectiveness of pre-trained morphological parsers for text classification /. Institutional Repository of Vilnius University, 2024. https://repository.vu.lt/VU:ELABAETD204759825&prefLang=en_US