Institutional Repository of Vilnius University
Lietuvių literatūros tekstais apmokytų „GloVe“ ir „Word2Vec“ modelių palyginimas ir pritaikymas /
Abstract
dc:descriptionIn this work, an analysis of the word embedding models “Word2Vec Skip-gram”, “Word2Vec Continuous Bag-of-Words” and “GloVe” is performed. These models were trained on Lithuanian literary works in order to apply them in the development of an educational tool. It was compared which models are more effective in capturing semantic similarity between words. For the purpose of the comparison, an adaptation of the “MEN” word pair set to the Lithuanian language was carried out. After conducting the comparison, a tendency was observed that, in the context of Lithuanian literature, the semantic properties of words are better captured by word embeddings constructed by “Word2Vec” models, which use the surrounding context, rather than “GloVe” models, which are based on word co-occurrence statistics. Models trained on Lithuanian literary works were used to develop an educational tool - a web-based fill-in-the-blank application for omitted words in Lithuanian literary works.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Gilys, Ignas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192846423&prefLang=en_US