| Abstract |
The paper details the current state of an ongoing collaboration between Hungarian lexicographers and computational linguists. Our goal is to provide a comprehensive and consistent description of Hungarian adjectives, benefiting lexical semantics, lexicography and NLP. This thread of research focuses on identifying systematic semantic patterns of Hungarian adjectives and their typical subcategorization frames, with a particular emphasis on polysemous meanings. The proposed methodology is entirely unsupervised, reducing reliance on human intuition. It is based on a graph representation derived from adjectival static embeddings. The algorithm models adjectival semantic domains by specific subgraphs, namely, connected graph components. In the next step, potential subcategorization frames for the detected adjectival semantic domains, so called meaning structures, are also derived from corpus data. Then, a sample of the meaning structures is compared to the entries of the Concise Dictionary of Hungarian, evaluating the pros and cons of the proposed algorithm. Finally, as a further improvement, the automatically derived subcategorization frames were generalized. |
| BibTex |
@inproceedings{euralex_2024_paper_20,
address = {Cavtat},
title = {Graph-based Detection of Hungarian Adjectival Meaning Structures via Monolingual Static Embeddings},isbn = {978-953-7967-77-2},
shorttitle = {Euralex 2024},
url = {},
language = {eng},
booktitle = {Lexicography and Semantics. Proceedings of the XXI EURALEX International Congress},
publisher = {Institut za hrvatski jezik},
author = {Héja, Enikő and Gábor, Kata and Simon, László and Lipp, Veronika},
editor = {Despot, Kristina Š. and Ostroški Anić, Ana and Brač, Ivana},
year = {2024},
pages = {251-265}
} |
