Title: Hypernym discovery for FarsNet using Hearst patterns and word embedding

Authors: Rahim Ahmadi Eslamloo; Hossein Shirazi; Maryam Hourali

Addresses: Faculty of Electrical and Computer Engineering, Malek Ashtar University of Technology, Tehran, Iran ' Faculty of Electrical and Computer Engineering, Malek Ashtar University of Technology, Tehran, Iran ' Faculty of Electrical and Computer Engineering, Malek Ashtar University of Technology, Tehran, Iran

Abstract: With the rapid growth of languages and technological innovations, new terms continuously enrich lexical resources. Dictionaries structure these concepts into synonym sets and link them through semantic relations. Many NLP applications, including translation, semantic analysis, summarisation, content creation, classification, and information retrieval, rely on FarsNet, the Persian WordNet. The lack of new lexical entries negatively impacts these applications. This study utilises multiple sources, including the Hamshahri corpus and a hypernym database, to help FarsNet identify hypernyms and semantic relations. By integrating traditional models with modern word embeddings, hypernyms are extracted from the Hamshahri sports section. This is the first such system applied to Persian. Results show promising accuracy. FarsNet, aligned with Persian's agglutinative structure and SOV word order, organises words hierarchically through hypernyms, supporting NLP tasks like translation and summarisation while enhancing Persian language processing.

Keywords: semantic relations; synonyms; hyponyms and hypernyms; FarsNet and WordNet; word embedding vectors.

DOI: 10.1504/IJAISC.2024.146625

International Journal of Artificial Intelligence and Soft Computing, 2024 Vol.8 No.4, pp.321 - 338

Received: 27 May 2024
Accepted: 02 Feb 2025
Published online: 09 Jun 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article