Open-source thesaurus development for under-resourced languages: a welsh case study

This paper introduces an open-access, userfriendly online thesaurus for the Welsh language, aimed at enriching digital resources for Welsh speakers and learners. Utilising advances in Natural Language Processing (NLP), our approach combines pre-existing word embeddings, a Welsh semantic tagger, and human evaluation to establish related terms. In this case, an initial list of 250 words was expanded
by adding 6,953 synonyms provided by linguists, creating a more extensive foundation for building the gold-standards. With this expanded list, when a user queries a particular word, the thesaurus presents all of its synonyms, allowing them to choose from a wider range of options. This is especially helpful when a
user is unsure of the exact word they want to use or wants to explore different ways to express a concept. The resulting thesaurus offers a comprehensive, reliable resource for Welsh language users, fostering enhanced communication and expression. Our work promotes Welsh NLP and showcases NLP’s potential to support under-resourced languages. The thesaurus will be accessible via a bilingual website, and the accompanying Python code will be available in a bilingual, public GitHub repository, and it will be available as a web service. Our approach presents a more efficient, cost-effective method for thesaurus creation, with potential applicability to other under-resourced languages.