A Bibliometric Analysis of Embedding Techniques for Addressing Meaning Conflation Deficiency in Low-Resourced Languages

Meaning Conflation Deficiency (MCD): the phenomenon whereby a singular vector representation amalgamates many meanings of a word—continues to pose a significant difficulty in Natural Language Processing, especially for morphologically complex and low-resourced languages. This shortcoming obstructs precise semantic comprehension, particularly in activities such as word meaning disambiguation and machine translation. This study performs a bibliometric analysis of academic research from 2014 to 2025 focused on embedding strategies aimed at addressing MCD. Using data sourced from Scopus and Web of Science and visualised via VOSviewer, the research delineates significant publishing patterns, identifies prominent authors and journals, and explores topic progression within embedding paradigms. Findings indicate a distinct evolution from static embeddings (e.g., Word2Vec, GloVe) to advanced context-aware and multi-sense models like ELMo, BERT, and multilingual transformers. The results highlight an increasing interest in graph-based and hybrid embeddings specifically designed for the linguistic intricacies of low-resourced languages. This work serves as a thorough reference for future research focused on creating more effective embedding techniques to alleviate meaning conflation in various linguistic contexts.