State-of-the-Art Approaches to Word Sense Disambiguation: A Multilingual Investigation

Word sense disambiguation (WSD) is determining the correct sense of an ambiguous word from its surrounding context. It’s a fundamental problem in NLP that has wide-ranging effects on critical tasks like information retrieval, machine translation, and question-answering systems. The inherent complexity of language is like, word homonymy, polysemy, and other linguistic features, making it challenging to develop precise word sense disambiguation solutions. Our goal in this study is to find out more about the most recent developments in word sense disambiguation. Notable progress has been recently made in this field. However, dealing with languages with complex morphological structures, numerous dialects, or limited linguistic resources continues to be a challenging task for WSD. To the best of our knowledge, we present a comprehensive understanding of WSD across different languages including English, Chinese, Arabic, selected low-resource Indian languages, and Amharic. We look into various WSD techniques, including Knowledge-based, supervised, unsupervised, Deep Learning, and Hybrid approaches. The finding of this study shows that deep learning approaches and pre-trained models show promise in improving multilingual Word Sense Disambiguation, We also draw attention to the significant effort needed to address cross-lingual challenges in this field.