CONTEXTUAL LEARNING APPROACHES FOR WORD SENSE DISAMBIGUATION

Abstract

Abstract: Word Sense Disambiguation (WSD) is a fundamental task in Natural Language Processing (NLP) that aims to identify the correct meaning of polysemous words based on their contextual usage. Accurate WSD is essential for improving the effectiveness of downstream NLP applications such as machine translation, information retrieval, text summarisation, and sentiment analysis. Recent advances in deep learning, particularly transformer-based architectures and contextualised word embeddings, have significantly redefined performance benchmarks in WSD by enabling richer semantic representation and improved generalisation across domains. This paper presents a structured study of contextual and deep learning approaches for Word Sense Disambiguation, including recurrent neural networks, convolutional models, transformer-based approaches, few-shot learning, and hybrid architectures. In addition, language-specific implementations, domain-adapted methods, and approaches for low-resource settings are analysed. Finally, the paper discusses open research challenges and highlights promising directions for future deep learning–driven WSD research. Keywords—Word Sense Disambiguation, Deep Learning, BERT, BiLSTM, Transformer, Low-Resource Languages, NLP