Comparative Study on Different Approaches Used in Word Sense Disambiguation—NLP

Our research integrates word semantic disambiguation (WSD), a pivotal natural language processing (NLP) approach. This article comprehensively explores the evolution of WSD methodologies, encompassing statistical, machine learning, and rule-based procedures. The study meticulously evaluates the performance of supervised models in Word Sense Disambiguation, scrutinizing their effectiveness across diverse parameters. A significant facet of proposed study involves an in-depth examination of the preprocessing stage, wherein various techniques are applied to the data. Notably, this study identified marked improvements in results and accuracies, particularly in the case of Naïve Bayes and Support Vector Machines. The accuracy of Naïve Bayes surged from 83.4% to an impressive 91%, while Support Vector Machines exhibited a notable rise from 81.01 to 81.8%. Findings underscore the pivotal role of preprocessing in refining the outcomes of WSD models and contribute valuable insights to the field of natural language understanding and semantic disambiguation. Furthermore, this study finds that supervised algorithms perform better than unsupervised algorithms, confirming their superiority in the WSD context. The results highlight how important preprocessing is to fine-tuning WSD model results and offer insightful new perspectives on the development of semantic disambiguation and natural language understanding in the broad field of NLP research.