A novel approach of word sense disambiguation for marathi language using machine learning
Ujwalla Gawande, Swati G. Kale, Chetana B. Thaokar · 2023
Semantic understanding is the key issue in natural language processing (NLP) systems. A single word with multiple interpretations is a common feature of NLP. Linguistic and structural ambiguity makes speech or written text open to multiple possible interpretations. Humans use the references provided by the world and language communities to solve the task of disambiguation. However, the machine does not have this capability. Machine translation, question answering systems, and information extraction and retrieval are some of the applications of NLP, and their accuracy depends on the accuracy achieved by the word sense disambiguation (WSD) system. The most challenging problem with WSD is to identify, which sense or meaning of a word is used in a particular sentence. There are other problems with predicting the correct word meaning, such as errors in this process and more time consuming. In this paper, we propose a WSD technique using Word’s multiple features-based approach for Marathi language. It is one of the morphologically rich languages in India. In this technique, a multi-class classifier is developed that will predict the exact sense of ambiguous words in each context. We conducted all the experiments on the benchmark WordNet dataset developed at Indian Institute of Technology Bombay (IIT Bombay). Experimental results show significant performance improvements in the area of NLP.