The Application of a Synonyms Dictionary Using Bi-LSTM in Bangla for Better Approximation

With the increasing internet usage, the volume of computerized data in Bangla is also growing exponentially. This vast repository of unstructured web data has various potential applications in Natural Language Processing (NLP). However, research in Bangla NLP faces significant challenges, including language complexity and limited availability of tools and datasets. These factors result in a lower accuracy rate when traditional methods used for English NLP are applied to Bangla. Specifically, the Bangla language has a rich lexicon of root words and characters, leading to an increase in out-of-vocabulary words when dealing with unseen texts. This paper proposes a novel method that employs a synonym dictionary to mitigate this problem. By converting unseen words into their synonyms that are recognized by the model, the method significantly reduces out-of-vocabulary errors. Our approach achieved an accuracy rate of 97.4%, compared to 95.3% without using the synonym dictionary, and had a loss value of 0.96. This methodology holds promise as an effective tool for solving classification problems in Bangla NLP.