Bhaashaadarshi: A Novel Computational Language Model for Indian Languages with a Case Study in Kannada

This paper presents ‘Bhaashaadarshi’, a novel computational language model for all Indian languages. It consists of a new numerical representation and the distance and word alignment algorithms for measuring word similarity. This model is capable of connecting the grammar and phonetics of Indian languages, thus unifying the written and spoken Indian language models. In this paper, we explain the numerical representation, the distance and the alignment algorithms, considering Kannada language as the case study. This model is trained, tested with a few text mining applications, and evaluated manually using our private Kannada data corpus. ‘Bhaashaadarshi’ is a simple and efficient model compared to many AI -driven language models. It can address the ‘curse of multi linguality’ and can be used in Large Language Models.