Artificial Intelligence Applications and Reconfigurable Architectures

5NLP-Based AI-Powered Sanskrit Voice Bot

Vedika Srivastava¹, Arti Khaparde¹, Akshit Kothari²* and Vaidehi Deshmukh¹

¹School of Electronics and Communication Engineering, MIT World Peace University, Pune, India

²Tech Mahindra Maker’s Lab, Pune, India

Abstract

The rapid evolution of human beings as a species can be credited to their ability to commune with one another and efficiently drive ideas, messages and intent past each other. One of the antediluvian and well-structured languages, Sanskrit, is being relegated only to use in scriptures during modern times. Our intent is to build a virtual assistant (voice/chat) which communicates through Sanskrit ensuring this language becomes the linchpin of understanding machines and relaying information and knowledge not only for an extensive heterogeneity of vernacular population but for the world. Studying various Machine Learning and Neural Network models, understanding their scope, underlying principles and application hence facilitating deep understanding of the scope of AI Assistants and aid in building a Sanskrit Voice Bot. Various algorithm explore include linear regression and logistic regression, whose reach is limited to linearly related/separable data, which was test by deploying gradient descent algorithm. Support Vector Machine kernels resolve this problem by providing linear as well as polynomial decision boundary. Principal Component Analysis finds its major application in dimensionality reduction and Anomaly Detection ...