Leveraging Bi-Directional LSTM for Robust Lyrics Generation in Telugu: Methodology and Improvements

Abstract

The paper aims to analyze the various steps involved in creating semi-automated lyrics generators for Indian languages such as Telugu. Our study also examined the effects of bi-directional LSTM (long short-term memory) on different genres. After trying out several methods, we found that bi-directional LSTM works well with all formats. We improved our model with the help of 50,000 parameters during the training period and this is one of the largest crops for the Telugu language.