Tagging Speech For Words In Low Resourced Monolingual Contexts of Sanskrit Shlokas
Thanmayi S. Hegde, V Vindhya, V R Badri Prasad · 2023
Natural Language Processing is widely used in Machine translation which helps in replacing many tedious and difficult tasks and producing qualitative and efficient results. Sanskrit is one of the primitive languages of India. It is also the oldest, purest and most systematic language in the world. It is a magical language which helps in generating different words from a single word by prefixing and suffixing the word with different helper words to same root word. Shloka in Sanskrit are phrases or words that represent a poem or a hymn. The famous texts written entirely in shlokas are the “Ramayana” and “Mahabharata.” This paper aims to generate the POS tags for each word in Sanskrit Shlokas using various machine learning and deep learning algorithms thus helping in understanding the actual verbal beauty of Sanskrit Shlokas in widely spoken language English. The proposed work implements Parts of Speech Tagging for words in Sanskrit Shlokas using Conditional Random Field, Gaussian NB and Bi-directional LSTM models. Of the implemented approaches, Bi-directional LSTM gives the best result.