Researchers at IIT Roorkee develop sentiment analysis method for Sanskrit texts

An effective technique for Sanskrit text sentiment analysis has been developed by researchers at the Indian Institute of Technology, Roorkee. The proposed method has a machine translation accuracy of 87.50% and a sentiment classification accuracy of 92.83%.

Due to a lack of enough labelled data, these technologies were not used to their full potential. The study suggested a methodology that includes sentiment analysis, machine translation, and translation evaluation models.

The cross-lingual mapping of the source and target languages has been done using machine translations. The sentences that were translated into English are as natural and mature as the original English.

The Valmiki Ramayana website, which was created and is still being updated by IIT Kanpur researchers, served as the source of the dataset for this study. To improve classification using only root words and their corresponding suffixes and prefixes, the researchers are going to study the morphological features of Sanskrit.

They also intend to assess how well the Sanskrit’s rich morphology is preserved in the English translation. In addition, they hope to acquire a model that can recognise the context of words in many languages and offer word embeddings with smaller dimensions. In the journal “Applied Intelligence,” a research paper describing the model has been published.