TY - JOUR
AU1 - Dhar, Sourish
AU2 - Gour, Vishal
AU3 - Paul, Arnab
AB - Emotion recognition is a process of extracting and analysing human emotion from text, audio, and video. Emotion recognition from song lyrics is gaining attention from the research community. However, very limited efforts have been made in low resource Indian languages due to the lack of benchmark datasets and tools. This paper makes an attempt in recognising and categorising the emotions arising from the lyrical text of Hindi songs. Most of the song emotion classification approaches deal with a maximum of three categories. This paper explores the concept of Navrasa of Indian aesthetics for categorising emotions from the song into nine categories. In this work, five different feature variations of a particular song’s lyrics are extracted. The feature vectors are represented using TF-IDF and Doc2vec. Four different classifiers namely Support Vector Machines, Logistic Regression, Multinomial Naïve Bayes, and K-Nearest Neighbours are used for classifying the emotions of the song. From the experiments, it is observed that the performance of the SVM classifier achieved the highest accuracy of 66.7%.
TI - Emotion recognition from lyrical text of Hindi songs
JF - Innovations in Systems and Software Engineering
DO - 10.1007/s11334-022-00520-z
DA - 2025-03-01
UR - https://www.deepdyve.com/lp/springer-journals/emotion-recognition-from-lyrical-text-of-hindi-songs-7OzHUDyA3A
SP - 227
EP - 235
VL - 21
IS - 1
DP - DeepDyve
ER -