Emotion Classification in Thai music using Convolutional Neural Networks
Sitdhibong Laokok, Subhorn Khonthapagdee · 2022
Music is an integral aspect of human life. We use music for a variety of objectives, including enjoyment, treatment, and inspiration. This study is interested in evaluating song extraction utilizing natural language processing and deep learning techniques. To this end, we propose a framework for analyzing Thai music emotion based on the Hourglass which classify mood of music into 5 categories: Disapproval, frustration, love, optimism and remorse. First, we explain how to determine mood label for each song using BabelSenticNet corpus. We believe that using mood labels from BabelSenticNet is comparable to using humans to determine the mood of each song. In order to automate song classification without semantic knowledge, Fasttext was used to embed song lyrics and fed into Convolutional Neural Networks. Learning rates were adjust to find the best accuracy.