Emotion Classification through Song Lyrics in Multi-Languages with Bert
Y.-W. Peter Hong, Yuqi Xue · Applied and Computational Engineering · 2025
This research explores emotion classification in song lyrics using BERT models for multi-language datasets, focusing on English and Chinese lyrics. The study emphasizes the application of music therapy techniques by utilizing song lyrics to assist clients in expressing and processing emotions. Sentiment analysis is conducted through a combination of CNN, LSTM, and GRU models, with a GRU + CNN hybrid model demonstrating enhanced performance in multilingual contexts. A comprehensive preprocessing process, including translation processing and tokenization, enables the effective analysis of both English and Chinese lyrics. Experimental results indicate that the GRU + CNN model outperforms traditional models, particularly in cross-lingual sentiment analysis, achieving significant improvements in accuracy and emotional classification.