cantnlp@DravidianLangTech2025: A Bag-of-Sounds Approach to Multimodal Hate Speech Detection
This work addresses hate speech detection for Dravidian language speakers, but it is incremental as it applies an existing method to new data with limited success.
The paper tackled multimodal hate speech detection in Dravidian languages by using a bag-of-sounds approach with Mel spectrograms on audio data, but the model performed poorly on the test set while showing promise during training for Malayalam and Tamil.
This paper presents the systems and results for the Multimodal Social Media Data Analysis in Dravidian Languages (MSMDA-DL) shared task at the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages (DravidianLangTech-2025). We took a `bag-of-sounds' approach by training our hate speech detection system on the speech (audio) data using transformed Mel spectrogram measures. While our candidate model performed poorly on the test set, our approach offered promising results during training and development for Malayalam and Tamil. With sufficient and well-balanced training data, our results show that it is feasible to use both text and speech (audio) data in the development of multimodal hate speech detection systems.