cantnlp@DravidianLangTech 2025: A Bag-of-Sounds Approach to Multimodal Hate Speech Detection

Sidney Wong, Andrew Li · 2025

Read cantnlp@DravidianLangTech 2025: A Bag-of-Sounds Approach to Multimodal Hate Speech Detection by Sidney Wong, Andrew Li (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-07862.txt.

Description

Sidney Wong and Andrew Li entered a shared task on hate speech in Dravidian languages with an ear-first system, in a 2025 paper. The task asked teams to flag hate in social media material that mixes text and speech, in Malayalam and Tamil among other languages. Their idea, a bag of sounds, trains on the audio alone, turning speech into pictures of its frequency content over time.

On held-out test data the system performed poorly, and the authors say so plainly. During training and development, though, the sound-only route looked promising for Malayalam and Tamil. Their claim is modest: with enough balanced training data, audio can stand alongside text in multimodal hate detection. The paper is a workshop report, valued for its negative test result and its clear next step, more and fairer data.

Catalog record

CREATORSidney Wong, Andrew Li
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-07862

Browse the full library · About the Tower of Babel