Shared Task on Multimodal content moderation in low-resource languages
This shared task focuses on multimodal content moderation in low-resource languages, with a particular emphasis on Nepali memes. The objective is to develop systems that can automatically determine whether a meme is hateful or not, as well as classify its sentiment into positive, negative, or neutral categories. The task will include six subtasks: (i) hate speech detection in Nepali-only memes, (ii) sentiment analysis in Nepali-only memes, (iii) hate speech detection in code-mixed Nepali memes, (iv) sentiment analysis in code-mixed Nepali memes, (v) hate speech detection in code-switched Nepali memes, and (vi) sentiment analysis in code-switched Nepali memes. The competition will be hosted on CodaLab, where participants will receive training and test sets, and system performance will be evaluated and ranked using the F1-score. A dataset for this task has already been prepared and published in ICWSM 2025 (Thapa, Naseem, Lee, & Razzak, 2024) and The Web Conference Companion 2025 (Thapa, Veeramani, Razzak, Lee, & Naseem, 2025).
Shared Task on Multlingual ASR for South Asian Languages
The shared task focuses on advancing Automatic Speech Recognition (ASR) for the linguistically diverse landscape of South Asia, explicitly tackling the challenge of resource disparity. We target a carefully selected set of eight languages: Balti, Bengali, Dhivehi, Marathi, Nepali, Saraiki, Tamil, and Torwali. These are selected from different regions and language families which are a mix of high, moderate, and low (speech) resources. ones from different families and regions, to foster robust, generalizable model development. The core task will involve building ASR systems for these languages with a primary evaluation on seen languages and a generalization challenge on one unseen language. The challenge will include development of models having large and small numbers of parameters to include small models (that can be used in mobile devices) in the competition.
For details, visit Shared Task 2.
Await more details...