This repo contains the data for our paper YUE-PUB-Speech: A Speech-based Pragmatic Understanding Benchmark for Cantonese (Interspeech 2026).
The benchmark consists of 14 tasks collected from 7 existing pragmatic datasets, covering different forms of pragmatic reasoning, including response classification… See the full description on the dataset page:
https://huggingface.co/datasets/Multilingual-NLP/YUE-PUB-Speech.