XTREME-S is a benchmark for evaluating universal cross-lingual speech representations in many languages. XTREME-S consists of:
📚 Paper
🤗 Datasets
🏆 Leaderboard and Submission
XTREME-S covers four task families: speech recognition, classification, speech-to-text translation and retrieval. Covering 102 languages from 10+ language families, 3 different domains and 4 task families, XTREME-S aims to simplify multilingual speech representation evaluation, as well as catalyze research in "universal" speech representation learning.
XTREME-S was proposed in the paper XTREME-S: Evaluating Cross-lingual Speech Representations by Conneua et. al. in 2022. For more information, see the official paper on Arxiv.