ZeroSpeech

The 'ZeroSpeech' dataset/benchmark contains unlabelled speech data and is used to evaluate methods for zero-resource speech segmentation and lexicon learning.

Papers using ZeroSpeech (3)

Word Segmentation on Discovered Phone Units with Dynamic Programming and Self-Supervised Scoring2022 · 2 citesEvaluating context-invariance in unsupervised speech representations2022 · 1 citesRevisiting speech segmentation and lexicon learning with better features2024