VoxLingua107 is a comprehensive speech dataset designed for training spoken language identification models. It comprises short speech segments sourced from YouTube videos, labeled based on the language indicated in the video title and description. The dataset covers 107 languages and contains a total of 6628 hours of speech data, averaging 62 hours per language. However, the actual amount of data per language varies significantly. Additionally, there is a separate development set consisting of 1609 speech segments from 33 languages, validated by at least two volunteers to ensure the accuracy of language representation.
No results indexed yet — be the first to submit a score.
Submit a checkpoint and a reproduction script. We will run it, publish the score, and — if it takes the top — annotate the step on the progress chart with your name.