Audio Dataset Index
Audio Dataset Index Every audio dataset Daniel Rosehill publishes on Hugging Face, on this profile and in the Hebrish organisation (the name of the English-Hebrew speech project), as one tree: which dataset is a superset of which, which is a subset, and how many hours each holds. Sentence datasets (the text that is read aloud) are listed in Hebrish/Dataset-Index, which also maps datasets to models. Machine-readable: index.json and growth.csv. 64.69 hours, 46,227 takes in the… See the full description on the dataset page: https://huggingface.co/datasets/danielrosehill/Audio-Dataset-Index.
About This Project
This project serves as a centralized index for all audio datasets published under the author's profile and the Hebrish organization. It documents the relationships between various datasets, explicitly identifying superset and subset structures, and tracks the total volume of audio data in hours and number of takes. The index is designed to provide transparency into data lineage and facilitate easier navigation of speech-related resources.
The project maintains machine-readable data through index.json and growth.csv files, allowing developers to programmatically track dataset evolution and dependencies. It is specifically designed for researchers and machine learning engineers working on automatic speech recognition (ASR) projects who require a clear overview of available training data and its organizational hierarchy.