Taxonomy Of AI
Analysis of Hugging Face audio-to-audio models: full Hub API metadata snapshot + refresh script
About This Project
Taxonomy-Of-AI addresses the limitations of broad Hugging Face pipeline tags by providing a structured, granular classification system for machine learning models. The project automates the retrieval of model snapshots from the Hub and facilitates a hybrid workflow where users can define canonical subcategories for specific tasks, such as differentiating between various audio-to-audio applications like speech enhancement and voice conversion.
Built with Python, the toolkit includes scripts for data synchronization, tag aggregation, and integration with Label Studio for human-in-the-loop annotation. It serves as a personal research utility for developers and data scientists who require a more precise inventory of model capabilities than default platform labels provide, enabling the future development of automated classifiers trained on these refined taxonomies.