README.md

metadata

title: Chunk Classifier
emoji: 🎵
colorFrom: purple
colorTo: pink
sdk: gradio
app_file: app.py
pinned: false

Classify audio chunks for music contamination and technical defects using the AST AudioSet model (527 classes).

Input: one or more HF chunk dataset repo IDs (e.g. fosters/my-book-chunks).

Output: <input>_classified dataset — all rows kept, new columns added: music_score, has_music, contaminated, flags, top_labels, rms_db, clipping_ratio, max_silence_sec. The audio column is preserved as-is so you can listen in the HF viewer and sort by music_score.

Set HF_TOKEN as a Space secret to read private datasets and write to fosters/.