Add metadata manifest and source download links

README.md

@@ -10,11 +10,19 @@ tags:
 
 # Speech Quality Test Labels (metadata-only)
 
-This repository intentionally does **not** contain the `data/` directory or any audio files. The audio cannot be redistributed as a single package because the 14 source datasets have different access and copyright terms. This README records the subset counts, the label schema, and the official access paths for obtaining the original data independently.
+This Hugging Face repository intentionally contains no `data/` directory and no audio files. It publishes only this README and [`metadata.jsonl`](./metadata.jsonl), because the 14 source datasets have different copyright and access terms.
 
-The sample counts below refer to the subset used by PICSAFEv1, not the full size of each source dataset.
+The manifest contains the 10,728-sample PICSAFEv1 subset. It is not a license to redistribute the underlying recordings.
 
-## Source datasets and official access paths
+## Metadata fields
+
+- `file_name`: audio basename.
+- `id`: globally unique sample identifier with the `[source]` prefix.
+- `source`: source dataset name.
+- `source_path`: path relative to the local `data/audio/` directory. For example, resolve an entry as `data/audio/{source_path}` after obtaining the corresponding source audio locally.
+- `tags`: list of labels.
+
+## Source datasets and official download paths
 
 | Source dataset | Samples | License / access note | Official download or access path |
 | --- | ---: | --- | --- |
@@ -39,16 +47,6 @@ The 33 labels are:
 
 `Nspk`, `distortion`, `emotional`, `enhanced`, `excessive_sibilance`, `fast_speed`, `female`, `high_pitch`, `imperceptible_noise_level`, `instantaneous_noise`, `low_noise_level`, `low_pitch`, `male`, `microphone_popping`, `mid_high_noise_level`, `music_or_effect`, `nb_noise`, `non_binary`, `non_speech`, `read_speech`, `real_recording`, `regular_pitch`, `regular_speed`, `reverberation`, `singing_voice`, `slow_speed`, `speech_overlap`, `spontaneous_speech`, `synthetic`, `vocal_sound`, `wb_noise`, `whispered_speech`, and `wrong_transcript`.
 
-## Metadata schema
-
-The local preparation manifest uses these fields:
-
-- `file_name`: audio basename.
-- `id`: globally unique sample identifier with the `[source]` prefix.
-- `source`: source dataset name.
-- `source_path`: path relative to the local `data/` directory, when the source audio is available locally.
-- `tags`: list of labels.
-
 ## Licensing and limitations
 
-This repository does not grant, aggregate, or override any source-dataset license. Obtain the audio directly from the source providers and follow their current terms, attribution requirements, access restrictions, and applicable privacy/publicity rules. In particular, do not treat this metadata-only repository as permission to redistribute the underlying recordings.
+This repository does not grant, aggregate, or override any source-dataset license. Obtain the audio directly from the source providers and follow their current terms, attribution requirements, access restrictions, and applicable privacy/publicity rules. Some sources require non-commercial research use, registration, a license, or direct permission from the authors.