Update training data description to 'publicly available recordings with aligned transcripts'

README.md

@@ -118,9 +118,9 @@ model.generation_config.forced_decoder_ids = None
 
 | Source | Repo | Hours | Role |
 |---|---|---|---|
-| New sehaj (YouTube captions) | [`gurbani-sehajpath-yt-captions-canonical`](https://huggingface.co/datasets/surindersinghssj/gurbani-sehajpath-yt-captions-canonical) | ~160h | Primary sehaj stream |
+| New sehaj (publicly available recordings with aligned transcripts) | [`gurbani-sehajpath-yt-captions-canonical`](https://huggingface.co/datasets/surindersinghssj/gurbani-sehajpath-yt-captions-canonical) | ~160h | Primary sehaj stream |
 | Old sehaj (studio) | [`gurbani-sehajpath`](https://huggingface.co/datasets/surindersinghssj/gurbani-sehajpath) | ~66h | Extra sehaj stream |
-| Kirtan (YouTube captions) | [`gurbani-kirtan-yt-captions-300h-canonical`](https://huggingface.co/datasets/surindersinghssj/gurbani-kirtan-yt-captions-300h-canonical) | ~420h | Aux kirtan stream |
+| Kirtan (publicly available recordings with aligned transcripts) | [`gurbani-kirtan-yt-captions-300h-canonical`](https://huggingface.co/datasets/surindersinghssj/gurbani-kirtan-yt-captions-300h-canonical) | ~420h | Aux kirtan stream |
 
 Training mix: **~220h sehaj** (36%) **+ ~420h kirtan** (64%) via oversampling aux at `AUX_TRAIN_PROBABILITY=0.64`. Text column normalized to `final_text` on all canonical sources. Old sehaj's `gurmukhi_text` column renamed at load.
 
@@ -189,7 +189,7 @@ Training mix: **~220h sehaj** (36%) **+ ~420h kirtan** (64%) via oversampling au
 
 - Not trained on **non-Gurbani Punjabi speech** — expect degradation on everyday Punjabi conversation
 - Not trained on **English / Hindi / other languages** — base Whisper multilingual ability is mostly lost after full fine-tune
-- **Speaker / mic / recording bias** — training data skews toward YouTube-sourced recordings; live Darbar Sahib PA reverb or noisy field recordings may WER worse
+- **Speaker / mic / recording bias** — training data skews toward publicly available web-sourced recordings with aligned transcripts; live Darbar Sahib PA reverb or noisy field recordings may WER worse
 - **Katha (spoken commentary)** is not in the training distribution — quality will vary
 
 ## How this compares to prior Surt versions