Upload README.md with huggingface_hub

README.md

@@ -53,7 +53,7 @@ Total: **60.06 minutes**.
 ## Evaluation (evidence, not just claims)
 
 - **Single-speaker check** (ECAPA-TDNN embeddings): same-speaker cosine 0.74 vs different-speaker 0.21 (separation 0.52, verification AUC 0.96 / EER 9.1%; 0/11 speakers flagged).
-- **Transcript reliability**: English cross-ASR agreement with Whisper = 6.8% WER / 4.5% CER (n=40) — strong. Realtime ASR language-ID matched the target language on 100% of EN and 100% of TE clips. Telugu cross-ASR is not a valid proxy (Whisper is weak in Telugu); Telugu transcripts are best audited by human review.
+- **Transcript reliability**: English cross-ASR agreement with Whisper = 6.8% WER / 4.5% CER (n=40), strong. Realtime ASR language-ID matched the target language on 100% of EN and 100% of TE clips. Telugu cross-ASR is not a valid proxy (Whisper is weak in Telugu); Telugu transcripts are best audited by human review.
 - **Emotion-tag reliability** (sarvam-30b vs sarvam-105b on 120 clips): 65% agreement, Cohen's κ 0.55.
 - **Phoneme coverage**: English 39 (100%), Telugu 44 (88%).
 - **Perceptual quality** (DNSMOS OVRL, published set): EN 3.21 (86% pass>3.0), TE 3.19 (86% pass>3.0). Filter `dnsmos_pass=True` for a stricter subset.
@@ -76,7 +76,7 @@ See the project report (GitHub repo) for full methodology and figures.
 5. Automated quality gates (clipping, SNR, silence, music/noise bed, ASR confidence, dedup).
 6. Hybrid emotion tagging: per-speaker-normalized acoustic features + Sarvam LLM,
    with an acoustic whisper override.
-7. Human review (listen, fix transcripts, relabel) — **human labels override automated ones**.
+7. Human review (listen, fix transcripts, relabel); **human labels override automated ones**.
 8. Light loudness normalization (dynamics preserved), balanced emotion selection.
 
 ## Audio