--- pretty_name: Webinars — Russian speech and ASR transcripts language: - ru license: mit task_categories: - automatic-speech-recognition - text-to-speech size_categories: - 100K`. ## Preparation 1. Original audio was downloaded from `ZeroAgency/shkolkovo-bobr.video-webinars-audio` at revision `c35640ea983658baa75047ad0d058c7912ff08ce`. This release uses the recordings, with newly generated local transcripts; it does not reuse the source repository's TXT transcripts. 2. FireRed VAD located speech and grouped it into segments targeting 20–30 seconds. Short tails remain; some boundaries were forced to satisfy the maximum duration. 3. ASR used `Systran/faster-whisper-large-v3`, revision `edaa852ec7e145841d8ffdb056a99866b5f0a478`, with FP16 inference, beam size 5, automatic language detection and temperature fallback. 4. Long internal quiet pauses were shortened using the saved, audited edit maps. Eligible quiet runs of at least 0.8 s retained 0.3 s, protecting VAD speech with 60 ms margins and trusted Unity2 word alignments with 80 ms margins. Untrusted alignments used a more conservative energy threshold. Edge silence was retained. No new alignment or transcription was run for this publication. 5. The ASR text was preserved verbatim. The 130 segments with empty or whitespace-only transcripts were omitted. The language/music filters below were then applied. No stress marks or prosody tags were added to the transcripts. 6. Audio was reconstructed at its native sample rate, converted to mono PCM16 and encoded as FLAC. Every original and pause-edited float32 chunk was checked against the waveform hashes saved during data preparation. FLAC headers, lengths and representative decode round trips were checked before upload. FLAC stores the exported PCM16 signal losslessly; the source recordings were already MP3. This export does not restore frequencies or detail lost by MP3. Full shard checksums and measured counts are in [manifest.json](manifest.json). ## Language and music filtering This filtered release retains only segments whose existing ASR language tag is `ru`. It removes 3,526 other-tagged segments and 8,643 additional Russian-tagged segments flagged for music, background music, singing, songs or musical instruments. [EfficientAT mn10_as](https://github.com/fschmid56/EfficientAT) scanned every Russian-tagged segment using 10-second windows with a 5-second hop, including the final window. Short clips were zero-padded. Analysis used temporary 32 kHz mono waveforms reconstructed from the same saved pause-edit spans as the release. [AST](https://huggingface.co/MIT/ast-finetuned-audioset-10-10-0.4593) independently checked all primary candidates (score >= 0.05 or background-music score >= 0.005) and 1,000 deterministic low-score audit samples at 16 kHz. Where these two detectors disagreed at a maximum target-class/window score of 0.15, EfficientAT mn40_as resolved the decision. A segment was removed when at least two detectors exceeded that threshold. The low-score audit found no additional music decisions under this rule. All shards were rebuilt from local source recordings and audited ASR/edit records. Retained audio signals, transcripts, identifiers and source spans are unchanged. Detector revisions, thresholds and counts are in [filtering.json](filtering.json); per-sample decisions and scores are in [filter_scores.parquet](filter_scores.parquet). The unfiltered version remains reproducible at revision `83c2459778926a7db6144abd552b084de1ac95e9`. Filtering is automatic; quiet music can remain and false exclusions are possible. Detector scores are not calibrated probabilities of a clean recording. ## Transcript-quality filtering (2026-11-14) This revision additionally excludes **1,430 audio/text pairs** (9.53 audio hours) from the preceding Russian/no-music release. Reference transcripts are kept exactly as they were; independent ASR text is used only as filtering evidence and does not replace our Whisper output. The screen combines source-specific extreme letter rates, severe anomalies in the existing Qwen word alignments, and large unaligned leading/tail spans. Unicode letter counts exclude whitespace, punctuation and stress markup. Audio duration is the exact released duration, including existing webinar pause edits. The duration-weighted mean is 10.39 letters/sec. Rates alone never establish that a transcript is wrong. 2,813 selected pairs, including fixed duration-stratified controls, were independently checked with Qwen3-ASR-1.7B and Silero speech detection. ASR received only audio, with automatic language detection and no reference text or hotwords. Comparison normalizes punctuation, case, ё and integer spellings. Qwen forced alignment places supplied words; timestamp margins and isolated zero-duration words are not evidence that the words were spoken or missing. Conservative exclusions require independent mismatch evidence. **714 uncertain cases are retained and marked for review**, including numeric ambiguity. The remainder of the corpus has not undergone exhaustive independent ASR checks; automatic filtering can miss errors and can exclude usable pairs. See [quality/report.json](quality/report.json), [quality/policy.py](quality/policy.py), [quality/decisions.parquet](quality/decisions.parquet), and [quality/exclusions.jsonl](quality/exclusions.jsonl). Integrated language, music and final transcript decisions are in [filter_scores.parquet](filter_scores.parquet). The preceding release remains available at revision `9c0dc32fc92ef6aa336783e2c838b1c0e3051ae9`. ## Transcript quality These are automatic transcripts, without manual correction or a measured corpus WER. Retained recordings can contain multiple speakers, background sounds, filler words, course-specific names or formulas, and music missed by the automatic filter. ASR may hallucinate, omit words or misidentify a language. Speech-aware segmenting and pause checks do not establish transcript correctness. This release preserves those limitations so users can apply task-specific filtering. ## Source and license The [upstream dataset card](https://huggingface.co/datasets/ZeroAgency/shkolkovo-bobr.video-webinars-audio) declares the MIT license and states that copyright for the course materials belongs to Shkolkovo.online. It describes the intended use as LLM and STT/TTS training. That attribution is retained here. See [UPSTREAM_README.md](UPSTREAM_README.md) for the original card and [NOTICE](NOTICE) for provenance. Please cite both the original dataset and this prepared release: ```bibtex @misc{zeroagency_webinars, author = {ZeroAgency}, title = {shkolkovo-bobr.video-webinars-audio}, howpublished = {Hugging Face dataset}, url = {https://huggingface.co/datasets/ZeroAgency/shkolkovo-bobr.video-webinars-audio} } @misc{teratts_webinars_2026, author = {TeraTTS}, title = {Webinars: segmented audio with ASR transcripts}, year = {2026}, howpublished = {Hugging Face dataset}, url = {https://huggingface.co/datasets/TeraTTS/webinars} } ```