README.md

---
configs:
- config_name: ko
  data_files:
  - split: train
    path: corpus/ko.train.clean.jsonl
  - split: eval
    path: corpus/ko.eval.clean.jsonl
- config_name: ko-5000h
  data_files:
  - split: train
    path: corpus/ko.train.coverage-5000h.jsonl
  - split: eval
    path: corpus/ko.eval.clean.jsonl
- config_name: audio-jp1924__broadcastspeech
  data_files:
  - split: train
    path: audio/jp1924__broadcastspeech/*/part-*.parquet
- config_name: audio-jp1924__chungcheongspeech
  data_files:
  - split: train
    path: audio/jp1924__chungcheongspeech/*/part-*.parquet
- config_name: audio-jp1924__ganwonspeech
  data_files:
  - split: train
    path: audio/jp1924__ganwonspeech/*/part-*.parquet
- config_name: audio-jp1924__gyeongsangspeech
  data_files:
  - split: train
    path: audio/jp1924__gyeongsangspeech/*/part-*.parquet
- config_name: audio-jp1924__jejuspeech
  data_files:
  - split: train
    path: audio/jp1924__jejuspeech/*/part-*.parquet
- config_name: audio-jp1924__jeollaspeech
  data_files:
  - split: train
    path: audio/jp1924__jeollaspeech/*/part-*.parquet
- config_name: audio-jp1924__kconfspeech
  data_files:
  - split: train
    path: audio/jp1924__kconfspeech/*/part-*.parquet
- config_name: audio-jp1924__klecspeech
  data_files:
  - split: train
    path: audio/jp1924__klecspeech/*/part-*.parquet
- config_name: audio-jp1924__koreanuniversitylecturedata
  data_files:
  - split: train
    path: audio/jp1924__koreanuniversitylecturedata/*/part-*.parquet
- config_name: audio-jp1924__koreaspeech
  data_files:
  - split: train
    path: audio/jp1924__koreaspeech/*/part-*.parquet
- config_name: audio-jp1924__krespspeech
  data_files:
  - split: train
    path: audio/jp1924__krespspeech/*/part-*.parquet
- config_name: audio-jp1924__ksponspeech
  data_files:
  - split: train
    path: audio/jp1924__ksponspeech/*/part-*.parquet
- config_name: audio-jp1924__patternedutterancewithnumber
  data_files:
  - split: train
    path: audio/jp1924__patternedutterancewithnumber/*/part-*.parquet
- config_name: audio-junhoee__stt_korean_dataset_80000
  data_files:
  - split: train
    path: audio/junhoee__stt_korean_dataset_80000/*/part-*.parquet
- config_name: audio-korean_speech_metadata__junhoee_stt_korean_dataset_80000
  data_files:
  - split: train
    path: audio/korean_speech_metadata__junhoee_stt_korean_dataset_80000/*/part-*.parquet
---

# ko cleaned speech corpus

This refresh provides a converged full train manifest, a frozen eval manifest,
and a token-coverage subset of approximately 5000 hours.
The validated FLAC payloads are reused without decoding or transcoding. Completely referenced Parquet shards are hard-linked locally; partial shards are filtered into exact row-only Parquet files.

## Corpus overview

| Variant | Rows | Hours |
|---|---:|---:|
| train | 20,137,680 | 35976.329 |
| coverage_5000h | 2,919,596 | 4992.635 |
| eval | 1,198 | 1.997 |

All audio locators are portable and have the form
`audio/<subset>/<group>/part-N.parquet#row=N`.

The release audit records exact input/output hashes, rejected audio mappings,
language-tag validation, and zero train/eval audio-identity overlap.

```python
from datasets import load_dataset
ds = load_dataset("YuKuanFu/ko", "ko-5000h", split="train")
```

## Packaging exclusions

Rows that could not be represented by the validated FLAC pool are omitted
and recorded here rather than silently replaced:

| Variant | Rows | Hours | Reasons |
|---|---:|---:|---|
| train | 40,733 | 41.434033 | all_zero_audio: 35, duration_mismatch: 40,698 |
| coverage_5000h | 7,338 | 7.364521 | all_zero_audio: 5, duration_mismatch: 7,333 |
| eval | 1 | 0.001478 | duration_mismatch: 1 |