README.md

---
license: apache-2.0
task_categories:
- automatic-speech-recognition
language:
- zh
configs:
- config_name: SPEECHIO_ASR_ZH00000
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00000/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00001
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00001/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00002
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00002/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00003
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00003/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00004
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00004/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00005
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00005/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00006
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00006/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00007
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00007/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00008
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00008/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00009
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00009/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00010
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00010/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00011
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00011/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00012
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00012/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00013
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00013/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00014
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00014/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00015
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00015/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00016
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00016/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00017
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00017/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00018
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00018/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00019
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00019/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00020
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00020/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00021
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00021/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00022
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00022/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00023
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00023/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00024
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00024/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00025
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00025/test-*.parquet
- config_name: SPEECHIO_ASR_ZH00026
  data_files:
  - split: test
    path: SPEECHIO_ASR_ZH00026/test-*.parquet
---

# SpeechIO ASR Test Sets (parquet)

Parquet repackaging of the [SpeechColab SpeechIO](https://github.com/SpeechColab/Leaderboard) Mandarin ASR benchmark,
re-exported from [`yuekai/speechio`](https://huggingface.co/datasets/yuekai/speechio) (Lhotse cuts) into standard
HuggingFace parquet with embedded 16 kHz audio.

- 27 test sets: `SPEECHIO_ASR_ZH00000` ... `SPEECHIO_ASR_ZH00026`, each a config with a single `test` split.
- ~43k utterances, ~66 hours total, evaluation only.

## Columns
| column | type | note |
|---|---|---|
| `segment_id` | string | utterance id |
| `speaker` | string | speaker id |
| `text` | string | reference transcript (Chinese) |
| `duration` | float32 | seconds |
| `audio` | Audio(16kHz) | embedded wav bytes |
| `original_full_path` | string | source path in original corpus |

## Usage
```python
from datasets import load_dataset
ds = load_dataset("yuekai/speechio_test", "SPEECHIO_ASR_ZH00001", split="test")
print(ds[0]["text"], ds[0]["audio"]["sampling_rate"])
```