dataset.yaml

---
languages:
- en
licenses:
- cc-by-4.0
multilinguality:
- multilingual
size_categories:
- 1K<n<10K
source_datasets:
- original
task_categories:
- automatic-speech-recognition
task_ids:
- speech-recognition
pretty_name: Whisper ASR Dataset
configs:
- name: whisper_v1
  description: Contains ASR from Whisper v1.
- name: whisper_v1_en
  description: English-translated datasets from Whisper v1.
- name: whisper_v2
  description: Contains ASR from Whisper v2.
- name: whisper_v2_en
  description: English-translated datasets from Whisper v2.
- name: whisper_v3
  description: Contains ASR from Whisper v3.
dataset_info:
  features:
  - name: segments
    feature: {
      "segment index (int)": [
        "start time in second (float)",
        "end time in second (float)",
        "transcribed text from ASR"
      ]
    }
  splits:
    whisper_v1:
      name: whisper_v1
      description: ASR data from Whisper v1
    whisper_v1_en:
      name: whisper_v1_en
      description: English-translated ASR data from Whisper v1
    whisper_v2:
      name: whisper_v2
      description: ASR data from Whisper v2
    whisper_v2_en:
      name: whisper_v2_en
      description: English-translated ASR data from Whisper v2
    whisper_v3:
      name: whisper_v3
      description: ASR data from Whisper v3
---

# Dataset Card for Whisper ASR Dataset

## Dataset Description

The Whisper ASR Dataset is a collection of automatic speech recognition (ASR) data generated using Whisper models. The dataset contains multiple splits, each corresponding to different versions of the Whisper model and language translations. The splits are as follows:

- **whisper_v1**: Contains ASR data from Whisper v1.
- **whisper_v1_en**: English-translated datasets from Whisper v1.
- **whisper_v2**: Contains ASR data from Whisper v2.
- **whisper_v2_en**: English-translated datasets from Whisper v2.
- **whisper_v3**: Contains ASR data from Whisper v3.

### Languages

The primary language of the dataset is English (`en`).

### Licenses

The dataset is licensed under the Creative Commons Attribution 4.0 International License ([CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)).

### Dataset Structure

Each JSON file in the dataset contains ASR segments with the following structure:

```json
{
  "segments": {
    "segment index (int)": [
      "start time in second (float)",
      "end time in second (float)",
      "transcribed text from ASR"
    ]
  }
}