Create dataset.yaml

dataset.yaml

@@ -0,0 +1,89 @@
+---
+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"
+    ]
+  }
+}