diarray commited on
Commit
5ab6cc6
·
1 Parent(s): 1e63c90

Update README and dataset viewer

Browse files
Files changed (1) hide show
  1. README.md +16 -78
README.md CHANGED
@@ -49,32 +49,32 @@ configs:
49
  data_files:
50
  - split: train
51
  path:
52
- - "jeli-asr-rmai/train/*"
53
- - "bam-asr-oza/train/*"
54
- - "oza-mali-pense/train/*"
55
- - "rt-data-collection/train/*"
56
  - split: test
57
  path:
58
- - "jeli-asr-rmai/test/*"
59
- - "bam-asr-oza/test/*"
60
  - config_name: jeli-asr
61
  data_files:
62
  - split: train
63
  path:
64
- - "jeli-asr-rmai/train/*"
65
- - "bam-asr-oza/train/*"
66
  - split: test
67
  path:
68
- - "jeli-asr-rmai/test/*"
69
- - "bam-asr-oza/test/*"
70
  - config_name: oza-mali-pense
71
  data_files:
72
  - split: train
73
- path: "oza-mali-pense/train/*"
74
  - config_name: rt-data-collection
75
  data_files:
76
  - split: train
77
- path: "rt-data-collection/train/*"
78
 
79
  description: |
80
  The **Bambara-ASR-All Audio Dataset** is a multilingual dataset containing audio samples in Bambara, accompanied by semi-expert transcriptions and French translations.
@@ -83,8 +83,6 @@ description: |
83
 
84
  ---
85
 
86
- # Documentation WILL be updated soon
87
-
88
  # All Bambara ASR Dataset
89
 
90
  This dataset aims to gather all publicly available Bambara ASR datasets. It is primarily composed of the **Jeli-ASR** dataset (available at [RobotsMali/jeli-asr](https://huggingface.co/datasets/RobotsMali/jeli-asr)), along with the **Mali-Pense** data curated and published by Aboubacar Ouattara (available at [oza75/bambara-tts](https://huggingface.co/datasets/oza75/bambara-tts)). Additionally, it includes 1 hour of audio recently collected by the RobotsMali AI4D Lab, featuring children's voices reading some of RobotsMali GAIFE books. This dataset is desihgned for automatic speech recognition (ASR) task primarily.
@@ -92,65 +90,6 @@ This dataset aims to gather all publicly available Bambara ASR datasets. It is p
92
  ## Important Notes
93
 
94
  1. Please note that this dataset is currently in development and is therefore not fixed. The structure, content, and availability of the dataset may change as improvements and updates are made.
95
- 2. The Dataset viewer has been disabled for this dataset since it uses a [custom loading script](bam-asr-all.py). You can safely load this dataset and all its features as a HF dataset object though (see usage section)
96
-
97
- ---
98
-
99
- ## **Directory Structure**
100
-
101
- ```
102
- bam-asr-all/
103
- |
104
- ├── README.md
105
- ├── metadata.csv
106
- ├── manifests/
107
- │ ├── jeli-asr-rmai-test-manifest.json
108
- │ ├── jeli-asr-rmai-train-manifest.json
109
- │ ├── oza-bam-asr-test-manifest.json
110
- │ └── oza-bam-asr-train-manifest.json
111
- │ └── oza-mali-pense-train-manifest.json
112
- │ └── reading-tutor-train-manifest.json
113
- │ └── train-manifest.json # jeli-asr-rmai-train-manifest.json + oza-bam-asr-train-manifest.json
114
- │ └── test-manifest.json # jeli-asr-rmai-test-manifest.json + oza-bam-asr-test-manifest.json
115
-
116
- ├── french-manifests/
117
- │ ├── jeli-asr-rmai-test-french-manifest.json
118
- │ ├── jeli-asr-rmai-train-french-manifest.json
119
- │ ├── oza-bam-asr-test-french-manifest.json
120
- │ └── oza-bam-asr-train-french-manifest.json
121
- │ └── oza-mali-pense-train-french-manifest.json
122
-
123
- ├── jeli-asr-rmai/ |
124
- │ ├── train/ |
125
- │ └── test/ |
126
- │ | These two subset are combined as jeli-asr
127
- ├── bam-asr-oza/ |
128
- │ ├── train/ |
129
- │ └── test/ |
130
- |
131
- ├── oza-mali-pense/
132
- │ ├── train/
133
- |
134
- ├── rt-data-collection/
135
- ```
136
-
137
- ### **manifests Directory**
138
- This directory contains the manifest files used for training speech recognition (ASR) and text-to-speech (TTS) models. Those are JSON files:
139
-
140
- Each line in the manifest files is a JSON object with the following structure:
141
- ```json
142
- {
143
- "audio_filepath": "bam-asr-all/rt-data-collection/zctn7pFmtmR45FKym7d5.wav",
144
- "duration": 10.24,
145
- "text": "ni birituban dɔ tun bɛ se ka piyano fɔ, a tun bɛ fara sɛ ka dɔnkilida la."
146
- }
147
- ```
148
- - **audio_filepath**: The relative path to the corresponding audio file.
149
- - **duration**: The duration of the audio file in seconds.
150
- - **text**: The transcription of the audio in Bambara.
151
-
152
- ### 3. **french-manifests/**
153
- This directory contains French equivalent manifest files for the dataset. The structure is similar to the `manifests/` directory but with French transcriptions
154
 
155
  ---
156
 
@@ -172,15 +111,14 @@ This directory contains French equivalent manifest files for the dataset. The st
172
 
173
  ## **Usage**
174
 
175
- The manifest files are specifically created for training Automatic Speech Recognition (ASR) models in NVIDIA NeMo framework, but they can be used with any other framework that supports manifest-based input formats or reformatted for other use cases.
176
-
177
- To use the dataset, simply load the manifest files (`train-manifest.json` and `test-manifest.json`) in your training script. The file paths for the audio files and the corresponding transcriptions are already provided in these manifest files You can also load it directly in a HuggingFace dataset object.
178
 
179
  ### Downloading the Dataset:
180
 
181
  ```bash
182
 
183
  # Clone dataset repository maintaining directory structure for quick setup with Nemo
 
184
  git clone --depth 1 https://huggingface.co/datasets/RobotsMali/bam-asr-all
185
  ```
186
 
@@ -191,13 +129,13 @@ git clone --depth 1 https://huggingface.co/datasets/RobotsMali/bam-asr-all
191
  from datasets import load_dataset
192
 
193
  # Load the dataset into Hugging Face Dataset object
194
- dataset = load_dataset("RobotsMali/bam-asr-all", trust_remote_code=True)
195
  # Note: You can also download only a specific subset if you with
196
  ```
197
 
198
  ## **Known Issues**
199
 
200
- This dataset also has most of the issues of Jeli-ASR, plus a few samples with missing french translations
201
 
202
  ---
203
 
 
49
  data_files:
50
  - split: train
51
  path:
52
+ - "rt-data-collection/train/*.arrow"
53
+ - "oza-mali-pense/train/*.arrow"
54
+ - "bam-asr-oza/train/*.arrow"
55
+ - "jeli-asr-rmai/train/*.arrow"
56
  - split: test
57
  path:
58
+ - "bam-asr-oza/test/*.arrow"
59
+ - "jeli-asr-rmai/test/*.arrow"
60
  - config_name: jeli-asr
61
  data_files:
62
  - split: train
63
  path:
64
+ - "bam-asr-oza/train/*.arrow"
65
+ - "jeli-asr-rmai/train/*.arrow"
66
  - split: test
67
  path:
68
+ - "bam-asr-oza/test/*.arrow"
69
+ - "jeli-asr-rmai/test/*.arrow"
70
  - config_name: oza-mali-pense
71
  data_files:
72
  - split: train
73
+ path: "oza-mali-pense/train/*.arrow"
74
  - config_name: rt-data-collection
75
  data_files:
76
  - split: train
77
+ path: "rt-data-collection/train/*.arrow"
78
 
79
  description: |
80
  The **Bambara-ASR-All Audio Dataset** is a multilingual dataset containing audio samples in Bambara, accompanied by semi-expert transcriptions and French translations.
 
83
 
84
  ---
85
 
 
 
86
  # All Bambara ASR Dataset
87
 
88
  This dataset aims to gather all publicly available Bambara ASR datasets. It is primarily composed of the **Jeli-ASR** dataset (available at [RobotsMali/jeli-asr](https://huggingface.co/datasets/RobotsMali/jeli-asr)), along with the **Mali-Pense** data curated and published by Aboubacar Ouattara (available at [oza75/bambara-tts](https://huggingface.co/datasets/oza75/bambara-tts)). Additionally, it includes 1 hour of audio recently collected by the RobotsMali AI4D Lab, featuring children's voices reading some of RobotsMali GAIFE books. This dataset is desihgned for automatic speech recognition (ASR) task primarily.
 
90
  ## Important Notes
91
 
92
  1. Please note that this dataset is currently in development and is therefore not fixed. The structure, content, and availability of the dataset may change as improvements and updates are made.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
93
 
94
  ---
95
 
 
111
 
112
  ## **Usage**
113
 
114
+ You can download and use this dataset with the Datasets library for download the archives to reconstruct the dataset with a specific structure for non HuggingFace workflows.
 
 
115
 
116
  ### Downloading the Dataset:
117
 
118
  ```bash
119
 
120
  # Clone dataset repository maintaining directory structure for quick setup with Nemo
121
+ # To be updated
122
  git clone --depth 1 https://huggingface.co/datasets/RobotsMali/bam-asr-all
123
  ```
124
 
 
129
  from datasets import load_dataset
130
 
131
  # Load the dataset into Hugging Face Dataset object
132
+ dataset = load_dataset("RobotsMali/bam-asr-all")
133
  # Note: You can also download only a specific subset if you with
134
  ```
135
 
136
  ## **Known Issues**
137
 
138
+ This dataset also has most of the issues of Jeli-ASR, including a few misaligned samples. Additionally a few samples from the mali pense subset and all the data from the rt-data-collection subset don't currently have french translations
139
 
140
  ---
141