nguyenvulebinh commited on
Commit
d4dfbb6
1 Parent(s): f110a12

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +65 -0
README.md CHANGED
@@ -224,4 +224,69 @@ configs:
224
  path: voxpopuli/train-*
225
  - split: valid
226
  path: voxpopuli/valid-*
 
 
 
 
 
227
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
224
  path: voxpopuli/train-*
225
  - split: valid
226
  path: voxpopuli/valid-*
227
+ language:
228
+ - en
229
+ pretty_name: Speech Recognition Alignment Dataset
230
+ size_categories:
231
+ - 10M<n<100M
232
  ---
233
+
234
+ # Speech Recognition Alignment Dataset
235
+
236
+ This dataset is a variation of several widely-used ASR datasets, encompassing Librispeech, MuST-C, TED-LIUM, VoxPopuli, Common Voice, and GigaSpeech. The difference is this dataset includes:
237
+ - Precise alignment between audio and text.
238
+ - Text that has been punctuated and made case-sensitive.
239
+ - Identification of named entities in the text.
240
+
241
+ # Usage
242
+
243
+ First, install the latest version of the 🤗 Datasets package:
244
+
245
+ ```bash
246
+ pip install --upgrade pip
247
+ pip install --upgrade datasets[audio]
248
+ ```
249
+
250
+ The dataset can be downloaded and pre-processed on disk using the [`load_dataset`](https://huggingface.co/docs/datasets/v2.14.5/en/package_reference/loading_methods#datasets.load_dataset)
251
+ function:
252
+
253
+ ```python
254
+ from datasets import load_dataset
255
+
256
+ # Available dataset: 'libris','mustc','tedlium','voxpopuli','commonvoice','gigaspeech'
257
+ dataset = load_dataset("nguyenvulebinh/asr-alignment", "libris")
258
+
259
+ # take the first sample of the validation set
260
+ sample = dataset["train"][0]
261
+ ```
262
+
263
+ It can also be streamed directly from the Hub using Datasets' [streaming mode](https://huggingface.co/blog/audio-datasets#streaming-mode-the-silver-bullet).
264
+ Loading a dataset in streaming mode loads individual samples of the dataset at a time, rather than downloading the entire
265
+ dataset to disk:
266
+
267
+ ```python
268
+ from datasets import load_dataset
269
+
270
+ dataset = load_dataset("nguyenvulebinh/asr-alignment", "libris", streaming=True)
271
+ # take the first sample of the validation set
272
+ sample = next(iter(dataset["train"]))
273
+ ```
274
+
275
+ ## Citation
276
+
277
+ If you use this data, please consider citing the [ICASSP 2024 Paper: SYNTHETIC CONVERSATIONS IMPROVE MULTI-TALKER ASR]():
278
+ ```
279
+ @INPROCEEDINGS{synthetic-multi-asr-nguyen,
280
+ author={Nguyen, Thai-Binh and Waibel, Alexander},
281
+ booktitle={ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
282
+ title={SYNTHETIC CONVERSATIONS IMPROVE MULTI-TALKER ASR},
283
+ year={2024},
284
+ volume={},
285
+ number={},
286
+ }
287
+
288
+ ```
289
+
290
+ ## License
291
+
292
+ This dataset is licensed in accordance with the terms of the original dataset.