ylacombe HF staff commited on
Commit
0c37939
1 Parent(s): b5f7630

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +215 -1
README.md CHANGED
@@ -231,4 +231,218 @@ pretty_name: Google English Dialects
231
  ---
232
  # Dataset Card for "english_dialects"
233
 
234
- [More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
231
  ---
232
  # Dataset Card for "english_dialects"
233
 
234
+
235
+ ## Table of Contents
236
+ - [Dataset Description](#dataset-description)
237
+ - [Dataset Summary](#dataset-summary)
238
+ - [Supported Tasks](#supported-tasks)
239
+ - [How to use](#how-to-use)
240
+ - [Dataset Structure](#dataset-structure)
241
+ - [Data Instances](#data-instances)
242
+ - [Data Fields](#data-fields)
243
+ - [Data Statistics](#data-statistics)
244
+ - [Dataset Creation](#dataset-creation)
245
+ - [Curation Rationale](#curation-rationale)
246
+ - [Source Data](#source-data)
247
+ - [Annotations](#annotations)
248
+ - [Personal and Sensitive Information](#personal-and-sensitive-information)
249
+ - [Considerations for Using the Data](#considerations-for-using-the-data)
250
+ - [Social Impact of Dataset](#social-impact-of-dataset)
251
+ - [Discussion of Biases](#discussion-of-biases)
252
+ - [Other Known Limitations](#other-known-limitations)
253
+ - [Additional Information](#additional-information)
254
+ - [Dataset Curators](#dataset-curators)
255
+ - [Licensing Information](#licensing-information)
256
+ - [Citation Information](#citation-information)
257
+ - [Contributions](#contributions)
258
+
259
+ ## Dataset Description
260
+
261
+ - **Homepage:** [Crowdsourced high-quality UK and Ireland English Dialect speech data set.](https://www.openslr.org/83/)
262
+ - **Repository:** [Google Language Resources and Tools](https://github.com/google/language-resources)
263
+ - **Paper:** [Open-source Multi-speaker Corpora of the English Accents in the British Isles](https://aclanthology.org/2020.lrec-1.804/)
264
+
265
+ ### Dataset Summary
266
+
267
+ This dataset consists of 31 hours of transcribed high-quality audio of English sentences recorded by 120 volunteers speaking with different accents of the British Isles. The dataset is intended for linguistic analysis as well as use for speech technologies. The speakers self-identified as native speakers of Southern England, Midlands, Northern England, Welsh, Scottish and Irish varieties of English.
268
+
269
+ The recording scripts were curated specifically for accent elicitation, covering a variety of phonological phenomena and providing a high phoneme coverage.
270
+ The scripts include pronunciations of global locations, major airlines and common personal names in different accents; and native speaker pronunciations of local words.
271
+ Overlapping lines for all speakers were included for idiolect elicitation, which include the same or similar lines with other existing resources such as the [CSTR VCTK corpus](https://huggingface.co/datasets/vctk) and the Speech Accent Archive to allow for easy comparison of personal and regional accents.
272
+
273
+ The data archives were restructured from the original ones from [OpenSLR](http://www.openslr.org/83) to make it easier to stream.
274
+
275
+
276
+ ### Supported Tasks
277
+
278
+ - `text-to-speech`, `text-to-audio`: The dataset can be used to train a model for Text-To-Speech (TTS).
279
+ - `automatic-speech-recognition`, `speaker-identification`: The dataset can also be used to train a model for Automatic Speech Recognition (ASR). The model is presented with an audio file and asked to transcribe the audio file to written text. The most common evaluation metric is the word error rate (WER).
280
+
281
+
282
+ ### How to use
283
+
284
+ The `datasets` library allows you to load and pre-process your dataset in pure Python, at scale. The dataset can be downloaded and prepared in one call to your local drive by using the `load_dataset` function.
285
+
286
+ For example, to download the Irish male config, simply specify the corresponding language config name (i.e., "irish_male" for Irish male speakers):
287
+ ```python
288
+ from datasets import load_dataset
289
+
290
+ dataset =load_dataset("ylacombe/english_dialects", "irish_male", split="train")
291
+ ```
292
+
293
+ Using the datasets library, you can also stream the dataset on-the-fly by adding a `streaming=True` argument to the `load_dataset` function call. Loading a dataset in streaming mode loads individual samples of the dataset at a time, rather than downloading the entire dataset to disk.
294
+ ```python
295
+ from datasets import load_dataset
296
+
297
+ dataset =load_dataset("ylacombe/english_dialects", "irish_male", split="train", streaming=True)
298
+
299
+ print(next(iter(dataset)))
300
+ ```
301
+
302
+ #### *Bonus*
303
+ You can create a [PyTorch dataloader](https://huggingface.co/docs/datasets/use_with_pytorch) directly with your own datasets (local/streamed).
304
+
305
+ **Local:**
306
+
307
+ ```python
308
+ from datasets import load_dataset
309
+ from torch.utils.data.sampler import BatchSampler, RandomSampler
310
+
311
+ dataset =load_dataset("ylacombe/english_dialects", "irish_male", split="train")
312
+ batch_sampler = BatchSampler(RandomSampler(dataset), batch_size=32, drop_last=False)
313
+ dataloader = DataLoader(dataset, batch_sampler=batch_sampler)
314
+ ```
315
+
316
+ **Streaming:**
317
+
318
+ ```python
319
+ from datasets import load_dataset
320
+ from torch.utils.data import DataLoader
321
+
322
+ dataset =load_dataset("ylacombe/english_dialects", "irish_male", split="train", streaming=True)
323
+ dataloader = DataLoader(dataset, batch_size=32)
324
+ ```
325
+
326
+ To find out more about loading and preparing audio datasets, head over to [hf.co/blog/audio-datasets](https://huggingface.co/blog/audio-datasets).
327
+
328
+ ## Dataset Structure
329
+
330
+ ### Data Instances
331
+
332
+ A typical data point comprises the path to the audio file called `audio` and its transcription, called `text`. Some additional information about the speaker and the passage which contains the transcription is provided.
333
+
334
+ ```
335
+ {'line_id': 'BI0057', 'audio': {'path': 'irm_02484_00388340153.wav', 'array': array([-1.22070312e-04, -1.52587891e-04, -1.22070312e-04, ...,
336
+ 1.52587891e-04, 9.15527344e-05, 1.83105469e-04]), 'sampling_rate': 48000}, 'text': 'It is thirteen degrees with drizzle in Exeter', 'speaker_id': 2484}
337
+ ```
338
+
339
+ ### Data Fields
340
+
341
+ - audio: A dictionary containing the audio filename, the decoded audio array, and the sampling rate. Note that when accessing the audio column: `dataset[0]["audio"]` the audio file is automatically decoded and resampled to `dataset.features["audio"].sampling_rate`. Decoding and resampling of a large number of audio files might take a significant amount of time. Thus it is important to first query the sample index before the `"audio"` column, *i.e.* `dataset[0]["audio"]` should **always** be preferred over `dataset["audio"][0]`.
342
+
343
+ - text: the transcription of the audio file.
344
+
345
+ - speaker_id: unique id of the speaker. The same speaker id can be found for multiple data samples.
346
+
347
+ - line_id: unique id of the transcription. The same line id can be found for multiple speakers.
348
+
349
+
350
+ ### Data Statistics
351
+
352
+
353
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/62611fcabbcbd1c34f1615f6/ony5ZDV7h1xP3tZCgh0Qj.png)
354
+
355
+
356
+
357
+ ## Dataset Creation
358
+
359
+ ### Curation Rationale
360
+
361
+ [Needs More Information]
362
+
363
+ ### Source Data
364
+
365
+ #### Initial Data Collection and Normalization
366
+
367
+ [Needs More Information]
368
+
369
+ #### Who are the source language producers?
370
+
371
+ [Needs More Information]
372
+
373
+ ### Annotations
374
+
375
+ #### Annotation process
376
+
377
+ [Needs More Information]
378
+
379
+ #### Who are the annotators?
380
+
381
+ [Needs More Information]
382
+
383
+ ### Personal and Sensitive Information
384
+
385
+ The dataset consists of people who have donated their voice online. You agree to not attempt to determine the identity of speakers in this dataset.
386
+
387
+ ## Considerations for Using the Data
388
+
389
+ ### Social Impact of Dataset
390
+
391
+ [More Information Needed]
392
+
393
+ ### Discussion of Biases
394
+
395
+ [More Information Needed]
396
+
397
+ ### Other Known Limitations
398
+
399
+ [Needs More Information]
400
+
401
+ ## Additional Information
402
+
403
+ ### Dataset Curators
404
+
405
+ [Needs More Information]
406
+
407
+ ### Licensing Information
408
+
409
+ License: ([CC BY-SA 4.0 DEED](https://creativecommons.org/licenses/by-sa/4.0/deed.en))
410
+
411
+ ### Citation Information
412
+
413
+ @inproceedings{demirsahin-etal-2020-open,
414
+ title = "Open-source Multi-speaker Corpora of the {E}nglish Accents in the {B}ritish Isles",
415
+ author = "Demirsahin, Isin and
416
+ Kjartansson, Oddur and
417
+ Gutkin, Alexander and
418
+ Rivera, Clara",
419
+ editor = "Calzolari, Nicoletta and
420
+ B{\'e}chet, Fr{\'e}d{\'e}ric and
421
+ Blache, Philippe and
422
+ Choukri, Khalid and
423
+ Cieri, Christopher and
424
+ Declerck, Thierry and
425
+ Goggi, Sara and
426
+ Isahara, Hitoshi and
427
+ Maegaard, Bente and
428
+ Mariani, Joseph and
429
+ Mazo, H{\'e}l{\`e}ne and
430
+ Moreno, Asuncion and
431
+ Odijk, Jan and
432
+ Piperidis, Stelios",
433
+ booktitle = "Proceedings of the Twelfth Language Resources and Evaluation Conference",
434
+ month = may,
435
+ year = "2020",
436
+ address = "Marseille, France",
437
+ publisher = "European Language Resources Association",
438
+ url = "https://aclanthology.org/2020.lrec-1.804",
439
+ pages = "6532--6541",
440
+ abstract = "This paper presents a dataset of transcribed high-quality audio of English sentences recorded by volunteers speaking with different accents of the British Isles. The dataset is intended for linguistic analysis as well as use for speech technologies. The recording scripts were curated specifically for accent elicitation, covering a variety of phonological phenomena and providing a high phoneme coverage. The scripts include pronunciations of global locations, major airlines and common personal names in different accents; and native speaker pronunciations of local words. Overlapping lines for all speakers were included for idiolect elicitation, which include the same or similar lines with other existing resources such as the CSTR VCTK corpus and the Speech Accent Archive to allow for easy comparison of personal and regional accents. The resulting corpora include over 31 hours of recordings from 120 volunteers who self-identify as native speakers of Southern England, Midlands, Northern England, Welsh, Scottish and Irish varieties of English.",
441
+ language = "English",
442
+ ISBN = "979-10-95546-34-4",
443
+ }
444
+ ```
445
+
446
+ ### Contributions
447
+
448
+ Thanks to [@ylacombe](https://github.com/ylacombe) for adding this dataset.