metadata

language:
  - eu
license: cc-by-4.0
size_categories:
  - 100K<n<1M
task_categories:
  - automatic-speech-recognition
pretty_name: Composite dataset for basque v2.1
tags:
  - asr
  - stt
  - dataset
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test_cv
        path: data/test_cv-*
      - split: test_parl
        path: data/test_parl-*
      - split: test_oslr
        path: data/test_oslr-*
      - split: dev
        path: data/dev-*
      - split: dev_cv
        path: data/dev_cv-*
      - split: dev_parl
        path: data/dev_parl-*
      - split: dev_oslr
        path: data/dev_oslr-*
dataset_info:
  features:
    - name: audio
      dtype: audio
    - name: sentence
      dtype: string
    - name: duration
      dtype: float64
  splits:
    - name: train
      num_bytes: 77939859505
      num_examples: 387426
    - name: dev
      num_bytes: 350063742
      num_examples: 1691
    - name: dev_cv
      num_bytes: 117244633
      num_examples: 620
    - name: dev_parl
      num_bytes: 116145832
      num_examples: 550
    - name: dev_oslr
      num_bytes: 116673276
      num_examples: 521
    - name: test_cv
      num_bytes: 2789895701
      num_examples: 14312
    - name: test_parl
      num_bytes: 328152751
      num_examples: 1521
    - name: test_oslr
      num_bytes: 118968255
      num_examples: 526

Composite dataset for Basque made from public available data

This dataset is composed of the following public available data:

Train split:

The train split is composed of the following datasets combined:

mozilla-foundation/common_voice_18_0/eu: "validated" split removing "test_cv" and "dev_cv" split's sentences. (validated split contains official train + dev + test splits and more unique data)
gttsehu/basque_parliament_1/eu: "train_clean" split removing some of the sentences that are repeated in "test_parl" and "dev_parl" splits (not the same recording but same text).
openslr: a train split made from the SLR76 (Basque recordings) subset, this split has been cleaned from acronyms, numbers and sentences that are repeated in the following "test_oslr" and "dev_oslr" splits.

Split tag Source Hours Sentences

- common_voice_18_0 300.05 h 198498

- basque_parliament_1 369.65 h 185699

- openslr 6.28 h 3229

train Total 675.98 h 387426

Split tag	Source	Hours	Sentences
-	common_voice_18_0	300.05 h	198498
-	basque_parliament_1	369.65 h	185699
-	openslr	6.28 h	3229
train	Total	675.98 h	387426

Test splits:

Those test splits are separated, and it is recommended to not evaluate them together in a single split:

mozilla-foundation/common_voice_18_0/eu: official "test" split.
gttsehu/basque_parliament_1/eu: official "test" split.
openslr: a test split made from the SLR76 (Basque recordings) subset, this split has been cleaned from acronyms, numbers and repeated sentences so all of them are only once in the split.

Split tag Source Hours Sentences

test_cv common_voice_18_0 24.2 h 14312

test_parl basque_parliament_1 2.85 h 1521

test_oslr openslr 1.03 h 526

Total 28.08 h 16359

Split tag	Source	Hours	Sentences
test_cv	common_voice_18_0	24.2 h	14312
test_parl	basque_parliament_1	2.85 h	1521
test_oslr	openslr	1.03 h	526
	Total	28.08 h	16359

Dev splits:

Theres a dev split composed by 3 dev subsplits that are also independently accesible. It is recommended to use the combined "dev" split for development tasks since it is accurately balanced in number of hours.

mozilla-foundation/common_voice_18_0/eu: a small dev split made from the official "dev" split.
gttsehu/basque_parliament_1/eu: a small dev split made from the official "dev" split.
openslr: a dev split made from the SLR76 (Basque recordings) subset, this split has been cleaned from acronyms, numbers and repeated sentences so all of them are only once in the split.

Split tag Source Hours Sentences

dev_cv common_voice_18_0 1.02 h 620

dev_parl basque_parliament_1 1.01 h 550

dev_oslr openslr 1.01 h 521

dev Total 3.04 h 1691

Split tag	Source	Hours	Sentences
dev_cv	common_voice_18_0	1.02 h	620
dev_parl	basque_parliament_1	1.01 h	550
dev_oslr	openslr	1.01 h	521
dev	Total	3.04 h	1691

Funding:

This project with reference 2022/TL22/00215335 has been parcially funded by the Ministerio de Transformación Digital and by the Plan de Recuperación, Transformación y Resiliencia – Funded by the European Union – NextGenerationEU ILENIA and by the project IkerGaitu funded by the Basque Government.