Datasets:
annotations_creators:
- expert-generated
language_creators:
- expert-generated
language:
- ajz
- bam
- bi
- bis
- bjn
- bm
- boz
- bze
- bzi
- cak
- ceb
- chd
- chp
- clo
- csw
- en
- eng
- es
- fli
- fr
- fra
- gu
- guj
- hbb
- hi
- hin
- id
- ind
- jmx
- jra
- kan
- kbq
- kek
- kjb
- kmu
- kn
- kqr
- kwu
- loh
- mai
- mal
- mam
- mar
- ml
- mle
- mr
- my
- mya
- myk
- nas
- nsk
- nsn
- oj
- oji
- omw
- por
- pt
- quc
- sdk
- snk
- spa
- stk
- ta
- taj
- tam
- tbj
- tdc
- tgl
- tl
- tpi
- tuz
- tzj
license:
- cc-by-nc-4.0
- cc-by-sa-4.0
- cc-by-nc-nd-4.0
- cc-by-nc-sa-4.0
multilinguality:
- multilingual
size_categories:
- 10K<n<100K
source_datasets:
- original
task_categories:
- automatic-speech-recognition
- text-to-speech
paperswithcode_id: null
pretty_name: BloomSpeech
extra_gated_prompt: >-
One more step before getting this dataset. This dataset is open access and
available only for non-commercial use (except for portions of the dataset
labeled with a `cc-by-sa` license). A "license" field paired with each of the
dataset entries/samples specifies the Creative Commons license for that
entry/sample.
These [Creative Commons
licenses](https://creativecommons.org/about/cclicenses/) specify that:
1. You cannot use the dataset for or directed toward commercial advantage or
monetary compensation (except for those portions of the dataset labeled
specifically with a `cc-by-sa` license. If you would like to ask about
commercial uses of this dataset, please [email us](mailto:sj@derivation.co).
2. Any public, non-commercial use of the data must give appropriate credit,
provide a link to the license, and indicate if changes were made. You may do
so in any reasonable manner, but not in any way that suggests the licensor
endorses you or your use.
3. For those portions of the dataset marked with an ND license, you cannot
remix, transform, or build upon the material, and you may not distribute
modified material.
In addition to the above implied by Creative Commons and when clicking "Access
Repository" below, you agree:
1. Not to use the dataset for any use intended to or which has the effect of
harming or enabling discrimination against individuals or groups based on
legally protected characteristics or categories, including but not limited to
discrimination against Indigenous People as outlined in Articles 2; 13-16; and
31 of the United Nations Declaration on the Rights of Indigenous People, 13
September 2007 and as subsequently amended and revised.
2. That your *contact information* (email address and username) can be shared
with the model authors as well.
extra_gated_fields:
I have read the License and agree with its terms: checkbox
Table of Contents
-
- Data Instances
- Data Fields
- Data Splits
Dataset Description
- Homepage: SIL AI
- Point of Contact: SIL AI email
- Source Data: Bloom Library
Dataset Summary
Bloom is free, open-source software and an associated website Bloom Library, app, and services developed by SIL International. Bloom’s primary goal is to equip non-dominant language communities and their members to create the literature they want for their community and children. Bloom also serves organizations that help such communities develop literature and education or other aspects of community development.
This version of the Bloom Library data is developed specifically for the automatic speech recognition and speech-to-text tasks. It includes data from 56 languages across 18 language families. There is a mean of 458 and median of 138 audio records per language.
Note: If you speak one of these languages and can help provide feedback or corrections, please let us know!
Note: Although data from bloom-lm was used in the training of the BLOOM model, the dataset only represents a small portion of the data used to train that model. Data from "Bloom Library" was combined with a large number of other datasets to train that model. "Bloom Library" is a project that existed prior to the BLOOM model, and is something separate. All that to say... We were using the "Bloom" name before it was cool. 😉
Languages
Of the 500+ languages listed at BloomLibrary.org, there are 56 languages available in this dataset. Here are the corresponding ISO 639-3 codes:
ajz, bam, bis, bjn, boz, bze, bzi, cak, ceb, chd, chp, clo, csw, eng, fli, fra, guj, hbb, hin, ind, jmx, jra, kan, kbq, kek, kjb, kmu, kqr, kwu, loh, mai, mal, mam, mar, mle, mya, myk, nas, nsk, nsn, oji, omw, por, quc, sdk, snk, spa, stk, taj, tam, tbj, tdc, tgl, tpi, tuz, tzj
Dataset Statistics
Some of the languages included in the dataset include few audio cuts. These are not split between training, validation, and test. For those with higher numbers of available stories we include the following numbers of stories in each split:
ISO 639-3 Name Train Cuts Validation Cuts Test Cuts ajz Amri Karbi 135 34 50 bam Bamanankan 203 50 50 bis Bislama 0 0 46 bjn Banjar 80 20 50 boz Bozo, Tieyaxo 427 50 52 bze Bozo, Jenaama 101 26 50 bzi Bisu 1363 50 157 cak Kaqchikel 989 50 115 ceb Cebuano 553 50 67 chd Chontal, Highland Oaxaca 205 50 50 chp Dene 0 0 14 clo Chontal, Lowland Oaxaca 120 30 50 csw Cree, Swampy 0 0 45 eng English 4143 48 455 fli Fali Muchella 59 15 50 fra French 261 49 50 guj Gujarati 27 0 48 hbb Nya Huba 558 50 67 hin Hindi 62 15 49 ind Indonesian 0 0 14 jmx Mixtec, Western Juxtlahuaca 39 0 50 jra Jarai 203 50 50 kan Kannada 281 43 50 kbq Kamano 0 0 27 kek Q’eqchi’ 1676 49 190 kjb Q’anjob’al 770 50 91 kmu Kanite 0 0 28 kqr Kimaragang 0 0 18 kwu Kwakum 58 15 50 loh Narim 0 0 15 mai Maithili 0 0 11 mal Malayalam 125 31 44 mam Mam 1313 50 151 mar Marathi 25 0 49 mle Manambu 0 0 8 mya Burmese 321 50 50 myk Sénoufo, Mamara 669 50 80 nas Naasioi 13 0 50 nsk Naskapi 0 0 15 nsn Nehan 0 0 31 oji Ojibwa 0 0 25 omw Tairora, South 0 0 34 por Portuguese 0 0 34 quc K’iche’ 1460 50 167 sdk Sos Kundi 312 50 50 snk Soninke 546 50 66 spa Spanish 1816 50 207 stk Aramba 180 45 50 taj Tamang, Eastern 0 0 24 tam Tamil 159 39 46 tbj Tiang 0 0 24 tdc Ẽpẽra Pedea 0 0 19 tgl Tagalog 352 48 50 tpi Tok Pisin 1061 50 123 tuz Turka 48 13 50 tzj Tz’utujil 0 0 41 Dataset Structure
Data Instances
The examples look like this for Hindi:
from datasets import load_dataset # Specify the language code. dataset = load_dataset('sil-ai/bloom-speech', 'hin', use_auth_token=True) #note you must login to HuggingFace via the huggingface hub or huggingface cli # A data point consists of transcribed audio in the specified language code. # To see a transcription: print(dataset['train']['text'][0])This would produce an output:
चित्र: बो और शैम्पू की बोतलWhereas if you wish to gather all the text for a language you may use this:
dataset['train']['text']Data Fields
The metadata fields are below. In terms of licenses, all stories included in the current release are released under a Creative Commons license (even if the individual story metadata fields are missing).
- file: the local path to the audio file
- audio: a dictionary with a path, array, and sampling_rate as is standard for Hugging Face audio
- text: the transcribed text
- book: title of the book, e.g. "बो मेस्सी और शैम्पू".
- instance: unique ID for each book/translation assigned by Bloom Library. For example the Hindi version of 'बो मेस्सी और शैम्पू' is 'eba60f56-eade-4d78-a66f-f52870f6bfdd'
- license: specific license used, e.g. "cc-by-sa" for "Creative Commons, by attribution, share-alike".
- credits: attribution of contributors as described in the book metadata, including authors, editors, etc. if available
- original_lang_tag: the language tag originally assigned in Bloom Library. This may include information on script type, etc.
Data Splits
All languages include a train, validation, and test split. However, for language having a small number of stories, certain of these splits maybe empty. In such cases, we recommend using any data for testing only or for zero-shot experiments.
Changelog
- 26 September 2022 Page initiated