Datasets:

sartifyllc
/

Sukuma-Voices

text stringlengths 20 362	audio audioduration (s) 1.4 30.4
Ū ng'wikanza lya matwajo ūtūma nsese wakwe, gīkī aje ū kū balīmi benabo, bakang'winhe matwajo kūfuma ū mu ngūnda.
Alīyo a balīmi benabo būūkandima būntūlagūla milangha, būneka ūja makono dalali.
Ū ng'wenekīlī ngūnda ūtūma nsese ūngī, lelo ū ng'wene būmelagūla nguma, nīyo būng'wītīla ya minala kībi no!
Ūtūma ūngī hangī, ū ng'wene būshikīla kūmūlaga! Akatūma na bangī bingī dūhū; bamò babo būūkatūlagūlwa, bangī būbūlagwa!
Walī ataalī nanghwe ūmò; ng'wana wakwe ntogwa, nose ū kū nghalīlo yaho ūntūma ng'wenūyo, ū kūnū akwiwīlaga gīkī, ‘Bakūnkūja ū ng'wana wane!’
Alīyo a balīmi benabo biwīla gīkī, ‘Ūyū angū hū ng'wene ū ng'wingīji gete, tūmūlagagi būbize wise ū wingīji!’
Na nghana būndima, būmūlaga, būūkamponya hanze ya ngūnda!”
“Ī gīsī lūūlū, ū ng'wenekīlī ngūnda akwīta mhayo kī? Tūhaya akwiza abakenaagūle a balīmi benabo, nū ngūnda ūgūtūūla mu makono ga balīmi bangī!
A bing'we mutaalī ū kūshisoma ī shandīkwa īsho shihayile gīkī, ‘Ī liwe īlo bakalīlema bazengi h'īlo lyabizile lya kwīgoloolela ī numba
Ū mhayo gwenūyū gūfumile kūlī Seeba, nīyo gūlī gwawiza no ū mu miso gise?’ ”
Bakadeeba gīkī, ū lūsumo lwenūlo akalūhaya kūlwa nguno yabo, hūna lūūlū bandya kūkooba nzīla ya kūndima, kwike būbogoha a banhū bīnga ho, būneka.
Ha numa ya yenīyo, būntūmīla Bafalisayo balebe, hamò na ban'ikūlū ba ng'wa Helode, gīkī bakangwashe mu kalūngalūngīle.
Bahayūshika būng'wīla, “Nangi, tūmanile gīkī ū bebe ūkashimīzīlaga mu nghana, nguno ūtibeelejaga kūlī munhū ose-ose gīkī ūnyeje; na hangī ūtadīlīlaga banhū mu kūlola būshū wabo, alīyo ū mhayo go ng'wa Mulungu ūkagūlangaga mu nghana. Angū, īzunīlīgije ū kūng'winha goodi ū Kaisali nūūlū yaya? Twikale tūtoba ī goodi nūūlū tooye?”
Alīyo īkī walī abūmanile ū wibeeleja wabo, ūbawīla gīkī, “Nibūlī mulīnigema? Nenhelagi aha hela īmò nīlole.”
Būng'wenhela; hanuma ūbabūja, “Ī suso īyī nū wandīke wenūbū, sha ng'wa nani?” Būnshokeja, “Sha ng'wa Kaisali.”
Hūna lūūlū ū Yesu ūbawīla, “Ī sha ng'wa Kaisali ng'winhagi Kaisali, nī sha ng'wa Mulungu ng'winhagi Mulungu.” Bakankumya no!
Hūna Basadukayo, abo bahayile gīkī būtīho būjūūko, būnsanga kūmūja; būhaya,
“Nangi, ū Musa atwandīkīlile gīkī, ūlū ndūgū ūzumalīka ūneka nke alīyo atalekile ng'wana, ng'wanong'wawe akūng'wingīla ū nkīma ng'wenūyo, amyalīle baana ū ndūgūye.
Ali lūūlū, baalīho bagosha mpungatī ba shikūlū na shizuna. Ū tangì wabo akatoola, alīyo ūcha atalekile būbyīle.
Ū wa kabīlī ūng'wingīla ū nkīma ng'wenūyo, nanghwe ūcha atalekile būbyīle, nū wa kadatū gīko dūhū.
Kūshisha būshila pye a bose mpungatī; batalekile būbyīle. Aho bakalīka, nose nū nkīma ūcha!
Ī gīsī lūūlū, ū ng'wikanza lya būjūūko, ūlū būjūūka, akūbiza nke o ng'wa nani, īkī pye a bose mpungatī baalī bantoola?”
Ū Yesu ūbawīla, “A bing'we ahene mutūhūbīlaga kūlwa kūbiza mutashimanile ī shandīkwa nū būdūla bo ng'wa Mulungu?
Ū Yosefu aho wamisha kūfuma ū mu tūlo, wīta gīt'ūmo akawīlīlwa nū malaika wa ng'wa Seeba; ūnsola ū nke wakwe.
Kwike atalaalile nanghwe kūshisha ūmyala ū Ng'wana, hūna ūng'wilīka ī lina lyakwe ī lya Yesu.
Ū Yesu akabyalīlwa mu nzengo gwa Betelehemu ya Būyudaya, mu shikū ja ntemi Helode. Kūhayimanīla biza ū mu Yelusalemu bamani ba mihayo ya sonda, bafumile mhandī ja kīya.
Aho bashika būbūja gīkī, “Alī halī ū Ntemi wa Bayudaya ūyo wabyalilwe? Nguno a bise ī kīya twabonile sonda yakwe, na lūūlū twizaga kūnamya.
Ū ntemi Helode aho wigwa ī mihayo yenīyo akazwangana no, pye na a bangī abo baalī hamò nanghwe ī Yelusalemu.
Ūbakuminga pye a bagabīji bataale na a balangi ba banhū, wandya kūbabūgīlīja ī lipandī lya kūbyalīlwa ū Kilisto.
Nabo būng'wīla, “Betelehemu ya Būyudaya, nguno ū mulī nghangi handīkilwe gīkī,
‘Bebe Betelehemu mu sī ya Būyudaya, ūtī ndo na hado yaya ū mu basugi ba Būyudaya, nguno ū kūfuma mulī bebe akwilonga Nsugi, ūyo akūbadīīma banhū bane, a Baisilaeli.’
Hūna ū Helode ūkalalwa kūmana kūfuma ū kū bamani ba mihayo ya sonda benabo. Ūbapūnja ha mhūjo kūbagīsīlīja chiza ī likanza lya kilongele ka sonda yenīyo.
Ha numa ya henaho ūbatūma baje ī Betelehemu; ūbawīla, “Jagi mukīgīsīlīje mu welwa-ngholó ī mihayo ya ng'wana. Ūlū mulamone ng'wize muniwīle, nguno nane natogilwe kūjūnamya!”
Aho bamala kūndegeleka ū ntemi, būbūūka bajile. Bībona hangī ī sonda ībatongeelile, gīt'ūmo yalī yabatongeelela ū kūfuma ī kīya. Yahayūshika a h'ipandī īlo walī ū Ng'wana, yoya ū kūselema.
Aho bībona ī sonda bakayega na kayegele kataale no!
Bahayingīla ū mu numba, būmona ū Ng'wana nū nina ū Malia, būtuja mazwi būnamya. Baalī na manongho gabo, būgataligūla būng'winha shakwinha: zahabu, būbbani, na ngazu.
Aho bahaya kūja kaya, ū Mulungu ūbahūgūla mu shilooti gīkī batizūbīta ū kūlī Helode. Na nghana būbītīla nzīla yīngī būja kū sī yabo
Aho bīnga, malaika wa ng'wa Seeba ūng'wilongela ū Yosefu mu shilooti; ūng'wīla gīkī, “Būūkaga ūnsole ū Ng'wana nū nina, ūpeelele Misili, ūkikale kwenūko kūshisha nakūwīle. Nguno ū Helode alī hihī kwandya kūnkooba ū Ng'wana amūlage.”
Na nghana ūmisha, ūnsola ū Ng'wana hamò nū nina, būbūūka kūja Misili ī būjikū yenīyo. 15Akikala kwenūko kūshisha ūcha ū ntemi Helode. Ī mihayo īkaja gīko kūlwa kūshikīlījiwa kwa būhangi bo ng'wa Seeba; ūmo nghangi wakwe ahayīlile gīkī, “Nakang'witana ū Ng'wana wane kūfuma Misili.”
Ū Helode akabona gīkī wagīmva nyachilū na a bamani ba mihayo ya sonda, na lūūlū ūpelana no. Ūlagīla babūlagwe a banīgīnī a bagosha, a ba mu Betelehemu, nī chalo īsho shibīmbīkanilwe nayo; pye a b'ilika lya kwandīja myaka ībīlī na kwika ha sīlīlī ya henaho, kūlenganīla nī likanza īlo bakang'wīla a bamani ba mihayo ya sonda, aho wabagīsīlīja.
Ū mu kwīta gīko gūbiza gwashikīīla ū mhayo ūyo gūhayiwe na nghangi Yelemia; gīkī,
“Ilaka līkigwiwa mu Lama, nghūngū na kūganghīla kūtaale;Laheli alīlīlīīla baana bakwe,nū kūhuumūjiwa alemile nguno batīho!”
Aho waacha ū Helode, malaika wa ng'wa Seeba ūng'wilongela hangī ū Yosefu mu shilooti aho alī Misili. Ūng'wīla gīkī,
“Būūkaga ūnsole ū Ng'wana nū nina ūshoke kū sī ya Isilaeli; nguno baachile abo balūbūkoobaga ū būpanga bo Ng'wana.”
Nghana aho wamisha ūnsola ū Ng'wana nū nina, būbūūka bajile; būshika ū mu sī ya Isilaeli.
Alīyo aho wigwa gīkī Akelao watemile ū mu sī ya Būyudaya a ha ng'wanya go ng'wa ise ū Helode, akogoha ū kūja ko ū kwenūka. Ūhūgūlwa mu shilooti, ūja mhandī ja Galilaya,
ūūkazenga Nazaleti. Gūbiza gwashikīīla ūyo gūhayiwe na bahangi gīkī, “Alitanwa muna Nazaleti.”
Ū mu shikū jenījo Yohana Matīīja akiza kū matogolo ga Būyudaya; ūyūlomeela
alīhaya, “Galūkagi, nguno ū būtemi bo ng'wigūlū wegeelaga.”
Ū ng'wenūyo hū ng'wene ūyo ahayiwe na nghangi Isaya, gīkī, “Ilaka lya munhū akūhamukaga mu matogolo, ‘Beeja-beejagi ikūūwa lya ng'wa Seeba, jigoloolagi ī nzīla jakwe.’ ”
Ī shizwalo sha ng'wa Yohana shalī sha booya bo ngamīla, ū mu nkīmbīlī witungaga ngwasho gwa nkoba, ī shilīwa shakwe shalī njige na būūkī bo mu matogolo.
Banhū bingī bakayiza ūko walī; bafumaga Yelusalemu na hose-hose ū mu sī ya Būyudaya, nū mu sī pyī ī ja kū mbalama ya mongo gwa Yolodani.
Būlī bene bashikaga batambūūla ī shibi shabo wababatīīja ū mu mongo gwenūyo.
Kwike aho wabona Bafalisayo bingī na Basadukayo baliza ū kū būbatīīja bokwe; ūbawīla, “Bing'we būbyīle wa shipīlī, nani wamuhūgūlile gīkī mugūpeele ū nsango ūyo gūliza?
Ali lūūlū, twajagi matwajo kūlwa koolecha būkaanīji bo kūgalūka kwing'we.
Mutiganikage mukwiwīlaga gīkī, ‘Tūlī na nkūlūgenji wise Abulahamu.’ Nguno nalīmuwīla nūūlū kūfuma mu mawe gīt'aya, ū Mulungu adūgije kūmūūkīja baana ū Abulahamu.
Nīyo ī haha īlīho mbasa ītūūlilwe nghana-nghana a ha matina ga mitī. Hū kūhaya būlī ntī ūyo gūtūtwaja matwajo gawiza, gūkūbbutwa gūponyiwe mu moto.
“Ū nene nghana nalīmubatīīja mu minzī kūlwa kūgalūka; alīyo a ha numa yane aliza ūyo anikīlile ū būdūla. Ū nene natigeleelilwe nūūlū ni kūmūūkīja shilatū shakwe, ū ng'wenūyo h'ūyo akūmubatīīja na Moyo Ng'wela, na moto.
Ū lūhūngo lokwe alī nalo mu nkono gokwe wei ng'wenekīlī, hama ū lūbūūga lokwe akūlūpyagūla alweje lwele pe. Ū būsīga abūkuminge chiza abūtūūle ng'wibīkīlo lyakwe; alīyo a meela a gene ūgapemba mu moto gūtī gwa kūjima!”
Hūna ū Yesu wīnga ī Galilaya, ūja ūko walī ū Yohana, ū kū mongo gwa Yolodani, nanghwe akabatīījiwe.
Alīyo ū Yohana ūnemeja alīhaya, “Nibūlī ūliza kūlī nene, alīyo ū nene hū īnigeleelile ūnibatīīje ū bebe?”
Nanghwe ū Yesu ūnshokeja ūng'wīla, “Nizunīlījage ī haha, nguno īlī yawiza tūbūshikīlīje ū būtūngīlīja ū bose.” Hūna ūnzunīlīja.
Ū Yesu aho wamala ū kūbatīījiwa, haho na haho ūfuma ū mu minzī. Ī likanza lyenīlo lyūkundūka ī ligūlū, ūmona ū Moyo wa ng'wa Mulungu alika gītī nghūūlū, izile a halī we.
N'ilaka lyūfuma ū ng'wigūlū līlīhaya gīkī, “Ūyū ng'wene Ng'wana wane Ntogwa, ūyo aniyegije hataale no!”
Ha numa ya yenīyo, ū Yesu ūtongeelwa na Moyo kūja kū matogolo, kūjūgeng'wa na Shetani.
Ūūkikala kwenūko shikū makūmi anè, akūdīlaga nzala līīmi na būjikū; nose ūtuuba!
Ū ngemi ūng'wizīīla ūng'wīla gīkī, “Ūlū ūlī Ng'wana wa ng'wa Mulungu, gawīlage a mawe genaya gabize shilīwa.”
Nanghwe ūshosha ūhaya, “Yandīkilwe gīkī, ‘Ū munhū atūdūla kūbiza mpanga kūlwa shilīwa kwike, alīyo kūlwa būlī mhayo ūyo gūfumile mu nomo go ng'wa Mulungu.’
Hūna ū Shetani ūntwala kū nzengo ng'wela, aho wanshisha koi ūntūūla higūlya ya hekalu.
Ūng'wīla gīkī, “Ūlū ūlī Ng'wana wa ng'wa Mulungu al'iponejage ha sī shi; nguno gwandīkilwe gīkī, ‘Akūlagīla bamalaika bakwe kūlwako, nabo bakūkūbūūcha mu makono gabo, ūtizipama lūpambala lwako h'iwe!’ ”
Nanghwe ū Yesu ūnshokeja ūng'wīla, “Yandīkilwe hangī gīkī, ‘Yaya ū kūngema ū Seeba Mulungu wako.’
Ha numa hangī ū Shetani ūntwala kū lūgūlū lūlīhu; ūūkandya kūng'oolekeja pye a mabūsugi ga mu sī, hamò nī likūjo lyago.
Hūna ūng'wīla gīkī, “Nakūkwinha bebe pye a genaya, ūlū ūgwa ha sī kūnilamya.”
A henaho ū Yesu ūng'wīla, “Īngaga a henaha Shetani, nguno gwandīkilwe gīkī, ‘Namyage Seeba Mulungu wako, na ūntūmamīle wei dūhū!’ ”
Hūna ū Shetani ūng'wīngīla, bamalaika biza kūntūmamīla.
Ū Yesu aho wigwa gīkī ū Yohana waponyiwe mu jeela, akaja Galilaya.
Kwike atalendile ī Nazaleti, wīnga koi ūja ūūkikala Kafalanaūmu, mhandī ja Zabuloni na Nafutali; hihī na nyanza.
Ū mhayo gwenūyū gūbiza gwashikīlīja īyo īhayiwe na nghangi Isaya; gīkī,
“Sī ya Zabuloni na sī ya Nafutali, nzīla ya kū nyanza, kū nkīlo gwa Yolodani; Galilaya ya banhū ba mahanga.
A banhū ba mu giiti baabonile isana itaale, alichene, isana lyabaakīlile abo wikalowabo būlī mu nengeeji gwa lūfu!”
Kūfuma ikanza lyenīlo, ū Yesu wandya kūlomeela alīhaya gīkī, “Galūkagi, nguno ū būtemi bo ng'wigūlū wegeelaga!”Kwitanwa kwa bahemba banè
Lūshikū lūmò aho alīshimīza kū mbalama ya nyanza ya Galilaya, akababona banhū babīlī ba shikūlū na shizuna, Simoni ūyo witanagwa Petelo, na Andelea ng'wanong'wawe; baategaga ntego mu nyanza, nguno baalī bategi ba ndīlo.
Ūbawīla gīkī, “Nilondeelagi, nakūng'wīta mubize bazubi ba banhū.”
Nghana haho na haho bīleka ī mitego yabo būnondeela.
Aho waja ha būtongi hado, ūbabona bangī babīlī ba shikūlū na shizuna, Yakobo ng'wana Zebedayo, na Yohana ng'wanong'wawe. Nabo baalī mu lyato hamò na saabo Zebedayo, bakūbeeja-beejaga mitego yabo: na a bene ūbitana.
Nabo haho na haho būnondeela, būneka ū saabo nī lyato lyabo.
Hūna wandya kūja hose-hose ū mu Galilaya akūlangaga mu masinagogi gabo, na kūlomeela Nghūlū Jawiza ja būtemi, hamò na kūpīja būsaatu bo banhū, na būlemehazu bo būlī mbika.
Ū lūkumo lokwe lūsambaala hose pye, kūshisha lūshika mu sī ya Sulia. Būyūng'wenhela pye abo baalī basaatu, na a babo baalī na būlemehazu būngī na būngī bo mīlī. Pye wabapījaga, na a babo baluhiagwa masamva, na a ba lūsalo lwa kūgwa, na a ba būsaatu bo kūpola mīlī.
Mabità mataale kūfuma ū mu Galilaya ng'wenūmo gūyūnondeela; na bangī bafumaga Dekapoli, bangī Yelusalemu, nī Būyudaya ī yose. Bamò bafumaga kū nkīlo gwa mongo gwa Yolodani.
Aho wagabona a mabità ga banhū, ūlinha mu lūgūlū ūūkigasha. Bahayūng'wegeela a bahemba bakwe,
wandya kūbalanga alīhaya gīkī,
“Balī na mbango a bahabī mu moyo;nguno ū būtemi bo ng'wigūlū būlī wabo.
Balī na mbango abo balī na būpīna; nguno bakūlūngūjiwa
Balī na mbango a bidohya; nguno bakūbiza bingīji ba sī.
Balī na mbango abo lūtuubo na notà yabo shilī ha kūkooba būtūngīlīja; nguno bakūshikīlīgījiwa ī nghūmbū jabo.
Balī na mbango a bachīji ba shigongo;nguno bakūchīlwa shigongo.
Balī na mbango a beela mu moyo;nguno bakūmona ū Mulungu.
Balī na mbango a batūūji ba mhola;nguno bakwitanwa balī baana ba ng'wa Mulungu.
Balī na mbango abo bakūluhiagwa kūlwa kwīta yawiza;nguno ū būtemi bo ng'wigūlū būlī wabo.
“A bing'we mulī na mbango ūlū bakūmusondagūlaga na kūmuluhya, na kūmulembekeja ya būbi bo būlī mbika kūlwa nguno yane.

End of preview. Expand in Data Studio

Sukuma Voices Dataset

Dataset Description

Sukuma Voices is the first publicly available speech corpus for Sukuma (Kisukuma), a Bantu language spoken by approximately 10 million people in northern Tanzania. This dataset addresses the critical gap in speech technology resources for one of Africa's most severely under-resourced languages.

Dataset Summary

Metric	Value
Total Samples	7,989
Total Duration	23.44 hours
Average Duration	10.56 ± 4.32 seconds
Duration Range	1.40 - 33.10 seconds
Total Words	168,022
Unique Vocabulary	24,398
Average Words/Sample	21.0
Speaking Rate	121.6 WPM

Supported Tasks

Automatic Speech Recognition (ASR): Converting Sukuma speech to text
Text-to-Speech (TTS): Synthesizing natural-sounding Sukuma speech from text
Cross-lingual speech processing: Research between Swahili and Sukuma

Languages

Sukuma (ISO 639-3: suk) - A Bantu language of the Niger-Congo family

Dataset Structure

Data Instances

Each instance contains:

audio: Audio recording in standard format (resampled to 16kHz for ASR, 24kHz for TTS)
transcription: Corresponding text transcription in Sukuma

Data Splits

Split	Samples	Percentage
Train	~6,391	80%
Test	~1,598	20%

Example

Language	Text
Sukuma	Umunhu ngwunuyo agabhalelaga chiza abhanhu bhakwe, kunguyo ya kikalile kakwe akagubhatogwa na gubhambilija abho bhali mumakoye.
English	This person raises his people well, because of his good behavior, of loving people and helping his colleagues who are in trouble, in their lives.

Dataset Creation

Source Data

The dataset was curated from audio recordings and textual transcriptions of the Sukuma New Testament 2000 translation, sourced from the Bible.com platform.

Why Biblical Text?

Standardized orthographic conventions ensuring transcription consistency
Diverse linguistic structures encompassing narrative, dialogue, and theological discourse
Cultural relevance to Sukuma-speaking communities
Availability of both audio recordings and verified textual transcriptions

Annotations

The data was rigorously annotated to ensure phonetic and orthographic consistency, with validation by native Sukuma speakers.

Personal and Sensitive Information

The dataset contains religious text (Bible readings) and does not include personal or sensitive information about individuals.

Considerations for Using the Data

Known Limitations

Domain Specificity: Data is primarily from biblical texts, which may not fully represent everyday conversational Sukuma
Diacritic Variations: Sukuma has two written forms (with and without diacritics); this dataset focuses on the non-diacritic version
Single Source: Limited speaker diversity from a single recording source

Linguistic Challenges

Sukuma is a tonal language with complex phonological features
The language lacks standardized orthographic conventions across written materials
Diacritic and non-diacritic text representations can affect vocabulary size and evaluation metrics

Baseline Results

ASR Performance

We evaluated both Whisper Large V3 and Wav2Vec2-large-XLSR-53 architectures. Whisper's sequence-to-sequence framework with pretrained multilingual representations proved significantly more effective for Sukuma's tonal and diacritic-rich phonology.

Whisper Large V3 (Best Model)

Metric	Original Speech	Synthetic Speech
Final WER	25.19%	32.60%
Min WER	22.01%	29.97%
WER Reduction	82.94%	78.93%

Key Findings:

Strong correlation between original and synthetic learning curves (Pearson's r = 0.997)
Performance gap narrows as training progresses (from 9.97 to 8.11 WER points)
Synthetic speech captures essential acoustic-phonetic characteristics despite ~28% relative performance gap

TTS Performance (Orpheus 3B v0.1)

Mean Opinion Score (MOS): 3.9 ± 0.15 (out of 5)
Human Recording MOS: 4.6 ± 0.1
Evaluation: Subjective quality study with native Sukuma speakers using 5-point Likert Scale

Usage

Loading the Dataset

from datasets import load_dataset

dataset = load_dataset("sartifyllc/Sukuma-Voices")

ASR Example

from transformers import WhisperProcessor, WhisperForConditionalGeneration

processor = WhisperProcessor.from_pretrained("openai/whisper-large-v3")
model = WhisperForConditionalGeneration.from_pretrained("your-finetuned-model")

# Process audio
audio = dataset["test"][0]["audio"]
input_features = processor(audio["array"], sampling_rate=16000, return_tensors="pt").input_features

# Generate transcription
predicted_ids = model.generate(input_features)
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)

Citation

If you use this dataset, please cite:

@inproceedings{mgonzo2025sukuma,
  title={Learning from Scarcity: Building and Benchmarking Speech Technology for Sukuma},
  author={Mgonzo, Macton and Oketch, Kezia and Etori, Naome and Mang'eni, Winnie and Nyaki, Elizabeth and Mollel, Michael S.},
  booktitle={Proceedings of the Association for Computational Linguistics},
  year={2026}
}

Additional Information

Authors

Macton Mgonzo - Brown University (macton_mgonzo@brown.edu)
Kezia Oketch - University of Notre Dame
Naome Etori - University of Minnesota - Twin Cities
Winnie Mang'eni - Pawa AI
Elizabeth Nyaki - Pawa AI, Sartify Company Limited
Michael Mollel - Sartify Company Limited (michael.mollel@sartify.com, michael.mollel@pawa-ai.com) — Correspondence

Dataset Curators

Acknowledgments

We would like to express our gratitude to Sartify Company Limited and Pawa AI for their instrumental role in initiating this project and for providing the data access necessary to develop and evaluate our models. We also extend our sincere thanks to all the volunteers who generously dedicated their time to the evaluation process, as their contributions were vital to the completion of this work.

Licensing Information

This dataset is released under CC-BY-4.0.

Contributions

We welcome contributions to expand and improve this dataset! Areas of interest include:

Additional Sukuma speech data beyond religious content
Conversational and everyday language recordings
Multi-speaker recordings
Diacritic-annotated transcriptions

Ethical Considerations

Consent was obtained from all human participants involved in data annotation
Participants were informed about the technology's limitations and potential impacts
The authors acknowledge that models trained on this data may inherit biases present in the source material

Contact

For questions or contributions, please open an issue on this repository.

This dataset represents an important step toward inclusive speech technology for African languages. We hope it will catalyze continued research on low-resource speech technologies.

Downloads last month: 83

Size of downloaded dataset files:

8.06 GB

Size of the auto-converted Parquet files:

8.06 GB

Number of rows:

7,989

Models trained or fine-tuned on sartifyllc/Sukuma-Voices

sartifyllc/sukuma-voices-asr

2B • Updated 2 days ago • 27