Datasets:

RobotsMali
/

an-be-kalan-bench

BookTitle stringclasses 22 values	sentenceID int64 1 147	text stringlengths 4 124	speakerAge int64 9 18	speakerGender stringclasses 2 values	speakerID stringclasses 8 values	duration float64 0.72 25.6
Aminata Fari Fisayara	1	Mɔgɔ caman tun bɛ yen.	9	female	spk_1778169797963_d3d546ba58	2.56
Aminata Fari Fisayara	2	An ye makɔnɔli kɛ yɔrɔ dɔ la sigilanw tun bɛ yɔrɔ min.	9	female	spk_1778169797963_d3d546ba58	6.4
Aminata Fari Fisayara	3	O kɔ, dɔgɔtɔrɔmuso dɔ ye an tɔgɔ wele.	9	female	spk_1778169797963_d3d546ba58	8.4
Aminata Fari Fisayara	4	Ne ni Mama taamana ka taa dɔgɔtɔrɔ ka biro kɔnɔ.	9	female	spk_1778169797963_d3d546ba58	6.96
Aminata Fari Fisayara	5	An taara dɔgɔtɔrɔso la.	9	female	spk_1778169797963_d3d546ba58	3.68
Aminata Fari Fisayara	6	A ye ne ɲininka ko:	9	female	spk_1778169797963_d3d546ba58	2.56
Aminata Fari Fisayara	7	“I tɔgɔ ye di?	9	female	spk_1778169797963_d3d546ba58	2.48
Aminata Fari Fisayara	8	Ne ye a fɔ a ye ko:	9	female	spk_1778169797963_d3d546ba58	3.84
Aminata Fari Fisayara	9	Ne tɔgɔ ye Aminata.	9	female	spk_1778169797963_d3d546ba58	3.12
Aminata Fari Fisayara	10	Tiyo jan dɔ tun bɛ a bolo.	9	female	spk_1778169797963_d3d546ba58	3.2
Aminata Fari Fisayara	11	A bɛ min wele ko “sitetosikɔpu”	9	female	spk_1778169797963_d3d546ba58	6.56
Aminata Fari Fisayara	12	A ye o da ne disi la.	9	female	spk_1778169797963_d3d546ba58	2.32
Aminata Fari Fisayara	13	A ye a fɔ ne ye ko:	9	female	spk_1778169797963_d3d546ba58	2.88
Aminata Fari Fisayara	14	Aminata, ninakili kosɛbɛ.	9	female	spk_1778169797963_d3d546ba58	3.6
Aminata Fari Fisayara	15	A ye a maga ne kɔnɔbara fana la.	9	female	spk_1778169797963_d3d546ba58	3.04
Aminata Fari Fisayara	16	Dɔgɔtɔrɔ in tun ye mɔgɔ ɲuman ye.	9	female	spk_1778169797963_d3d546ba58	4
Aminata Fari Fisayara	17	“Sumayabana de bɛ Aminata la wa?	9	female	spk_1778169797963_d3d546ba58	4.96
Aminata Fari Fisayara	18	Nka dɔgɔtɔrɔ ko :	9	female	spk_1778169797963_d3d546ba58	2.16
Aminata Fari Fisayara	19	Ayi, kɔnɔna bana de bɛ Aminata la.	9	female	spk_1778169797963_d3d546ba58	4.4
Aminata Fari Fisayara	20	A ye sɛbɛnni kɛ sɛbɛn dɔ kan.	9	female	spk_1778169797963_d3d546ba58	3.2
Aminata Fari Fisayara	21	A ye o di Mama ma.	9	female	spk_1778169797963_d3d546ba58	2.32
Aminata Fari Fisayara	22	Mama ko: Nin ye ɔrɔdonansi de ye.	9	female	spk_1778169797963_d3d546ba58	4.32
Aminata Fari Fisayara	23	Fura ninnu bɛna i kɛnɛya,	9	female	spk_1778169797963_d3d546ba58	3.12
Aminata Fari Fisayara	24	Aminata.	9	female	spk_1778169797963_d3d546ba58	1.6
Aminata Fari Fisayara	25	Mama ye dɔgɔtɔrɔ ɲininka ko:	9	female	spk_1778169797963_d3d546ba58	2.72
Aminata Fari Fisayara	26	A tun falen bɛ bɔnbɔnw la.	9	female	spk_1778169797963_d3d546ba58	2.88
Aminata Fari Fisayara	27	A ko ne ka kelen ta.	9	female	spk_1778169797963_d3d546ba58	2.08
Aminata Fari Fisayara	28	Ne ye bɔnbɔn bilenman dɔ ta.	9	female	spk_1778169797963_d3d546ba58	3.04
Aminata Fari Fisayara	29	Bilenman ka di ne ye.	9	female	spk_1778169797963_d3d546ba58	1.68
Aminata Fari Fisayara	30	Bɔnbɔn in diyara ne ye kosɛbɛ.	9	female	spk_1778169797963_d3d546ba58	2.56
Aminata Fari Fisayara	31	Dɔgɔtɔrɔ in tun kaɲi.	9	female	spk_1778169797963_d3d546ba58	2.4
Aminata Fari Fisayara	32	A ma ne tɔɔrɔ.	9	female	spk_1778169797963_d3d546ba58	2.4
Aminata Fari Fisayara	33	Saki dɔ tun bɛ dɔgɔtɔrɔ bolo.	9	female	spk_1778169797963_d3d546ba58	3.6
Aminata Fari Fisayara	34	ne bɛ fɛ ka taa dɔgɔtɔrɔ in fɛ walasa ka bɔnbɔn dɔwɛrɛ sɔrɔ.	9	female	spk_1778169797963_d3d546ba58	8.4
Aminata Fari Fisayara	35	Siɲɛ wɛrɛ ni ne banana,	9	female	spk_1778169797963_d3d546ba58	2.72
Aminata Fari Fisayara	36	Ɲininkaliw	9	female	spk_1778169797963_d3d546ba58	1.52
Aminata Fari Fisayara	38	o tununna. Sabula Aminata kunkolo ni a kɔnɔ tun bɛ a dimi.	9	female	spk_1778169797963_d3d546ba58	10.16
Aminata Fari Fisayara	39	Sabula Aminata tun bɛ a fɛ ka tulonkɛ dɔgɔtɔrɔ fɛ.	9	female	spk_1778169797963_d3d546ba58	8.64
Aminata Fari Fisayara	40	Ɲininkaliw Munna Aminata ni a ba taara dɔgɔtɔrɔso la?	9	female	spk_1778169797963_d3d546ba58	8.24
Aminata Fari Fisayara	42	Ɲininkaliw Dɔgɔtɔrɔ ye mun kɛ ka Aminata dusukun mankan lamɛn?	9	female	spk_1778169797963_d3d546ba58	7.76
Aminata Fari Fisayara	44	Ɲininkaliw Dɔgɔtɔrɔ ye mun bɔ a ka saki kɔnɔ ka a di Aminata ma?	9	female	spk_1778169797963_d3d546ba58	7.76
Aminata Fari Fisayara	45	o tɔgɔ ye di?	9	female	spk_1778169797963_d3d546ba58	1.44
Aminata Fari Fisayara	47	Ɲininkaliw Dɔgɔtɔrɔ ye fɛn min da Aminata disi kan walasa ka a dusukun mankan lamɛn,	9	female	spk_1778169797963_d3d546ba58	11.2
Aminata Fari Fisayara	48	o tɔgo ye di?	9	female	spk_1778169797963_d3d546ba58	1.92
Aminata Fari Fisayara	50	Ɲininkaliw Dɔgɔtɔrɔ ye sukaro ma fɛn tinin min bɔ a ka saki kɔnɔ ka a di Aminata ma,	9	female	spk_1778169797963_d3d546ba58	12.32
Aminata Fari Fisayara	53	Ɲininkaliw Mɔgɔ min ye Aminata ni a ba wele ka don dɔgɔtɔrɔso kɔnɔ,	9	female	spk_1778169797963_d3d546ba58	9.68
Aminata Fari Fisayara	54	Aminata kunkolo ni a kɔnɔbara tun tɛ ka a dimi.	9	female	spk_1778169797963_d3d546ba58	6.88
Aminata Fari Fisayara	55	Aminata siranna ka taa dɔgɔtɔrɔso la.	9	female	spk_1778169797963_d3d546ba58	3.28
Aminata Fari Fisayara	56	Dɔgɔtɔrɔ ye tulonkɛfɛn dɔ kɛ ka Aminata dusukun mankan lamɛn.	9	female	spk_1778169797963_d3d546ba58	7.36
Aminata Fari Fisayara	57	Dɔgɔtɔrɔ ye fɛn bilenman dɔ bɔ a ka saki kɔnɔ ka di Aminata ma.	9	female	spk_1778169797963_d3d546ba58	9.52
Aminata Fari Fisayara	58	Mama tun ye a miiri ko kɔnɔna bana bɛ Aminata la.	9	female	spk_1778169797963_d3d546ba58	5.36
Aminata Fari Fisayara	61	Ne ye a fɔ n ba ye ko:	9	female	spk_1778169797963_d3d546ba58	2.72
Aminata Fari Fisayara	62	Mama, ne man kɛnɛ.	9	female	spk_1778169797963_d3d546ba58	2.96
Aminata Fari Fisayara	63	Ne kunkolo bɛ ka n dimi.	9	female	spk_1778169797963_d3d546ba58	3.36
Aminata Fari Fisayara	64	Ne kɔnɔ fana bɛ ka n dimi.	9	female	spk_1778169797963_d3d546ba58	4.96
Aminata Fari Fisayara	65	’’ A ye a fɔ ne ye ko an ka kan ka taa dɔgɔtɔrɔso la.	9	female	spk_1778169797963_d3d546ba58	7.2
Aminata Fari Fisayara	66	Ne tun tɛ fɛ ka taa.	9	female	spk_1778169797963_d3d546ba58	3.84
Aminata Fari Fisayara	67	Ne tun sirannen don.	9	female	spk_1778169797963_d3d546ba58	2.32
Aminata Fari Fisayara	68	Ne tun bɛ a miiri ko dɔgɔtɔrɔ bɛ se ka ne jogin.	9	female	spk_1778169797963_d3d546ba58	7.52
Aminata Fari Fisayara	69	Ne ye Mama ɲininka ko:	9	female	spk_1778169797963_d3d546ba58	3.6
Aminata Fari Fisayara	70	Dɔgɔtɔrɔ bɛna mun kɛ?	9	female	spk_1778169797963_d3d546ba58	2.56
Aminata Fari Fisayara	71	Mama ko ne ma:	9	female	spk_1778169797963_d3d546ba58	3.04
Aminata Fari Fisayara	72	Dɔgɔtɔrɔ bɛna i lajɛ,	9	female	spk_1778169797963_d3d546ba58	4
Aminata Fari Fisayara	73	Aminata. A bɛna fura di i ma.	9	female	spk_1778169797963_d3d546ba58	4.08
Aminata Fari Fisayara	74	O la bɛ i kɛnɛya.	9	female	spk_1778169797963_d3d546ba58	3.76
Bakɔrɔnin Saba	1	A kɛra teliya ani nɔgɔya la.	15	male	spk_1778156944522_5bdcbdca99	3.84
Bakɔrɔnin Saba	2	O ye a to Awa fɛrɛla a ka tulonkɛ ma.	15	male	spk_1778156944522_5bdcbdca99	4.4
Bakɔrɔnin Saba	3	U bɛɛ la dɔgɔnin Awa ye a ka so jɔ ni karata ye.	15	male	spk_1778156944522_5bdcbdca99	6.64
Bakɔrɔnin Saba	4	ale ye a ka so jɔ ni bɔgɔ ye.	15	male	spk_1778156944522_5bdcbdca99	3.76
Bakɔrɔnin Saba	5	A jɔli mɛɛnna ka tɛmɛ,	15	male	spk_1778156944522_5bdcbdca99	2.72
Bakɔrɔnin Saba	6	karata so in kan.	15	male	spk_1778156944522_5bdcbdca99	2.72
Bakɔrɔnin Saba	7	Nka Bintu ka so tun sinsinnen don.	15	male	spk_1778156944522_5bdcbdca99	5.68
Bakɔrɔnin Saba	8	Bintu tun ye cɛmancɛ balimamuso ye,	15	male	spk_1778156944522_5bdcbdca99	4.4
Bakɔrɔnin Saba	9	A jɔli mɛɛnna,	15	male	spk_1778156944522_5bdcbdca99	2
Bakɔrɔnin Saba	10	nka a tun sinsinnen do ka tɛmɛ tɔw bɛɛ ta kan.	15	male	spk_1778156944522_5bdcbdca99	5.76
Bakɔrɔnin Saba	11	Mariyamu, u bɛɛ la kɔrɔmuso ye a ka so jɔ ni biriki ye.	15	male	spk_1778156944522_5bdcbdca99	5.52
Bakɔrɔnin Saba	12	Surukuba nana dugu kɔnɔ.	15	male	spk_1778156944522_5bdcbdca99	4.16
Bakɔrɔnin Saba	13	A ye Awa ka karata so ye,	15	male	spk_1778156944522_5bdcbdca99	3.44
Bakɔrɔnin Saba	14	a ye o yɛ ka o ci ka bɔ yen.	15	male	spk_1778156944522_5bdcbdca99	4.16
Bakɔrɔnin Saba	15	Awa bolila ka taa Bintu ka so.	15	male	spk_1778156944522_5bdcbdca99	4
Bakɔrɔnin Saba	16	Don dɔ,	15	male	spk_1778156944522_5bdcbdca99	1.68
Bakɔrɔnin Saba	17	A ye o fana yɛ ka o ci ka bɔ yen.	15	male	spk_1778156944522_5bdcbdca99	4.88
Bakɔrɔnin Saba	18	Awa ni Bintu la bɛɛ bolila ka taa Mariyamu ka so.	15	male	spk_1778156944522_5bdcbdca99	5.52
Bakɔrɔnin Saba	19	Surukuba nana Bintu ka so.	15	male	spk_1778156944522_5bdcbdca99	3.2
Bakɔrɔnin Saba	20	So in tun sinsinnen don ka ɲɛ kosɛbɛ.	15	male	spk_1778156944522_5bdcbdca99	4.8
Bakɔrɔnin Saba	21	Surukuba sɛngɛnna, a taara.	15	male	spk_1778156944522_5bdcbdca99	3.92
Bakɔrɔnin Saba	22	Balimamuso ninnu ɲuman bɔra.	15	male	spk_1778156944522_5bdcbdca99	3.92
Bakɔrɔnin Saba	23	Surukuba ye Mariyamu ka so birikima yɛ kosɛbɛ nka fosi ma kɛ a la.	15	male	spk_1778156944522_5bdcbdca99	7.12
Bakɔrɔnin Saba	24	Balimamuso ninnu nisɔndiyara,	15	male	spk_1778156944522_5bdcbdca99	3.68
Bakɔrɔnin Saba	25	wa u ye u ka ɲɛnamaya kɛ lakana la u ka biriki sow kɔnɔ.	15	male	spk_1778156944522_5bdcbdca99	8.8
Bakɔrɔnin Saba	26	Dugu kɔnɔ bakɔrɔnninw bɛɛ ye ɲɛnajɛ kɛ balimamuso ninnu ka cɛfarinya ni u ka so barikama sinsinnew la.	15	male	spk_1778156944522_5bdcbdca99	12.8
Bakɔrɔnin Saba	27	O kɔfɛ, Awa ni Bintu ye so barikamaw dilanni dege i n'a fɔ Mariyamu.	15	male	spk_1778156944522_5bdcbdca99	9.92
Bakɔrɔnin Saba	30	Ɲininkaliw Awa ye a ka so dilan ni mun ye?	15	male	spk_1778156944522_5bdcbdca99	4.72
Bakɔrɔnin Saba	32	Ɲininkaliw Jɔn ma se ka Mariyamu ka so yɛ ka a ci ka bɔ yen?	15	male	spk_1778156944522_5bdcbdca99	6.8
Bakɔrɔnin Saba	33	Awa ni Bintu ye mun jɔli dege?	15	male	spk_1778156944522_5bdcbdca99	3.68
Bakɔrɔnin Saba	35	Ɲininkaliw Surukuba taalen kɔ,	15	male	spk_1778156944522_5bdcbdca99	4.16
Bakɔrɔnin Saba	37	Ɲininkaliw Jɔn ye a ka so jɔ ni karata ye?	15	male	spk_1778156944522_5bdcbdca99	4.8
Bakɔrɔnin Saba	38	Surukuba Bintu Mariyamu	15	male	spk_1778156944522_5bdcbdca99	2.4
Bakɔrɔnin Saba	39	Nin ye jɔn ye?	15	male	spk_1778156944522_5bdcbdca99	1.68
Bakɔrɔnin Saba	41	Ɲininkaliw Mariyamu ye a ka sɔ jɔ ni mun ye?	15	male	spk_1778156944522_5bdcbdca99	4.8

End of preview. Expand in Data Studio

Bambara Educational Speech Dataset

This dataset is a collection of READ Bambara text based on educational children's books from RobotsMali's GAIFE project. It is designed to support the training and benchmarking of Automatic Speech Recognition (ASR) models, with a particular focus on child speech, regional acoustics, and repetitive text structures (inherent to the domain).

The dataset is structured into two separate subsets to support specialized training and evaluation paradigms:

main: Contains non-overlapping text and speakers divided into standard training and test splits.
duplicate: A highly dense, multi-speaker redundant training set featuring multiple recordings of the same source literature by a diverse pool of speakers.

Dataset Architecture & Splits

The dataset enforces a strict split philosophy to ensure structural and evaluation integrity. A book present in the test split (the benchmark) never appears in another split. Though there might be rare speaker overlap due to inconsistent speaker identity handling araising from the data collection pipeline.

Summary Statistics Table

Metric	`main` (Train)	`main` (Test)	`duplicate` (Train)	Combined
Total Unique Speakers	8	11	113	113
Mean Audio Duration	4.86s	4.42s	7.73s	4.74s
Audio Duration Range	0.72s – 25.60s	0.32s – 16.48s	0.08s – 37.52s	0.08s – 37.52s
Mean Sentence Length	8.06 words	7.18 words	7.85 words	7.85 words
Sentence Length Range	1 – 26 words	1 – 21 words	1 – 26 words	1 – 26 words
Missing Speaker Metadata	0 utterances	38 utterances	0 utterances	0 utterances

Demographic Distributions

Age Profile: The dataset exclusively features children and young adolescents. Across the entire combined footprint, speaker age ranges from a minimum of 8 to a maximum of 19 years old.
Gender Splits (Combined): Female: 17,471 utterances | Male: 17,213 utterances.
Gender Splits (main Test): Female: 522 | Male: 164 | Unknown: 38.
Gender Splits (main Train): Female: 783 | Male: 420.

Detailed Book Inventory & Duplicate Metrics

The duplicates subset is derived from 20 core books.

`duplicate` (Train) Subset Book Inventory (44.02 h)

Book Title	Total Reads (Proxied)	Unique Speaker IDs	Female Reads	Male Reads	Age Distribution Split (<10 / 10-15 / 15-20)
Aminata Fari Fisayara	36	34	17	19	3 / 19 / 14
Bakɔrɔnin Saba	37	36	18	19	3 / 20 / 14
Bɛnkɛ Tɔm Ka So	26	24	14	12	1 / 13 / 12
Dawuda ni a Mɔkɛ	30	30	15	15	1 / 15 / 14
Dɔgɔtɔrɔ ni Farafinfurabɔla a	23	22	14	9	2 / 10 / 11
Filomani	32	31	18	14	2 / 16 / 14
Gawusu ni Masakɛ Sidiki	29	29	14	15	3 / 16 / 10
Gerenadi-Feerew	29	27	14	15	1 / 17 / 11
Gesedala Musa	27	26	15	12	2 / 15 / 10
Gundola Kuma	31	27	18	13	3 / 18 / 10
Kalo la Taama	26	25	13	13	1 / 15 / 10
Kan Orobotik	27	27	15	12	2 / 12 / 13
Korokara Yɛrɛdɔnbali	24	24	12	12	0 / 14 / 10
Kurun	25	25	11	14	0 / 14 / 11
Lamini Ka Don Kɛrɛnkɛrɛnnen	24	24	10	14	1 / 13 / 10
Mama ka Sama	24	24	10	14	0 / 13 / 11
Ne ni Mama ka Gafe Kalan	27	26	12	15	3 / 16 / 8
Subahana Daga	29	28	12	17	2 / 17 / 10
Sɔminiminɛnw	29	29	13	16	3 / 13 / 13
Yɛlɛ Ka Di Npogotiginin Mi Ye	21	20	9	12	0 / 13 / 8

`main` (Train) Subset Book Inventory (1.62 h)

Made of unique readings of 22 books, the 20 books in duplicate + 2:

Saratu
Tulonkɛw

Test Subset Book Inventory (0.89 h)

The following distinct literary works compose the main test subset footprint (no repetition):

Anw Bɛ Baara Kɛ!
Bako Cɛnin Ŋaniya Ɲuman
Bama Miirina
Cɛni Tulogɛlɛn
Donfɛnw
Fali Nalonma Ni Ba Kegunma ani Ɲininkaliw ni u j
Jate
Kogo
Kulɔriw
Ne ni Papa ka Gafe Kalan
Ni a tun bɛ se ...

Critical Considerations: Features vs. Weaknesses

When developing models on this dataset, users should balance its unique profile against known recording constraints:

1. The High-Volume Duplication Matrix

As a Feature: Due to the severe scarcity of open-source text and educational literature in Bambara, collecting deep audio variations on a finite set of text was a deliberate design choice. This split provides a robust playground for specific speech experiments, such as acoustic multi-speaker verification, text-constrained acoustic profiling, and downstream speech representation probing.
As a Weakness: The textual diversity in the duplicate subset is inherently bottlenecked by the underlying literature. Models trained aggressively on the duplicate split without constraint can rapidly overfit to the vocabulary, tone structures, and phonetic bounds of these 20 specific books.

2. Known Metadata Inconsistencies

The Identifier Inconsistency: The log metrics show 113 unique speaker IDs in the combined dataset slice. However, on-the-ground project coordinators reported a real-world count of approximately 60 unique speakers.
Implication: This discrepancy highlights an operational metadata inflation error where individual speakers were assigned differing tracking IDs across different recording sessions, days, or environments. Users should exercise caution when benchmarking strict zero-shot speaker verification algorithms on this dataset without manual speaker clustering.

Dataset Format

Data manifests are deployed using the standard JSON Lines (.jsonl) configuration natively compatible with deep learning toolkits like NVIDIA NeMo and ESPnet:

{
  "BookTitle": "Aminata Fari Fisayara",
  "sentenceID": 1,
  "text": "Mɔgɔ caman tun bɛ yen.",
  "speakerAge": 15,
  "speakerGender": "male",
  "speakerID": "spk_1778182779107_12bb3ea9c8",
  "duration": 2.56,
  "audio_filepath": "data/audios/dup_Aminata_Fari_Fisayara_1_1200.wav"
}