Ewe - UNICEF WCARO NLP

🌍 General Information

ISO 639-3ewe

ISO 639-1ee

French Nameéwé

Glottocodeewee1241

Language FamilyNiger–Congo

Writing SystemLatin (Ewe alphabet) Ewe Braille

L1 Speakers5.0 million

L2 Speakers—

Population3820000 (10⁶+)

Endangermentsafe

CountriesBenin, Ghana, Togo

WikipediaEwe on Wikipedia

🗺️ Language Use

Data from CLEAR Global's Language Use Data Platform.

👥 Actors (8)

AdwumaTech AI (Startup)

AdwumaTech AI provides enterprise-grade AI training data infrastructure powered by African engineering talent. Core offerings include RLHF for code ge...

Farmerline / Darli AI (Startup)

Ghanaian agritech company building technology solutions for smallholder farmers. Core platform Mergdata is a cloud-based farm management system used b...

Ghana NLP / Khaya AI (Research Community)

Ghana NLP is an open-source initiative focused on Natural Language Processing for Ghanaian and African languages. They developed Khaya, the first AI-p...

HCI and Persuasive Systems Lab, University of Ghana (University)

Research lab within the Department of Computer Science at University of Ghana, led by Prof. Isaac Wiafe. Focused on Human-Computer Interaction, persua...

LAfricaMobile (Startup)

Communication platform serving 300+ companies across Africa. Provides messaging (SMS, WhatsApp Business API, Voice), USSD, and AI-powered Text-to-Spee...

ML Collective (Ngo)

Non-profit research lab providing research opportunities to independent and underrepresented researchers. Runs multiple African language NLP projects ...

UBC Deep Learning & NLP Lab (University)

Research lab at the University of British Columbia led by Muhammad Abdul-Mageed, focused on deep representation learning and natural language socio-pr...

Umbaji (Startup)

Togo-based organization building NLP resources for Togolese languages. Developed the Eyaa-Tom multilingual speech corpus (30.9 hours across 10 Togoles...

🤖 NLP & Tech Resources

✓ AfroLID

✓ Google Translate: 2022

✓ NLLB-200

✓ Mozilla Common Voice: ☑️

✓ MADLAD-400 Docs: 4500 MAD-Docs

✓ MADLAD-400 Sentences: 246700 MAD-Sents

✓ Fineweb2: 💬

✓ CMU Wilderness

✓ MAFAND

✓ MMS-ASR

✓ MMS-TTS

✓ MMS-LID

✓ Omnilingual ASR

✓ SimbaBench

✓ WaxalNLP

Benchmarks

ASR (Speech-to-Text)

Waxal-test(<40s)

Model	WER	CER	Source
facebook/omniASR-LLM-3B	50.03	16.89	evaluated
facebook/omniASR-LLM-7B	41.4	12.97	evaluated

Waxal

Model	WER	CER	Source
facebook/mms-1b-all	50.35	15.37	evaluated
UBC-NLP/Simba-S	113.11	51.12	evaluated

TTS (Text-to-Speech)

SimbaBench (bibleTTS, Ewe)

Model	WER	UTMOS	PESQ	Source
facebook/mms-tts	15.94	3.01	1.5	reported

MT (Machine Translation)

FLORES-200

Model	direction	chrF	BLEU	Source
AfriqueQwen-14B	eng→ewe	6.22	—	reported
AfriqueQwen-14B	ewe→eng	21.97	—	reported
Qwen3-14B	eng→ewe	6.26	—	reported
Qwen3-14B	ewe→eng	22.35	—	reported
AfriqueGemma-12B	eng→ewe	8.47	—	reported
AfriqueGemma-12B	ewe→eng	19.71	—	reported
Gemma3-12B	eng→ewe	7.45	—	reported
Gemma3-12B	ewe→eng	13.92	—	reported
AfriqueLlama-8B	eng→ewe	6.8	—	reported
AfriqueLlama-8B	ewe→eng	20.78	—	reported
google/madlad400-3b-mt	ewe→eng	33.3	10.3	reported
google/madlad400-3b-mt	eng→ewe	26.9	5.2	reported
google/madlad400-7.2b-mt	ewe→eng	37.4	14.3	reported
google/madlad400-7.2b-mt	eng→ewe	29.2	6.8	reported
google/madlad400-10.7b-mt	ewe→eng	36.6	13.3	reported
google/madlad400-10.7b-mt	eng→ewe	28.4	5.9	reported
google/madlad400-8b-lm	ewe→eng	24.3	4.4	reported
google/madlad400-8b-lm	eng→ewe	5.9	0.1	reported

FLORES+ 10-sample

Model	direction	SpBLEU	ChrF	Source
anthropic/claude-sonnet-4.5	ewe→eng	16.0	41.3	reported
anthropic/claude-sonnet-4.5	eng→ewe	12.1	37.4	reported
google/gemini-2.5-pro	ewe→eng	15.6	43.2	reported
google/gemini-2.5-pro	eng→ewe	15.4	43.5	reported
google/gemini-3-pro-preview	ewe→eng	24.0	48.3	reported
google/gemini-3-pro-preview	eng→ewe	16.1	44.6	reported
google/translate-v2	eng→ewe	17.3	46.0	reported
google/translate-v2	ewe→eng	19.2	40.5	reported
openai/gpt-5	eng→ewe	24.1	46.1	reported
openai/gpt-5	ewe→eng	8.6	27.9	reported

GATONES

Model	direction	BLEU	chrF	Source
google/madlad400-3b-mt	ewe→eng	8.6	25.8	reported
google/madlad400-3b-mt	eng→ewe	4.1	23.7	reported
google/madlad400-7.2b-mt	ewe→eng	11.1	28.9	reported
google/madlad400-7.2b-mt	eng→ewe	6.1	27.1	reported
google/madlad400-10.7b-mt	ewe→eng	10.4	28.3	reported
google/madlad400-10.7b-mt	eng→ewe	6	26.6	reported
google/madlad400-8b-lm	ewe→eng	3.1	16.4	reported
google/madlad400-8b-lm	eng→ewe	0	2.1	reported

NTREX

Model	direction	BLEU	chrF	Source
google/madlad400-3b-mt	ewe→eng	13.2	35.1	reported
google/madlad400-3b-mt	eng→ewe	6.1	29.1	reported
google/madlad400-7.2b-mt	ewe→eng	16.3	38.4	reported
google/madlad400-7.2b-mt	eng→ewe	7.4	31.6	reported
google/madlad400-10.7b-mt	ewe→eng	16.7	38.5	reported
google/madlad400-10.7b-mt	eng→ewe	7.2	30.5	reported
google/madlad400-8b-lm	ewe→eng	2.8	19.6	reported
google/madlad400-8b-lm	eng→ewe	0.1	4.5	reported

LLM (Large Language Models)

FLORES-200

Model	Log-perplexity	Source
goldfish-models/ewe_latn_full	128.71	reported

AfroBench

Model	Intent	Topic	Source
AfriqueQwen-14B	14.88	46.52	reported
Qwen3-14B	15.38	47.07	reported
AfriqueGemma-12B	13.84	43.23	reported
Gemma3-12B	15.97	47.47	reported
AfriqueLlama-8B	9.66	38.32	reported

AfriMGSM(Math)

Model	Accuracy	Source
Aya-101	3.2	reported
Gemma 2 27B	7.6	reported
LLaMa 3.1 70B	8.0	reported
GPT-4o	8.8	reported

AfriXNLI

Model	Accuracy	Source
Aya-101	43.2	reported
Gemma 2 27B	36.8	reported
LLaMa 3.1 70B	34.3	reported
GPT-4o	45.0	reported

AfriMMLU

Model	AfriMMLU-ElemMath	AfriMMLU-GlobalFacts	AfriMMLU-Geography	AfriMMLU-MicroEcon	AfriMMLU-IntLaw	Source
GPT-4o	55.0	32.0	25.0	26.0	36.0	reported

SIB-200 10-sample

Model	Classification Acc.	Source
anthropic/claude-opus-4.6	100.0	reported
google/gemini-2.5-pro	100.0	reported
meta-llama/llama-4-maverick	100.0	reported

SAHARA

Model	Lang. Avg.	XNLI Acc	MMLU Acc.	spBLEU fra→ewe	MGSM	NER F1	Source
gemini-3-pro-preview	68.97	75.81	95.24	17.63	76.19	80.0	reported
gemini-2.5-pro	63.83	64.52	92.06	14.75	73.02	74.79	reported
gpt-5-2025-08-07	46.35	50.0	74.6	8.36	26.98	71.79	reported
gpt-5.1-2025-11-13	46.66	54.84	73.02	7.79	26.98	70.69	reported
gemini-2.5-flash	47.41	54.84	79.37	0.81	65.08	36.97	reported
claude-sonnet-4-5@20250929	38.1	61.29	44.44	8.61	36.51	39.65	reported
claude-4-sonnet-20250514	44.48	58.06	69.84	9.38	34.92	50.21	reported
gpt-4.1	30.28	56.45	52.38	6.88	7.94	27.76	reported
CohereForAI/c4ai-command-a-03-2025	28.09	46.77	57.14	6.35	11.11	19.1	reported
meta-llama/Llama-4-Scout-17B-16E-Instruct	24.4	43.55	47.62	5.22	9.52	16.08	reported
meta-llama/Llama-3.3-70B-Instruct	23.52	33.87	52.38	5.08	3.17	23.08	reported
google/gemma-3-27b-it	22.02	38.71	49.21	5.69	0.0	16.51	reported
google/gemma-2-27b-it	22.41	43.55	39.68	4.2	0.0	24.6	reported
deepseek-ai/DeepSeek-R1-Distill-Qwen-32B	22.6	37.1	49.21	3.64	1.59	21.47	reported
meta-llama/Llama-3.1-70B-Instruct	23.62	43.55	50.79	5.31	4.76	13.7	reported
Qwen/Qwen3-30B-Instruct-2507	22.86	41.94	49.21	2.71	3.17	17.28	reported
deepseek-ai/DeepSeek-R1-Distill-Llama-70B	22.74	43.55	41.27	5.34	4.76	18.79	reported
Tower-Babel/Babel-9B-Chat	17.2	35.48	41.27	3.75	1.59	3.92	reported
Tower-Babel/Babel-83B-Chat	20.51	29.03	55.56	2.57	1.59	13.79	reported
Qwen/Qwen3-4B-Instruct-2507	19.01	45.16	34.92	2.77	4.76	7.45	reported
google/gemma-2-9b-it	18.54	40.32	39.68	1.88	1.59	9.25	reported
google/gemma-3-12b-it	19.65	43.55	38.1	6.29	0.0	10.29	reported
meta-llama/Llama-3.1-8B-Instruct	16.92	43.55	38.1	1.38	1.59	0.0	reported
CohereForAI/aya-23-35B	14.51	35.48	28.57	2.27	1.59	4.62	reported
CohereForAI/c4ai-command-r-plus-08-2024	16.33	33.87	30.16	2.4	6.35	8.89	reported
CohereForAI/c4ai-command-r7b-12-2024	13.81	35.48	30.16	1.81	1.59	0.0	reported
microsoft/Phi-4-mini-instruct	14.18	35.48	30.16	3.69	1.59	0.0	reported
meta-llama/Llama-3.2-1B-Instruct	14.17	33.87	30.16	2.07	4.76	0.0	reported
microsoft/Phi-3.5-mini-instruct	14.61	32.26	28.57	3.49	1.59	7.14	reported
google/gemma-2-2b-it	12.35	29.03	26.98	2.44	0.0	3.28	reported
CohereForAI/aya-23-8B	12.15	33.87	20.63	1.77	3.17	1.32	reported
google/gemma-3-4b-it	16.81	43.55	28.57	3.91	0.0	8.0	reported
meta-llama/Llama-3.2-3B-Instruct	12.92	33.87	28.57	0.58	1.59	0.0	reported

Noteworthy models without benchmark

Model	Task	Notes
Ghana-NLP/naafi-tts-demo	TTS	Ewe TTS demo by GhanaNLP. No public benchmarks.

📝 Notes

Observations

Survey Focus 1 for Togo (2026 UNICEF WCAR Local Languages Survey)
Also in Use Cases Document: Ghana AGOO platform (3rd most listened language on current IVR platform)

Data

GhanaNLP Parallel Corpora — 4,985 human-translated Ewe-English sentence pairs. Multi-domain, dialect-aware. CC BY-NC-SA 4.0. Datasets on HuggingFace but not tagged with language IDs: https://huggingface.co/collections/Ghana-NLP/data-collection-project

ASR Models (23)

Model	Size	Downloads	Likes
MahmoudAshraf/mms-300m-1130-forced-aligner	0.3B	4.0M	83
facebook/mms-1b-all	1.0B	2.0M	195
facebook/mms-1b-fl102	1.0B	8.7K	29
UBC-NLP/Simba-S	2B	2.2K	4
facebook/mms-1b-l1107	1.0B	280	12
espnet/xeus	—	60	146
UBC-NLP/Simba-W	2B	44	0
UBC-NLP/Simba-M	—	33	0
bc7ec356/heep-universal	0.8B	31	0
onnx-community/mms-300m-1130-forced-aligner-ONNX	—	25	0
johntsi/ZeroSwot-Medium_asr-cv_en-to-200	0.4B	17	0
johntsi/ZeroSwot-Large_asr-cv_en-to-200	0.4B	16	1
UBC-NLP/Simba-SLID-49	94.6M	15	0
johntsi/ZeroSwot-Medium_asr-mustc_en-to-200	0.4B	14	0
johntsi/ZeroSwot-Large_asr-mustc_en-to-200	0.4B	12	0
UBC-NLP/Simba-H	—	5	0
romara-labs/mms-300m-1130-forced-aligner-ONNX	—	4	0
UBC-NLP/Simba-X	—	3	0
dodziraynard/Shona	0.2B	1	1
dodziraynard/Shona2	0.2B	1	1

View on HuggingFace: ee: 22 | ewe: 1

TTS Models (6)

Model	Size	Downloads	Likes
analist/oute_ewe_r64_16bit	1B	177	0
analist/gaia_oute_ewe_r64_16bit	1B	38	0
rnjema-unima/mms-tts-ewe-baseline	36.3M	19	0
Dapor/Dapson	—	0	4

View on HuggingFace: ee: 3 | ewe: 3

MT Models (31)

Model	Size	Downloads	Likes
facebook/nllb-200-distilled-600M	—	1.1M	880
google/madlad400-3b-mt	3B	329.0K	192
facebook/nllb-200-distilled-1.3B	—	147.0K	181
facebook/nllb-200-3.3B	—	81.6K	443
Helsinki-NLP/opus-mt-mul-en	—	61.6K	89
facebook/nllb-200-1.3B	—	31.4K	71
Xenova/nllb-200-distilled-600M	—	4.4K	50
Helsinki-NLP/opus-mt-en-mul	—	3.3K	28
JustFrederik/nllb-200-distilled-600M-ct2-int8	—	925	1
facebook/nllb-moe-54b	—	557	123
Helsinki-NLP/opus-mt-fr-ee	—	60	0
Helsinki-NLP/opus-mt-fi-ee	—	55	0
Helsinki-NLP/opus-mt-sv-ee	—	52	0
Helsinki-NLP/opus-mt-en-nic	—	42	0
Helsinki-NLP/opus-mt-nic-en	—	41	0
Helsinki-NLP/opus-mt-en-ee	—	40	0
Helsinki-NLP/opus-mt-ee-fi	—	39	0
vonjack/opus-mt-mul-en-big	0.3B	39	4
Helsinki-NLP/opus-mt-ee-sv	—	38	0
ychenNLP/nllb-200-3.3B-easyproject	—	36	8

View on HuggingFace: ee: 30 | ewe: 1

LLM Models (27)

Model	Size	Downloads	Likes
UBC-NLP/toucan-base	0.6B	2.5K	5
UBC-NLP/toucan-1.2B	1B	1.9K	1
google/madlad400-8b-lm	9B	183	11
UBC-NLP/cheetah-1.2B	—	33	4
ZeppelinCorp/Charm_15	—	18	0
jbochi/madlad400-8b-lm	9B	14	8
asigalov61/MIDI_Qwen2.5-0.5B-Instruct	0.5B	11	1
Satyam-Singh/LLaVa-Large-Language-Virtual-Assistant-v2.1.6	47B	10	0
goldfish-models/ewe_latn_5mb	—	9	0
Kolyadual/Newton-bot-3-text-mini-8B	8B	7	1
goldfish-models/ewe_latn_10mb	—	7	0
masakhane/zephyr-7b-gemma-sft-african-alpaca	9B	4	1
McGill-NLP/gemma-2-9b-it-Injongo-intent	9B	2	0
McGill-NLP/gemma-2-9b-it-Injongo-slot	9B	2	0
ghananlpcommunity/opani-chat_1b-merged-16bit	1B	2	0
goldfish-models/ewe_latn_full	—	2	0
ChiomaBless/Chiomascreation	—	0	3
Xois/BuuClientPublicV1-MD-BOT	—	0	0
misterlakaz/Ceramic-Base-856b	—	0	6
misterlakaz/Midas-1922b	—	0	8

View on HuggingFace: ee: 20 | ewe: 7

Datasets

ASR Datasets (30)

Dataset	Rows	Downloads	Likes
google/WaxalNLP	2.6M	12.3K	219
ghananlpcommunity/navigation-corpus-speech-full-ewe	1.6K	6.3K	0
sarulab-speech/yodas2_sidon	1.5M	6.0K	60
claudefitz/WaxalNLP	2.6M	3.6K	0
BolajiJsAI/WaxalNLP	2.6M	2.5K	1
ngoloan/WaxalNLP	2.0M	2.1K	0
novelwolde36/WaxalNLP	2.5M	2.0K	1
Phsntom/WaxalNLP	2.6M	1.9K	0
Ephraimmm/WaxalNLPr	2.6M	1.8K	0
fiifinketia/WaxalNLP	2.5M	1.7K	1
zoelaenskii/zinny	2.6M	1.4K	1
0xzanuee/WaxalNLP	1.9M	1.4K	0
suryatmodulus/yodas2_sidon	—	1.1K	0
UBC-NLP/SimbaBench_dataset	748.0K	975	0
WueNLP/sib-fleurs	94.4K	260	11
bdallhrajh371/WaxalNLP	2.6M	250	0
YuRiVeRTi/V1Q	186.0K	201	3
fiifinketia/navigation-corpus-ewe-speech	49.3K	79	0
ghananlpcommunity/navigation-corpus-ewe-speech	49.3K	48	0
worldboss/ewe_bible_v2_tts	1.1K	41	1

View on HuggingFace: ewe: 16 | ee: 14

TTS Datasets (29)

Dataset	Rows	Downloads	Likes
google/WaxalNLP	2.6M	12.3K	219
ghananlpcommunity/navigation-corpus-speech-full-ewe	1.6K	6.3K	0
sarulab-speech/yodas2_sidon	1.5M	6.0K	60
claudefitz/WaxalNLP	2.6M	3.6K	0
BolajiJsAI/WaxalNLP	2.6M	2.5K	1
ngoloan/WaxalNLP	2.0M	2.1K	0
novelwolde36/WaxalNLP	2.5M	2.0K	1
Phsntom/WaxalNLP	2.6M	1.9K	0
Ephraimmm/WaxalNLPr	2.6M	1.8K	0
fiifinketia/WaxalNLP	2.5M	1.7K	1
zoelaenskii/zinny	2.6M	1.4K	1
0xzanuee/WaxalNLP	1.9M	1.4K	0
suryatmodulus/yodas2_sidon	—	1.1K	0
UBC-NLP/SimbaBench_dataset	748.0K	975	0
WueNLP/sib-fleurs	94.4K	260	11
bdallhrajh371/WaxalNLP	2.6M	250	0
YuRiVeRTi/V1Q	186.0K	201	3
fiifinketia/navigation-corpus-ewe-speech	49.3K	79	0
ghananlpcommunity/navigation-corpus-ewe-speech	49.3K	48	0
worldboss/ewe_bible_v2_tts	1.1K	41	1

View on HuggingFace: ee: 15 | ewe: 14

Translation Datasets (48)

Dataset	Rows	Downloads	Likes
facebook/flores	—	43.4K	97
openlanguagedata/flores_plus	893.0K	15.8K	124
gtak1/panlex-meanings	64.1M	6.3K	0
google/smol	798.0K	2.8K	94
Muennighoff/flores200	—	1.6K	20
aiana94/polynews-parallel	5.4M	1.4K	13
ayymen/Pontoon-Translations	3.6M	1.1K	18
Helsinki-NLP/tatoeba	—	866	56
masakhane/mafand	143.0K	463	15
sbmaruf/forai_ml_masakhane_mafand	136.0K	365	2
mteb/NTREXBitextMining	2.0K	309	1
mteb/FloresBitextMining	2.0K	303	1
visheratin/laion-coco-nllb	894.0K	255	44
cointegrated/panlex-meanings	157.0M	255	13
mteb/NTREX	2.0K	242	1
YuRiVeRTi/V1Q	186.0K	201	3
ymoslem/Tatoeba-Translations	8.6M	192	5
Helsinki-NLP/bible_para	—	191	22
slone/nllb-200-10M-sample	10.0M	168	13
bible-nlp/biblenlp-corpus	—	164	37

View on HuggingFace: ee: 30 | ewe: 18