---
language:
  - luy
  - en
  - sw
license: cc-by-4.0
task_categories:
  - translation
  - text-generation
pretty_name: Luhya Multilingual Dataset
size_categories:
  - 10K<n<100K
tags:
  - luhya
  - luyia
  - kenyan-languages
  - african-languages
  - low-resource
  - multilingual
  - translation
  - indigenous-languages
  - project-tafsiri
---
 
# Luhya Multilingual Dataset
 
**Curator:** Dr. Moody Amakobe  
**Project:** [Project Tafsiri — Bridging Indigenous Languages and AI](https://www.riskinfo.ai/post/project-tafsiri-bridging-indigenous-languages-and-ai)  
**Version:** 2.0  
**License:** Creative Commons Attribution 4.0 (CC BY 4.0)
 
---
 
## Dataset Overview
 
This dataset is a comprehensive, structured multilingual corpus for the **Luhya language** 
(also written Luyia), a Bantu language cluster spoken primarily in western Kenya by the 
Abaluhya people — Kenya's second largest ethnic group with approximately 7 million speakers.
 
The dataset was created as part of **Project Tafsiri**, an initiative to bridge indigenous 
Kenyan languages with modern AI systems by building high-quality, culturally grounded 
training data for low-resource language models.
 
The corpus covers **4 domains**, **9 dialects**, and provides trilingual alignment across 
**Luhya ↔ English ↔ Swahili** — making it suitable for machine translation, language 
modelling, dialect classification, and cross-lingual transfer learning tasks.
 
---
 
## Dataset Statistics
 
| Metric | Count |
|--------|-------|
| Total rows | 22,269 |
| Dialects covered | 9 |
| Domains covered | 4 |
| Rows with English | 22,268 (99.9%) |
| Rows with Swahili | 6,534 (29.3%) |
| Fully trilingual rows | 6,533 |
 
---
 
## Domains
 
| Domain | Rows | Description |
|--------|------|-------------|
| **Dictionary** | 11,493 | Word-level entries with English definitions, part of speech, and notes |
| **Conversational** | 6,447 | Naturally occurring sentences from community speech |
| **Proverbs** | 2,331 | Traditional Luhya proverbs with cultural annotations |
| **Bible** | 1,998 | Parallel scripture verses (Luhya ↔ English) |
 
---
 
## Dialects Covered
 
The Luhya language is a cluster of 16–18 mutually intelligible dialects. This dataset 
currently covers the following:
 
| Dialect | Description |
|---------|-------------|
| **Bukusu** | Widely spoken in Bungoma County; famous for circumcision ceremonies and oral literature |
| **Wanga / Luwanga** | Kakamega region dialect with rich oral traditions |
| **Maragoli** | Spoken in Vihiga County; known for formal writing and traditional music |
| **Marachi** | Spoken by the Marachi people in Butula, Busia County |
| **Kisa** | Close to Idakho; spoken in Kakamega County |
| **Marama** | Central Luhya cluster, Kakamega County |
| **Tsotso** | Closely related to other central dialects, Kakamega County |
| **Isukha** | Kakamega region; closely related to Idakho |
| **Nyala** | Spoken by the Nyala people in Busia district |
 
---
 
## Dataset Fields
 
| Field | Type | Description |
|-------|------|-------------|
| `id` | string | Unique UUID for each row |
| `luhya_text` | string | Source text in Luhya (always present) |
| `english_text` | string | English translation or definition (99.9% coverage) |
| `swahili_text` | string | Swahili translation (29% coverage, primarily conversational domain) |
| `dialect_id` | string | UUID linking to the dialect reference table |
| `dialect_name` | string | Human-readable dialect name (e.g. "Bukusu", "Wanga") |
| `domain` | string | Content domain: Bible / Dictionary / Proverbs / Conversational |
| `subdomain` | string | Finer-grained context (e.g. "Genesis 1:1", "Bukusu Dictionary", "general_wisdom") |
| `pos` | string | Part of speech — Dictionary rows only (n, v, adj, adv, etc.) |
| `quality_score` | float | Quality score where available (Bible domain) |
| `is_validated` | bool | Whether the entry has been human-validated |
| `notes` | string | Cultural context, noun class, cross-references, etymology |
| `source_table` | string | Originating database table for provenance tracing |
| `source_id` | string | Row ID in the originating table |
 
---
 
## Data Sources
 
This dataset was compiled from the following primary sources:
 
### 1. Lubukusu-English Dictionary (Academic)
**Marlo, Michael; Sifuna, Adrian; Wasike, Aggrey (2008)**  
A comprehensive ~6,200-word Bukusu-English dictionary compiled at Indiana University and 
Michigan State University. Peer-reviewed and tone-marked. Available on Academia.edu.
 
### 2. Wanga-English Dictionary (Academic)
**Compiled with Alfred Anangwe (2008)**  
A ~4,000-word preliminary Wanga-English dictionary compiled as part of the Luyia Dictionary 
Project at the University of Michigan. Available on Academia.edu.
 
### 3. IBIBILIA INDAKATIFU — The New Testament in Luhya
*Matayo–Obufwimbuli | Oluluyia*  
**Bible Society of Kenya, 2019**  
Provides formal religious language structure and vocabulary. A foundational text for 
standardised Luhya orthography.
 
### 4. KenTrans: A Parallel Corpora for Swahili and Local Kenyan Languages
**Wanzare, Lilian D.A; Indede, Florence; McOnyango, Owen; Ombui, Edward; Wanjawa, Barack; 
Muchemi, Lawrence (2022)**  
[https://doi.org/10.7910/DVN/NOAT0W](https://doi.org/10.7910/DVN/NOAT0W) — Harvard Dataverse, V2  
Offers professionally validated translation pairs and linguistic patterns across Swahili and 
local Kenyan languages including Luhya.
 
### 5. Collection of 100 Nyala Proverbs
**Kevin Namatsi Okubo**  
Africa Proverb Working Group, Nairobi, Kenya, April 2016  
Preserves traditional wisdom and cultural knowledge systems specific to the Nyala dialect.
 
### 6. Luyia Proverbs from Kisa, Marama, Tsotso and Wanga
**Tim Wambunya**  
Provides cross-dialectal cultural content and variation patterns across four Luhya dialects.
 
### 7. Community Dictionary Contributions
Structured word entries collected from native speakers across Wanga and Luwanga dialects, 
including noun class information, etymology, pronunciation notes, and cultural context.
 
---
 
## About Project Tafsiri
 
[Project Tafsiri](https://www.riskinfo.ai/post/project-tafsiri-bridging-indigenous-languages-and-ai) 
is an initiative dedicated to building AI infrastructure for indigenous Kenyan languages. 
The project focuses on creating high-quality, culturally authentic training data to ensure 
that AI systems can serve speakers of low-resource languages — particularly those in 
communities that have historically been excluded from the benefits of language technology.
 
The LuhyaAI system built on this dataset is grounded in both contemporary usage and 
traditional knowledge. By combining conversational sentences, formal scripture, academic 
dictionaries, and community proverbs, the dataset captures the full linguistic and cultural 
range of the Luhya language — creating a foundation for AI systems that are not only 
accurate but culturally resonant.
 
---
 
## Citation
 
If you use this dataset in your research, please cite:
 
```
@dataset{amakobe2025luhya,
  author    = {Amakobe, Moody},
  title     = {Luhya Multilingual Dataset},
  year      = {2025},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/datasets/mamakobe/luhya-multilingual-dataset},
  note      = {Part of Project Tafsiri — https://www.riskinfo.ai/post/project-tafsiri-bridging-indigenous-languages-and-ai}
}
```
 
---
 
## Contact
 
For questions, contributions, or collaboration opportunities related to Project Tafsiri 
or the LuhyaAI initiative, please reach out via the 
[Project Tafsiri page](https://www.riskinfo.ai/post/project-tafsiri-bridging-indigenous-languages-and-ai).