Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Thiomi Dataset: A Large-Scale Multimodal Corpus for Low-Resource African Languages

Hillary Mutisya, John Mugane, Gavin Nyamboga, Brian Chege, Maryruth Gathoni

Abstract

We present the Thiomi Dataset, a large-scale multimodal corpus spanning ten African languages across four language families: Swahili, Kikuyu, Kamba, Kimeru, Luo, Maasai, Kipsigis, Somali (East Africa); Wolof (West Africa); and Fulani (West/Central Africa). The dataset contains over 601,000 approved sentence-level text annotations and over 385,000 audio recordings across nine languages, collected through a dedicated community data collection platform involving over 100 contributors. The Thiomi platform collected data for nine languages; Swahili data was supplemented with existing Common Voice recordings. A multi-tier quality assurance pipeline achieves 86-100% text approval rates for the six primary languages. To validate the dataset's utility, we train and evaluate ASR, MT, and TTS models, establishing baselines across all ten languages. Our best ASR system achieves 3.24% WER on Swahili (Common Voice), reducing prior academic SOTA from 8.3% to 3.24% (5.1 percentage point absolute, 61% relative reduction), and 4.3% WER on Somali. The dataset will be published on HuggingFace. We describe the collection platform, quality assurance workflows, and baseline experiments, and discuss implications for African language technology infrastructure.

The Thiomi Dataset: A Large-Scale Multimodal Corpus for Low-Resource African Languages

Abstract

We present the Thiomi Dataset, a large-scale multimodal corpus spanning ten African languages across four language families: Swahili, Kikuyu, Kamba, Kimeru, Luo, Maasai, Kipsigis, Somali (East Africa); Wolof (West Africa); and Fulani (West/Central Africa). The dataset contains over 601,000 approved sentence-level text annotations and over 385,000 audio recordings across nine languages, collected through a dedicated community data collection platform involving over 100 contributors. The Thiomi platform collected data for nine languages; Swahili data was supplemented with existing Common Voice recordings. A multi-tier quality assurance pipeline achieves 86-100% text approval rates for the six primary languages. To validate the dataset's utility, we train and evaluate ASR, MT, and TTS models, establishing baselines across all ten languages. Our best ASR system achieves 3.24% WER on Swahili (Common Voice), reducing prior academic SOTA from 8.3% to 3.24% (5.1 percentage point absolute, 61% relative reduction), and 4.3% WER on Somali. The dataset will be published on HuggingFace. We describe the collection platform, quality assurance workflows, and baseline experiments, and discuss implications for African language technology infrastructure.

Paper Structure

This paper contains 28 sections, 2 figures, 10 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Multilingual NLP Datasets for African Languages
  4. Community Data Collection Approaches
  5. Multimodal African Language Resources
  6. Language Selection
  7. Data Collection Platform and Methodology
  8. Platform Architecture
  9. Collection Approach 1: Translation-Based (Text-First)
  10. Collection Approach 2: Audio-First (Transcription-Based)
  11. Audio Recording Infrastructure
  12. Quality Assurance
  13. Text Quality Assurance
  14. Audio Quality Assurance
  15. Inter-Annotator Agreement
  16. ...and 13 more sections

Figures (2)

  • Figure 1: Two-pipeline data collection workflow. Text-first (left) produces aligned English--target pairs with recorded audio. Audio-first (right) captures naturalistic speech with community transcription. Both feed into the same 4-stage QA pipeline.
  • Figure 2: Approved text sentences per language. Six languages exceed 64,000 approved sentences. Kipsigis, Maasai, and Wolof are in earlier collection stages. Total: 601,409 approved sentences across 9 languages.