AfriVoices-KE: A Multilingual Speech Dataset for Kenyan Languages

2026-04-09 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Intermediate, quick

Summary

AfriVoices-KE is a new large-scale multilingual speech dataset offering approximately 3,000 hours of audio across five Kenyan languages: Dholuo, Kikuyu, Kalenjin, Maasai, and Somali. It includes 750 hours of scripted speech and 2,250 hours of spontaneous speech, collected from 4,777 native speakers across various regions and demographics. The dataset addresses the significant underrepresentation of African languages in speech technology. Data collection involved both scripted recordings from compiled text corpora and unscripted speech elicited via textual and image prompts, utilizing a customized mobile application for contributors. Quality assurance included automated signal-to-noise ratio validation and human content review. Despite challenges like unreliable infrastructure and device compatibility, the project successfully created a foundational resource for inclusive speech technology and linguistic preservation.

Key takeaway

For NLP Engineers and AI Scientists developing speech technologies for African languages, AfriVoices-KE offers a critical resource to build more inclusive automatic speech recognition and text-to-speech systems. Your work can directly benefit from this dataset to improve model performance and address linguistic bias. Consider integrating this dataset into your training pipelines to enhance representation for Dholuo, Kikuyu, Kalenjin, Maasai, and Somali.

Key insights

AfriVoices-KE provides a large, diverse speech dataset for five underrepresented Kenyan languages.

Principles

• Diverse data collection improves linguistic representation.
• Community engagement is vital for low-resource data projects.

Method

Scripted speech was collected from translated text corpora, while spontaneous speech used textual/image prompts via a custom mobile app, with multi-layer quality assurance.

In practice

• Use mobile apps for distributed data collection.
• Implement multi-layered quality assurance.
• Partner with local mobilizers for community trust.

Topics

• AfriVoices-KE
• Multilingual Speech Dataset
• Kenyan Languages
• Automatic Speech Recognition
• Text-to-Speech

Best for: AI Scientist, NLP Engineer, Research Scientist

Related on AIssential

• WAXAL: A large-scale open resource for African language speech technology — WAXAL provides a large, open-access dataset for 27 African languages, fostering inclusive speech technology development.
• From A to B to A: Palindromic Zero-Shot Voice Conversion with Non-Parallel Data — A voice conversion framework uses KNN over WavLM to create synthetic training pairs from non-parallel data for zero-shot, multilingual VC.
• Task-Lens: Cross-Task Utility Based Speech Dataset Profiling for Low-Resource Indian Languages — Cross-task profiling of Indian speech datasets reveals untapped utility and critical resource gaps for low-resource languages and specific tasks.
• AfriSUD: A Dependency Treebank Collection for Evaluating Models on African Languages — Models exhibit a significant "syntax gap" in processing African languages, struggling with their structural diversity.
• Automatic Speech Recognition for Child Reading: A Phonemic Approach using Isolated Words in Brazilian Portuguese — A phonemic approach with fine-tuned Wav2Vec2.0 significantly improves ASR for child reading assessment in Brazilian Portuguese.
• KVoiceBench, KOpenAudioBench, and KMMAU: Agent-Driven Korean Speech Benchmarks for Evaluating SpeechLMs — Multilingual SpeechLM evaluation requires language-specific benchmarks to reveal true performance gaps and task-specific weaknesses.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.