From Speech to Text Corpora: Evaluating ASR-Based Data Acquisition for Low-Resource Fongbe and Hausa

2026-06-20

AI Summary

This study addresses the scarcity of high-quality text corpora for low-resource African languages by systematically evaluating automatic speech recognition (ASR) for both tonal (Fongbe) and non-tonal (Hausa) languages. The authors fine-tune MMS-300M and Whisper-Small models to transcribe speech from YouTube videos, incorporating domain-diverse data selection and human-in-the-loop quality scoring. Their approach substantially improves transcription accuracy: Fongbe achieves a word error rate of 9.48%—a 78% relative reduction—while preserving critical tonal diacritics; Hausa reaches a quality score of 57.4/100, approaching usability thresholds. The project publicly releases 12.3 hours of Fongbe text, 45.49 hours of transcribed audio, and a complete video catalog, establishing a new paradigm for resource-constrained language modeling.

Abstract

Low-resource African languages lack text corpora needed for language model training. We investigate whether ASR pipelines can extend text resources for two typologically distinct West African languages: Fongbe (tonal, diacritic-rich) and Hausa (non-tonal). We fine-tune MMS-300M on a curated 12.3-hour Fongbe dataset, achieving 9.48% WER on the ALFFA benchmark - a 78% relative reduction from the prior 44.04% baseline - while preserving tonal diacritics critical to the language. For Hausa, we apply an existing fine-tuned Whisper-Small model. We catalog 1,553 YouTube videos (236 hours) and process a subset of 424 videos (45.49 hours) selected to balance domain diversity with available computational resources, producing 6,770 transcribed segments. Human evaluation on 50 randomly sampled segments per language shows mean quality scores of 57.4/100 for Hausa and 36.5/100 for Fongbe, indicating that while Hausa transcriptions approach acceptable quality for corpus construction, Fongbe transcriptions require post-processing or improved models for production use. We release the curated dataset, fine-tuned model, transcribed corpus, and full video catalog following platform terms and ethical guidelines.

Problem

• low-resource languages
• text corpora
• language model training
• Fongbe
• Hausa

Innovation

• low-resource languages
• automatic speech recognition (ASR)
• tonal diacritics preservation
• corpus construction
• model fine-tuning