Sagalee: an Open Source Automatic Speech Recognition Dataset for Oromo Language

2025-02-01

AI Summary

Oromo lacks high-quality automatic speech recognition (ASR) data, severely hindering ASR research for this low-resource language. To address this, we introduce the first open-source Oromo ASR dataset—comprising 100 hours of real-world speech from multiple speakers, capturing phonetic variation and spanning both clean and noisy acoustic conditions. Methodologically, we propose and evaluate Conformer-based models trained with hybrid CTC–attention or pure CTC loss, alongside fine-tuned Whisper variants, enhanced by speech preprocessing and forced alignment. Experimental results show the best Conformer achieves a word error rate (WER) of 15.32%, while fine-tuned Whisper attains a new state-of-the-art WER of 10.82%. This work establishes the first authoritative ASR baseline for Oromo, empirically validating the efficacy of large-model fine-tuning for low-resource ASR and providing a reproducible, transferable methodology for related under-resourced languages.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Other
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

We present a novel Automatic Speech Recognition (ASR) dataset for the Oromo language, a widely spoken language in Ethiopia and neighboring regions. The dataset was collected through a crowd-sourcing initiative, encompassing a diverse range of speakers and phonetic variations. It consists of 100 hours of real-world audio recordings paired with transcriptions, covering read speech in both clean and noisy environments. This dataset addresses the critical need for ASR resources for the Oromo language which is underrepresented. To show its applicability for the ASR task, we conducted experiments using the Conformer model, achieving a Word Error Rate (WER) of 15.32% with hybrid CTC and AED loss and WER of 18.74% with pure CTC loss. Additionally, fine-tuning the Whisper model resulted in a significantly improved WER of 10.82%. These results establish baselines for Oromo ASR, highlighting both the challenges and the potential for improving ASR performance in Oromo. The dataset is publicly available at https://github.com/turinaf/sagalee and we encourage its use for further research and development in Oromo speech processing.

Problem

• Oromo language
• Automatic Speech Recognition (ASR)
• Data scarcity

Innovation

• Oromo Language
• Speech Recognition Database
• Conformer and Whisper Models