Design and Development of Afan Oromo Speech Recognition and integrating with Dialogue System Using Whisper and RASA

Abstract Automatic Speech Recognition (ASR) for low-resource languages remains a significant challenge due to limited transcribed data and the lack of systematic guidance on model selection. This paper presents the development of an Afan Oromo ASR system using fine-tuned Whisper-small, integrated with a RASA-based conversational agent. We curated a 76.88-hour corpus, implemented persistent dataset splits, applied data augmentation, and developed a robust checkpointing system. The fine-tuned model achieved a Word Error Rate (WER) of 16.88% on the validation set and 17.42% on the held-out test set. We further detail the integration of this model with RASA, creating a complete speech-to-intent pipeline. Our results demonstrate the viability of state-of-the-art ASR for Afan Oromo and provide a practical template for deploying conversational AI in low-resource settings.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-13
DOI
https://doi.org/10.21203/rs.3.rs-10010573/v1
Primary Topic
Speech Recognition and Synthesis
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Design and Development of Afan Oromo Speech Recognition and integrating with Dialogue System Using Whisper and RASA

Adugna Deksiso, Teklu Urgessa, Worku Jifara
5 links
Research Square
Speech Recognition and Synthesis
preprint

Design and Development of Afan Oromo Speech Recognition and integrating with Dialogue System Using Whisper and RASA

Adugna Deksiso, Teklu Urgessa, Worku Jifara
preprint en

Abstract

Abstract Automatic Speech Recognition (ASR) for low-resource languages remains a significant challenge due to limited transcribed data and the lack of systematic guidance on model selection. This paper presents the development of an Afan Oromo ASR system using fine-tuned Whisper-small, integrated with a RASA-based conversational agent. We curated a 76.88-hour corpus, implemented persistent dataset splits, applied data augmentation, and developed a robust checkpointing system. The fine-tuned model achieved a Word Error Rate (WER) of 16.88% on the validation set and 17.42% on the held-out test set. We further detail the integration of this model with RASA, creating a complete speech-to-intent pipeline. Our results demonstrate the viability of state-of-the-art ASR for Afan Oromo and provide a practical template for deploying conversational AI in low-resource settings.

Research Square
Adama Science and Technology University (ET)
Adama Science and Technology University
Speech Recognition and Synthesis
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon