cs.CLSep 26, 2026

Automatic Speech Recognition for the Basaà Language: A Low-Resource Approach

Authors: Sophie Gertrude Ngo Mock, Charles Moudina Varmantchaonala, Paul Dayang, Jean Michel Nlong, Christopher Gies

Organizations: Department of Mathematics and Computer Science, Faculty of Science, University of Ngaoundere, Ngaoundere, 454, Cameroon · Institute for Physics, Faculty V, Carl von Ossietzky University of Oldenburg, 26129 Oldenburg, Germany

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

The rapid advancement of Artificial Intelligence (AI) and Natural Language Processing (NLP) has revolutionized the way humans interact with machines. Among the most impactful developments is Automatic Speech Recognition (ASR), which enables computers to convert spoken language into text. Systems such as those built on deep neural networks, transformer architectures, and self-supervised learning have achieved near-human performance for well-resourced languages such as English and French. Yet, these advances have disproportionately benefited a small fraction of the world's languages.

Figures & tables

Fig. 1 : An overview of the regions of Cameroon where the Basaà language is spoken, adapted from [ 7 ] .
 VowelShort FormMeaningLong FormMeaning
I/i/títo givetííto touch
U/u/húaudacityhúúto return
E/e/éto clear landpééviper
O/o/sóantelopesòòto hide
E/E/étreeEEto cry
A/a/bàto skinbààto filter
TABLE I : Short/long vowel oppositions in Basaà [ 5 ]
Fig. 2 : End-to-end ASR pipeline for the Basaà language — Fine-tuned XLS-R-300M.
EpochTrain LossVal. LossWER (Val.)
5001.30050.751693.77%
10000.51230.276960.90%
15000.35270.222252.34%
20000.27500.206950.12%
25000.22720.172743.29%
30000.19100.161740.54%
TABLE II : Evolution of WER on the validation set during training
 Time (sec.)WERReferencePredictedError type
13.90.200mudaa a ék ndap libiimudaa a yék ndap libiiChar. substitution (é →\to yé)
24.00.375u ntop bé nok maébau ntop bénok mayébaMissing space + vowel change
33.90.000men me nlok lipondo li malépmen me nlok lipondo li malépPerfect transcription
43.30.000me nnek bé nyeme nnek bé nyePerfect transcription
53.80.500bajo sañ ba mbébnabajosañ ba mbébnaMissing word boundary
63.00.000nyañ a yé malétnyañ a yé malétPerfect transcription
TABLE III : Sample transcriptions from the test set
Fig. 3 : Registration and login interface
Fig. 4 : Simple user interface
Fig. 5 : Contributor Interface
Fig. 6 : Validator Interface
Fig. 7 : Validator interface
More in this topic

Automatic Speech Recognition

Automatic speech recognition converts spoken language into text across speakers, languages, domains, and recording conditions.

See all 480 papers

Explore similar work

CardsList

  1. Automatic Speech Recognition for Low-Resource Sinhala: A Critical Review of Methods, Challenges, and Future Directions

    Oct 5, 2026Chanuka Dinuwan, Sanath Jayasena, Buddhika KarunarathneASR EvaluationAutomatic Speech Recognition

    MSc in Data Science and Artificial Intelligence Department of Computer Science & Engineering University of Moratuwa, Sri Lanka · University of Moratuwa, Sri Lanka

  2. LoASR-Bench: Evaluating Large Speech Language Models on Low-Resource Automatic Speech Recognition Across Language Families

    Mar 20, 2026Jianan Chen, Xiaoxue Gao, Tatsuya Kawahara +1ASR EvaluationLow-Resource Speech Recognition

    Kyoto University, Kyoto, Japan · Institute for Infocomm Research, Agency for Science, Technology and Research, Singapore

  3. Responsible ASR: Overcoming Challenges of Foundational Models in Narrow-Band and Low-Resource Settings

    Jun 17, 2026Tejas Godambe, Nutan Choudhary, Sanket Shah +2ASR EvaluationAutomatic Speech Recognition

    Applied AI, Krutrim, India