Bemba Speech Translation: Exploring a Low-Resource African Language

2025/05/05 by Farouq, Muhammad Hazim Al, Wassie, Aman Kassahun, Moslem, Yasmin
#Audio and Speech Processing (eess.AS) #Computation and Language (cs.CL) #FOS: Computer and information sciences #FOS: Electrical engineering #Sound (cs.SD) #electronic engineering #information engineering

paper · doi:10.48550/arxiv.2505.02518

Abstract

This paper describes our system submission to the International Conference on Spoken Language Translation (IWSLT 2025), low-resource languages track, namely for Bemba-to-English speech translation. We built cascaded speech translation systems based on Whisper and NLLB-200, and employed data augmentation techniques, such as back-translation. We investigate the effect of using synthetic data and discuss our experimental setup.

Citations

• SpeechT: Findings of the First Mentorship in Speech Translation
• Findings of the IWSLT 2024 Evaluation Campaign
• Leveraging Synthetic Audio Data for End-to-End Low-Resource Speech Translation
• SEACrowd: A Multilingual Multimodal Data Hub and Benchmark Suite for Southeast Asian Languages
• AfriMTE and AfriCOMET: Enhancing COMET to Embrace Under-resourced African Languages
• BIG-C: a Multimodal Multi-Purpose Dataset for Bemba
• Robust Speech Recognition via Large-Scale Weak Supervision
• No Language Left Behind: Scaling Human-Centered Machine Translation
• Survey of Low-Resource Machine Translation
• The FLORES-101 Evaluation Benchmark for Low-Resource and Multilingual Machine Translation
• BembaSpeech: A Speech Recognition Corpus for the Bemba Language
• The Tatoeba Translation Challenge -- Realistic Data Sets for Low\n Resource and Multilingual MT
• Optuna: A Next-generation Hyperparameter Optimization Framework
• Adaptation of Machine Translation Models with Back-translated Data using Transductive Data Selection Methods
• Improving Neural Machine Translation Models with Monolingual Data

Related

• BembaSpeech: A Speech Recognition Corpus for the Bemba Language
• IMS' Systems for the IWSLT 2021 Low-Resource Speech Translation Task
• University of Cape Town's WMT22 System: Multilingual Machine Translation for Southern African Languages
• BIG-C: a Multimodal Multi-Purpose Dataset for Bemba
• Automatic Speech Recognition for African Low-Resource Languages: Challenges and Future Directions