WeleTeklay/Tigrinya-ASR-Research

Research on low-resource Tigrinya ASR systems conducted during ML research internship at Lesan.AI # Tigrinya ASR Research Research on low-resource Tigrinya Automatic Speech Recognition (ASR) systems conducted during my paid Machine Learning Research Internship at Lesan.AI, a German based AI company specifically working on African languages. ## Research Overview This repository contains research materials, experiments, reports, and presentation resources related to the evaluation of low-resource ASR systems for Tigrinya. ## Research Contributions - Curated and processed 10,000+ Tigrinya speech utterances - Conducted ASR benchmarking and evaluation - Performed linguistic and acoustic error analysis - Worked with Whisper-based ASR systems - Contributed to establishing baseline ASR benchmarks for Tigrinya ## Conference Presentation This work was presented at: **International Conference on Artificial Intelligence for Sustainable Development (AI4SD 2026)** Mekelle University, Ethiopia April 27–28, 2026 The camera-ready manuscript is under preparation for publication. ## Repository Structure ```text docs/ -> Reports and certificates slides/ -> Presentation slides notebooks/ -> Research notebooks and experiments scripts/ -> Supporting Python scripts ``` ## Authors Weldesemayat Teklay Gebre Gebregergs Mekonen - LinkedIn: linkedin.com - GitHub: github.com