Bidirectional Machine Translation Between Amharic and Chaha: Enhancing Performance Using Deep Learning

Machine translation for low-resource languages presents significant challenges due to the limited availability of parallel corpora and linguistic resources. This study focuses on bidirectional machine translation between Amharic, a widely spoken language in Ethiopia, and Chaha, a language from the Gurage family, using deep learning techniques to improve translation performance. A dataset of 6,200 parallel sentences was employed for model training and evaluation. Despite this relatively small corpus, various approaches were explored to enhance translation quality. The research addresses the lack of machine translation resources for Chaha, a low-resource language in the Sebat Bet Gurage dialect group of Ethiopia. By leveraging an encoder-decoder architecture with Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), and Gated Recurrent Unit (GRU) models, the study aims to create an efficient translation system that works in both directions—Amharic to Chaha and Chaha to Amharic. A parallel corpus of 6240 Amharic-Chaha sentence pairs was compiled from religious texts and publicly available datasets. The model's performance was evaluated using the BLEU score, with the Bi-LSTM model achieving the highest accuracy. This system demonstrates the feasibility of machine translation for under-resourced languages and contributes to the digital presence of the Chaha language, helping preserve and widen its use. The results show promise in improving communication between speakers of Amharic and Chaha while highlighting areas for future research, such as expanding the dataset and integrating morphological analysis. This research lays a foundation for further developments in bidirectional machine translation systems that can translate both Chaha to Amharic and Amharic to Chaha. This would enhance mutual understanding and preserve the linguistic nuances of both languages addressing these gaps, we develop Bidirectional Machine Translation Between Amharic and Chaha.