Demozrahwa/Developing-a-Tigrinya-English-Translation-System-Enhancing-Accessibility-for-Underrepresented-Lang

# Tigrinya-English Translation System ## Project Overview This project focuses on developing a machine translation system between Tigrinya and English. Tigrinya, a low-resource language spoken in Eritrea and Ethiopia, has limited digital tools for linguistic support. The project leverages transformer-based models such as Helsinki-NLP and M2M100, fine-tuned on a custom dataset of over 350,000 bilingual sentence pairs. --- ## Features - **Tigrinya-to-English Translation:** Translate Tigrinya sentences into English accurately. - **English-to-Tigrinya Translation:** Translate English sentences into Tigrinya. - **Custom Fine-Tuning:** Fine-tuned pre-trained models to improve translation quality using a diverse dataset. --- ## File Structure ```plaintext Capstone/ │ ├── Dataset_csv/ │ ├── en_to_ti_sampled_test.csv # Sampled English-to-Tigrinya test dataset │ ├── en_to_ti_sampled_train.csv # Sampled English-to-Tigrinya training dataset │ ├── ti_to_en_sampled_train.csv # Sampled Tigrinya-to-English training dataset │ ├── ti_to_en_sampled_val.csv # Sampled Tigrinya-to-English validation dataset │ ├── en_to_ti_val.csv # Full English-to-Tigrinya validation dataset │ ├── ti_to_en_val.csv # Full Tigrinya-to-English-to-validation dataset │ ├── ti_to_en_test.csv # Full Tigrinya-to-English test dataset │ ├── en_to_ti_test.csv # Full English-to-Tigrinya test dataset │ ├── en_to_ti_train.csv # Full (English-to-Tigrinya) training dataset │ ├── ti_to_en_train.csv # Full (Tigrinya-to-English) training dataset │ ├── models/ │ ├── m2m100_model.ipynb # M2M100 model notebook │ ├── En_Ti_Helsinki-NLP_model.ipynb # Helsinki-NLP model notebook (English to Tigrinya) │ ├── Ti_En_Helsinki-NLP_model.ipynb # Helsinki-NLP model notebook (Tigrinya to English) | |── T5_model.ipynb # T5_BERT …