Ramalheira07/kmbPtMT

Carefully constructed and treated parallel dataset can contribute to improving the performance of multilingual machine translation models in the Kimbundu–Portuguese language pair. # Angolan Low-Resource Language Translation Project This repository contains resources and scripts for our research on low-resource language translation, focusing on Kimbundu to Portuguese. All datasets and models are hosted on Hugging Face, and scripts are provided for Google Colab to facilitate reproducibility. ## 📂 Repository Structure ## 📝 Datasets The datasets used in this project were manually curated due to the lack of existing Kimbundu-Portuguese parallel corpora. The datasets include: - Cleaned and preprocessed parallel corpora - Metadata describing the source and licensing All datasets are publicly available on Hugging Face: Datasets on Hugging Face Due to the scarcity of digital data in the Kimbundu language, this work used diverse sources for the purposes of cultural preservation and non-profit linguistic research. The dataset is made available under a scientific use license, respecting the integrity of the original sources through data fragmentation. ## 🧠 Pre-trained Models We provide pre-trained models for Kimbundu to Portuguese translation. These models were trained and fine-tuned on the curated datasets. You can find the models here: Models on Hugging Face ## 💻 Scripts All scripts for training, fine-tuning, and evaluation are available as Google Colab notebooks for easy execution: - `traine_eval_kmbpt.ipynb` — Training and fine-tuning scripts; Evaluation and testing scripts. ## 🚀 Usage 1. Clone this repository: ```bash git clone github.com 2.Open the desired Colab notebook in Google Colab. 3.Follow the instructions in the notebook to download datasets, train models, and perform evaluations. If you use these resources in your research, please cite: ## License This project and datasets are licensed under CC-BY 4.0. Purpose: This dataset was compiled exclusively for the purposes of linguistic preservation, scientific research and development of Natural Language Processing (NLP) technologies for t …