WendkuuniArzouma/AI4Dev-Lab-Low-Resource-Language-Project

A brief presentation of the tools and dataset used in the experiments for the article "An explorative Investigation into Neural Machine Translation: the Case of Low-Resource Language Pairs in Burkina Faso". # AI4Dev-Lab Data and code release for Zakaria Sawadogo, Wendkuuni A. M. Christian Ouedraogo, Aminata Sabané, Idriss T. Tinto and Tegawendé F. Bissyandé, "An explorative Investigation into Neural Machine Translation: the Case of Low-Resource Language Pairs in Burkina Faso" # Dataset The French-Moorée Bible # The dataset presentation train_data represents the training data pre-processed (27,909 lines) valid_data represents the validation data pre-processed (3 000 lines) test_data represents the pre-processed test data (597 lines) bpe_data represents the vocabulary generated with the BPE used during the training phase with OpenNMT-py align_data represents the data alignment model generated with fast_align used during the training phase with OpenNMT-py # Tools used OpenNMT-py the framework for training and testing github.com subword-nmt for the BPE github.com fast_align for word alignment github.com