zeyneb-chiha/NMT_Tunisian-Dialect

# NMT_Tunisian-Dialect ### About the Project### * The Tunisian dialect written in Latin is different from the standard languages. * so Using the Natural Language Process (NLP) and Deep learning we gonna create a model that can understand this dialect * and translate it to standard Arabic language and others like French and English. ### DATASET ### * The dataset will be self-created with help of the bilingual selected sentence pairs from * the Tatoeba Project “English-Arabic”, English sentences will be replacedby Tunisian dialect sentences. ### DATASET PREPROCESSING ### * Cleanup data * Convert to lowercase * Exclude punctuation * Generate synthetic data * Feature Engineering(create feature dictionories for input and target data) * Convert sentences to one-hot vectors ### Model ### * Seq2seq (Encoder-Decoder) model impelementation with bahadnau Attention Mechanism