JABU-2022/English_to_Kinyarwanda_Translation_with_RNNs

# English to Kinyarwanda Translation using RNNs This project aims to build a model that translates text from English to Kinyarwanda using Recurrent Neural Networks (RNNs). ## Dataset The dataset used for this project was sourced from publicly available resources: - Dataset 1 - Dataset 2 - Dataset 3 ## Preprocessing - Lowercasing - Removing Non-Alphanumeric Characters - Adding Special Tokens ( , ) - Tokenization - Padding ## Model Architecture The model uses a Bidirectional GRU for the encoder and a GRU for the decoder, with embedding layers for converting integer sequences to dense vectors. ### Encoder from tensorflow.keras.layers import Embedding, GRU, Bidirectional from tensorflow.keras.models import Model import tensorflow as tf class Encoder(Model): def __init__(self, vocab_size, embedding_dim, enc_units, batch_sz): super(Encoder, self).__init__() self.batch_sz = batch_sz self.enc_units = enc_units self.embedding = Embedding(vocab_size, embedding_dim) self.gru = Bidirectional(GRU(self.enc_units, return_sequences=True, return_state=True, recurrent_initializer='glorot_uniform')) def call(self, x, hidden): x = self.embedding(x) output, forward_h, backward_h = self.gru(x, initial_state=hidden) state = tf.concat([forward_h, backward_h], axis=-1) return output, state def initialize_hidden_state(self): return [tf.zeros((self.batch_sz, self.enc_units)) for _ in range(2)] ### Decoder class Decoder(Model): def __init__(self, vocab_size, embedding_dim, dec_units, batch_sz): super(Decoder, self).__init__() self.batch_sz = batch_sz self.dec_units = dec_units self.embedding = Embedding(vocab_size, embedding_dim) self.gru = GRU(self.dec_units, return_sequences=True, return_state=True, recurrent_initializer='glorot_uniform') self.fc = Dense(vocab_size) def call(self, x, hidden, enc_output): x = self.embedding(x) output, state = self.gru(x, initial_state=hidden) output = self.fc(output) return output, state ## Training Process and Hyperparameters ### Training Step …