The development of a sepedi text generation model using long-short term memory
Mercy Moila, Thipe Isaiah Modipa · 2020
Recurrent Neural Network (RNN) is a type of deep neural network that is developed to generate a set of text given the features to produce the output based on features provided to the model. RNN is difficult to train when dealing with long memory dependencies. Long-short term memory (LSTM) is a recurrent neural network technique that works with memory cells which were developed to deal with the difficulties that were experienced by RNN. The system for word generation in text messages deduce the words based on statistical analysis from a large corpus. Moreover, the word in a sentence depends on its context which requires the use of machine learning algorithms to generate word sequences. The system to generate Sepedi text has not been developed especially using RNN techniques. The aim of the study is to use the LSTM RNN technique to generate Sepedi language text sequences. The data used for the experiments is a version of the NCHLT Sepedi text corpus. We obtained the accuracy of 50.3% with limited data and only 20 epochs.