In this project I built a recurrent neural network model in Python using TensorFlow/Keras to work with a dataset of phrases to test if it can remember long phrases. This notebook goes over the fundamenals of how Recurrent Neural Networks work, and how Recurrent Neural Networks can handle sequential data.
Our dataset consisted of an array of phrases, which served as input for training our RNN model. However, the dataset required preprocessing and modifications to make it suitable for training.
Our model achieved an accuracy of 98.75% in predicting the next set of characters to generate, demonstrating its ability to fully generate complete text from a given starting character, word, or phrase.
Our model was capable of generating the full phrase "pennies are shiny" just from the word "pen".
Feel free to download the notebook if you want to explore more about RNN models and their ability to handle sequential data.