NLP Text Generation with LSTM (RNN)

Enhanced a text generation model by introducing a regularized LSTM architecture on the Human Numbers dataset, which comprises the first 10,000 numbers written in English. The primary objective of the project was to predict each word based on the previous three words in the sequence. This optimization resulted in a substantial accuracy improvement from an initial 50%, achieved with an RNN model created from scratch using PyTorch, to an impressive 86%.