Used Python “keras” package to develop a unidirectional LSTM based deep learning model “encoder-decoder networks” for title generation on news data set. Improved the architecture by using attention. Created tensorboard visualizations to show the validation accuracy and loss. Evaluated the summaries using the ROUGE score.