Deep Learning has been the go-to tool for text summarization in the recent times.Traditional deep learning research focuses on performing abstractive text summarizationwithout considering the user’s interests to personalize the summaries.This problem motivated us to develop a deep learning based text summarizationsystem which can curate personalized summaries. In this work, we propose an LSTMbased Bi-Directional Recurrent Neural Network model to perform extractive textsummarization. Our new deep learning approach focuses on personalizing the extractivesummaries based on user’s interests to make the summaries more intriguing tothe user. We performed the experiments on CNN and Daily Mail news dataset. Wealso have experimented with a new set of semantic word vectors called ConceptnetNumberbatch. Out of domain evaluation was done on the Signal-Media one millionnews articles dataset. Experimental results on the two summarization datasetsdemonstrate that our models obtain results comparable to the state of the art. Thepersonalization framework curates interesting summaries based on user’s interestswhile retaining the important information from the source document.
