Text Summarization of Articles Using LSTM and Attention-Based LSTM

This research work focuses on developing a model for performing text summarization of articles, depending on the novelty of the information that is arranged in the final summary, where summarization is roughly categorized into two types: extractive and abstractive summarization. In today's fast-paced society, most people disregard if they are deemed to be rather extensive due to a lack of time. Therefore, having a brief overview of this article would not only encourage readers to read it, but if the topic is of interest, it would but also allow them to go deeper into it. As a result, this research work intends to bridge this research gap with the proposed model. Initially, this research project developed a reference RNN model and tested its accuracy. It was discovered that the summary loses a lot of real terms and reduces and confuses the overall meaning, therefore it was decided to build an LSTM model for the same and test it with three distinct datasets. The first experiment is carried out with a small dataset, whereas the second experiment is carried out with a comparatively bigger dataset by applying LSTM to it. We decided to carry it forward and improve the model by increasing the size of our dataset and choosing a large dataset with an attention layer to the proposed LSTM model by comparing the accuracy of all the models with the existing model, and it will be evaluated by using ROUGE metrics and proving the proposed solution to the world. Furthermore, the future research directions will be provided in Future Scope regarding where and how the proposed model can achieve more efficiency in the future.