Text Summarization Using Deep Learning: An Empirical Analysis of Various Algorithms

Text Summarization is rephrasing the text into a shorter, concise form while preserving its original meaning. Various researchers have worked on this domain using deep learning techniques, but still there is a scope to produce a concise and meaningful summary. In this work, we perform empirical study of neural models like seq2seq and Transformers. The focus is on architectural advancements and their impact on the summarization. Also, we discuss the introduction of the Attention module and its impact on the performance of the algorithm. Through these experiments, a comparison is drawn among various models. A reference sentence is taken and various algorithms are applied on the sentence. The dataset used to train the model is amazon food review. The generated summary using each algorithm is evaluated on ROUGE score and conclusion is drawn based on results obtained. Although Transformer outperforms seq2seq models however, seq2seq can be used for certain tasks. Also, the significance of attention module on summary can be visulaized through results.