A Hybrid Extractive-Abstractive Framework with Pre & Post-Processing Techniques To Enhance Text Summarization
Rohan Habu, Rohit Ratnaparkhi, Anjali Askhedkar, Sunita Kulkarni · 2023
The goal of this paper is to enhance text summarization using a hybrid methodology. The process of producing a condensed version of a text while keeping its essential details is known as text summarization. In this study, we have presented a method for training the T5 model on the SAMSum dataset with conversation sentences to increase its effectiveness in text summarization. To determine the impact of training the T5 model on the dataset, the model is assessed using the ROUGE metric both before and after training. ROUGE is a set of metrics used to evaluate the quality of automatic summaries by comparing them to reference summaries based on the overlap of n-grams, word sequences, and other linguistic units. In order to enhance the quality of the generated summary, our hybrid approach makes use of extractive and abstractive summarization techniques as well as pre-and post-processing techniques. There is an improvement in the ROUGE metrics of the model before and after training. ROUGE1 before training was observed to be 25.53 while ROUGE1 after training is calculated to be 45.17.