Abstract
Text summarization is a process of getting useful and relevant information from the original text. There are two types of text summarization: Extractive summarization, which extracts some important sentences from the original text and the other is Abstractive summarization which requires a sophisticated intermediate representation of text. In this research, a hybrid approach is proposed for text summarization. Several challenges are associated with text summarization systems in educational institutions. Firstly, there is an under-representation of the research performed in this field. Secondly, most of the models that are used for text summarization have shown results with very low accuracy. These include either extractive techniques or transformer models. An efficient deep-learning technique is needed to predict the summary of educational books with enhanced accuracy. In the current study, we used the T5(Text-To-Text Transfer Transformer) and BART(Bidirectional and Auto-Regressive Transformers) models to predict the summary of educational books. To enhance the performance of these techniques we used extractive and abstractive models. The main purpose of this research is to fine-tune the T5 and Bart models that are used for text summarization. The dataset to be used for the implementation of hybrid models contains the topics from educational books. The results are evaluated through the well-known ROUGE (Recall-Oriented Understudy for Gisting Evaluation) metric. Our best model achieved R-1 score of 0.73, R-2 score of 0.63 and R-L score of 0.42, as the F score calculated by ROUGE is evident to be the best as compared to the already developed techniques shown in the literature review. The significance of this study is that it can turn large amounts of data into easy-to-use summary information, and it can aid students at higher institutions in comprehending complex concepts by providing concise and accessible explanations. Similarly, teachers can take advantage of summarized material by making notes from different sources for students or taking important highlights, which is helpful for students.
Replies