A Hybrid Model for Text Summarization Using Natural Language Processing

Abstract

Text summarization plays an important role in the area of natural language processing. The need for information all over the world to solve specific problems keeps on increasing daily. This poses a greater challenge as data stored on the internet has gradually increased exponentially over time. Finding out the relevant data and manually summarizing it in a short time is a challenging and tedious task for a human being. Text Summarization aims to compress the source text into a more concise form while preserving its overall meaning. Two major categories of text summarization methods exist namely: extractive and abstractive. The extractive technique concentrates on determining key themes using frequency analysis of sentences in the corpus of the text. Abstractive methods write a new summary with newly generated texts which do not appear in the corpus itself. This paper presents a hybrid model for text summarization using both extractive and abstractive techniques. Term Frequency (TF) – Inverse Document Frequency (IDF) was used for extractive and T5 Transformers for abstractive summarization. Iterative Incremental Methodology was adopted in the study. The hybrid model emerged as not the best choice compared to the extractive and abstractive as it had been hypothesized in the study when the accuracy and execution time of the summary generated was considered.