Text Summarization for Big Data Analytics: A Comprehensive Review of GPT 2 and BERT Approaches

The goal of approaches to automatic text summarization is to construct summaries while extracting the essential information from one or more input texts. Large models could be trained thanks to the ability to examine text non-sequentially, which led to the Transformer becoming the most well-known NLP model. Big data and associated methodologies are frequently used to handle and alter these massive volumes of information. This chapter looks at large data methodologies and method such as Bidirectional Encoder Representations from Transformers (BERT) and Generative Pre-trained Transformer 2 (GPT 2) models for multi-document summarization. The Transformer, BERT and GPT and GPT 2 models in text summarization give very close results in terms of accuracy and they need to be compared to give a model that performs better. In this chapter, the two models have been compared and our results have shown that BERT performs better than GPT 2. This is found based on the results given by ROUGE metrics on a news article dataset containing 100 text files to summarize.