Text Summarization Quality Detection Based on GPT-3

Abstract

To summarize lengthy text in a concise and accurate manner is called text summarization. With the increasing amount of information available, text summarization has become increasingly important in information retrieval, knowledge management, and sentiment analysis. The research on text summarization dates to the 1960s, and the methods have evolved from traditional template-based generation to statistical and neural network-based methods. Modern language model GPT-3 has demonstrated outstanding linguistic ability, including the ability to produce text that is cohesive and grammatically correct. However, evaluating the caliber of text produced by GPT-3 is challenging and requires careful evaluation criteria. This study evaluated the text summarization ability of GPT-3 using multiple evaluation models (ROUGE, BLEU, and CIDER), and found that the generated summaries exhibited high quality and accuracy.