A COMPARATIVE EVALUATION OF CHATGPT AND DEEPSEEK IN AUTOMATIC TEXT SUMMARIZATION
Loading...
Date
2026
Journal Title
Journal ISSN
Volume Title
Publisher
DEPARTMENT OF ENGLISH AND APPLIED LINGUISTICS
Abstract
This study compared the AI-generated summaries of five selected short stories: The Last Leaf by O Henry(1907), The Tell-Tale Heart by Edgar Allan Poe(1843), Dead Men‘s Path by Chinua Achebe(1953), A Sorrowful Woman by Gail Godwin(1971) and After Twenty Years by O Henry(1906). The study focused on summaries, generated by ChatGPT and DeepSeek, evaluating them based on qualitative criteria of coherence, cohesion and informativeness, supported by quantitative measures of compression and readability. Previous studies examined AI-based text summarization using various evaluation metrics; however, limited research had comparatively analyzed ChatGPT and DeepSeek on texts using the textuality standards of coherence, cohesion and informativeness.The research was based on Beaugrande and Dressler‘s Theory of Textuality (1981). Out of the seven standards of text, Coherence, Cohesion, Informativity, Intertextuality, Intentionality, Acceptability and Situationality, only three were selected for evaluation: coherence, cohesion and informativeness. Through a qualitative comparative research approach, the study examined how well each AI model structured information and maintained the original texts' meanings. In addition, quantitative measures of compression ratios and standardized readability scores were calculated for each summary to complement the qualitative discourse analysis. The study concluded that ChatGPT outperformed DeepSeek making it a more effective and reliable tool for generating coherent, cohesive, and informative summaries of texts, therefore, contributing to the growing discussion on the use of artificial intelligence in language-related tasks.