The Phenomena Identification and Ranking Table (PIRT) serves as a fundamental tool in nuclear reactor safety analysis. It provides a systematic methodology for gathering information from experts on a specific concept, and ranking the importance of the information, in order to meet some decision-making objective. However, traditional PIRT methodologies are limited by expert subjectivity, which may lead to distorted assessment results. This paper presents an innovative PIRT methodology that incorporates text mining techniques, employing a data-driven approach to evaluate key phenomena with the aim of reducing expert subjective bias. This methodology leverages the advantages of text mining techniques through a three-step process: first, collecting and preprocessing large volumes of data; second, identifying key phenomena through feature extraction and cluster analysis; and finally, classifying and ranking these phenomena using keyword recognition, thereby overcoming the subjectivity and uncertainty inherent in traditional methods. By implementing a dual validation mechanism that combines text mining with expert assessment, this study establishes a more objective and reliable PIRT methodology, providing robust support for nuclear reactor safety analysis.