A large number of licensee event reports are available in the nuclear power generation sector. A comprehensive analysis of the reports will provide valuable insights for improving nuclear power plant operation and safety. However, the free-text format of the reports poses great challenges to the analysis of the tens of thousands of reports generated. To address this issue, we propose an automated method for the analysis based on natural language processing techniques. Specifically, the objective is to automatically extract the causal relationships from free-text reports. The proposed method relies on a set of keywords that indicates causal relationships and the rules associated with the keywords for identifying the causal relationships, both of which can be identified based on manual analysis of sampled reports and sentences. The rules are described using the parts of speech of the words in a sentence and the dependencies between these words. The keywords and the rules constitute a rule-based expert system, Causal Relationship Identification (CaRI). The proposed method is applied to the analysis of the abstract section of the reports from the U.S. Nuclear Regulatory Commission Licensee Event Report database. We identified 11 keywords and developed 184 rules. The developed system, CaRI, is tested and the result shows that 86% of the causal relationships in the test data can be captured automatically. Application of the proposed method is foreseen in a number of areas, for instance, in the analysis of performance-shaping factors and in reconstruction of the scenario in an event.