With the development of artificial intelligence, symptom recognition in the medical field has come into public view; to address issues such as the insufficient robustness of pure dictionary matching and the weak generalization ability of single rule-based systems, this paper designs and implements a multi-module symptom recognition system. By comparing the performance of pure dictionary matching, regular expressions, rule-based methods, and their combined approach, the paper verifies the effectiveness of a hierarchical collaborative strategy in symptom recognition tasks. Specifically, this method enhances the precise matching ability of pure dictionary matching through regular expressions, and improves the matching ability of pure dictionary matching for fuzzy problems through rule-based methods. The experiment adopts a comparative approach, conducting quantitative evaluations from four dimensions—accuracy, precision, recall and processing speed—and concludes with the advantages of the hybrid model.