Abstract: When using genetic algorithms to solve optimization problems in semantic-based applications, we find that these methods cannot interpret semantic relations and hence overlook useful information in evolution. Therefore, genetic algorithms are insufficient to satisfy the requirements in this case. We propose to use formal semantics of ontology to improve genetic algorithm in several aspects and make it more adaptive to solve semantic-based problems. In this study, we present a semantic-based genetic algorithm to incorporate domain knowledge into the algorithm and perform evolution based on the ontology semantics. The advantages of the algorithm include expressing semantic information in chromosome representation and preserving the information by applying genetic operators in evolution. We illustrate the usage of the algorithm by applying it to solve the problem of sub-ontology evolution. Our experiments with a large-scale traditional Chinese medicine ontology as the benchmark demonstrate the feasibility of the algorithm in solving semantic-based problems.