This paper describes a technique to optimize a similarity level of documents retrieved by using keyword selection in genetic algorithm (GA). an evaluation function for the fitness of each chromosome was selected based on dice's formulation, and to compare with the result of the research before (jaccard and cosine's formulation). by these formulations we can improve the performance of the system. to intialize a population, we need first to decide the number of genes for each individual and the total number of chromosomes (pop size) in the initial population. GA is bacically based on natural biological evaluation. we have implemented those techniques in a prototype of journal browser search (JBS). by the similarity level of documents, the user can retrieve the most similar document form a database.