Abstract: Today, with the development of database systems and the data stored in these systems, there is a need for an instrument that the stored data to be processed so information obtained from this process will be available to users. Using simple questions in databases and various tools for regular reports, Information can be made available to users to make conclusions about the data and the logical relationships among them, But when the volume of data is high, not matter how professional and experienced users are, they Cannot recognize useful patterns among large amount of data Or even if they are able to do it The cost of operation is very high in terms of human and financial resources. So for now there is a paradigm shift from modeling and classical analysis based on basic principles to Ongoing models and directly data analysis. Data mining is one of the most important methods which useful patterns in data with a minimum user intervention would known.in data mining the Information is available to users and analysts so According to it, Critical decisions would be made in organizations. BM25 algorithm is considered as a successful ranking algorithm. The use of this algorithm in Persian literature and also in the text that contains a combination of Persian and Latin , caused reduction in the precision, accuracy and the readout of this algorithm. IN this study we offer a way to fix the problem and the results will be evaluated.