Abstarct: Creating a system that can effectively determine the distance between two exxpressions has been a concern of researchers in artificial intelligence and data retrieval. The similarity measure of the two terms are used in a wide range of applications, such as natural language processing, query correction, semantic spelling, plagiarism detection, document comparisons and other areas of data retrieval. The input of detecting systems can be a text or a set of documents and texts, and the output is the result of a system's judgment about the similarity of the input sentences. Ultimately, the proximity of the judgment of the system to human judgment is indicative of the good functioning of the system. In this paper, we propose a method for calculating the uncertainty-baxsed lexical similarity. To achieve this purpose, we use fingerprinting-baxsed algorithms and Winnowing techniques as similarity calculation measures. Also, by considering several parameters to calculate the number of phrases, we use the fuzzy inference system to decide on uncertainty conditions. In this method, the characteristics are baxsed on the technique of hashing and general transformation of sentences that these criteria have a good speed and accuracy. After calculating the similarity of the two inputs, similar sentences are extracted by the system. The method has been evaluated on the PAN databaxse and the Shahrood University of Technology Similarity assessment, which with a precision of 78%, separates the documents into three different, relatively similar and similar classes.