Medical diagnosis is an important domain of research which aids to identify the occurrence of a disease. The paper proposes a novel glide path to knowledge discovery in medical systems by acquiring relevant information from the data set. This approach makes diagnosis easier. Using naive bayes, the overall speed and accuracy of the algorithm increased and extract high quality data set from an unstructured text. The primary advantage of the scheme is that it can be used to whatever sort of dataset whether it is a predefined dataset or not.