Massive volume of data is generated in the healthcare industry that includes numerous unstructured information and texts about the hospitals, patients, diseases, medical equipment, drugs, etc. Deployment of extensive data mining and data treatments will refine the data and provide more actionable insights than what is observed in the raw medical data and documents. This paper proposes a methodology of handling such medical records as a semi-supervised learning problem to formulate an assistive technology for predicting the topographic classes (Digestive system, Cardiovascular, Nervous, General Pathological Conditions, Neoplasms) of the diseases discussed in the records by leveraging natural language processing techniques. A topographic class classifier model has been built that predicts the class based on the contents described in the text. Models such as Support Vector Machine, Random Forest, XGBoost, and a few LLMs were tested for this purpose. Unified Medical Language Systems (UMLS) is a repository of medical terms and vocabulary that helps in tagging the type of a text based on the predefined classes. Using UMLS, the text is analyzed for identifying underlying diseases, symptoms, treatment, and drug-related information and tagged. This forms the base data to perform rule mining on the text to identify the core connections between diseases, symptoms, drug, and so on to find their co-occurrence. The reasoning behind this analysis is to uncover the hidden co-occurrences of diseases, symptoms, treatments that otherwise go unnoticed. This will provide more insights about drug behaviors, adverse events, disease-drug reactions, and so on that would help in refining the clinical trials.