The Machine Learning (ML) field has gained its momentum in almost any domain of research and just recently has become a reliable tool in the medical domain. Extracting medical relations is very trivial task since the medical information is stored in textual format and the database of medical information is also very large in size for example Medline is the medical database that contains 21 million records from 5000 selected publications. In addition to that web page containing medical information also contains some unrelated contents like advertisements, scroll bars, quick links, related searches etc., manually extracting only relevant information from such a huge database is very difficult task. To reduce user overhead of extracting useful information current approach is proposed. This approach presents the efficient machine learning algorithm and techniques used in extracting disease symptom and treatment related sentences from Medline. In this approach Multinomial Naive Bayes algorithm and several other techniques are used to extract semantic relation between disease symptom and their associated treatment. The proposed system gives the user exactly the Disease Symptom and Treatment related sentences by avoiding unnecessary information and this technique can be integrated with any medical management system to make better medical decisions.