Generation of answer in QA system isone of the most challenging research areas in natural language and performing in the medical domain is more difficult. The main reason is: patient has faith in doctor’s information but may doubt on the machine. Also, accuracy of the system is restricted in natural language. The computer may find the information of a disease but not as efficient as human. In this paper, various well known question answering systems are discussed and analyzed with their benefits and limitations. Finally, in this review, a new approach is visualized to overcome the limitations of present question answering system.