Although search engines are good at short query search, they are currently not effective for long query search or natural language question answering. Automatic question answering systems (AQAS) have been developed to solve this problem in recent years. However, general AQAS still suffer from a few problems. For example, laypersons usually use non-professional terms to express medical questions, and thus this will make general AQAS extract inappropriate answers. Besides, according to our observation, we found that most of medical questions contain not only medical terms but also question intention. Generally, AQAS is composed of three modules: question analysis, document retrieval and answer extraction. In this paper, we particularly focus on dealing with the two parts question analysis and answer extraction. Our main contributions are to propose the Question Intention Model (QIM) for question analysis and Entropy-based Paragraph Extraction Model (EPEM) for answer extraction.