The rise of question answering (QA) research is mainly due to the demand of people that access to information quickly and accurately. Instead of returning “hits”, as information retrieval systems do, QA systems respond to natural language questions with concise, precise answer. In this work, we addressed on generating an exact answer in natural language for cooking QA system. We first reviewed the previous work of question analysis. Then, we presented the annotation scheme for knowledge database. Finally, we proposed the answer planning methodology for answer generation. The method mainly includes two steps: answer content planning and answer surface realization. An evaluation has been conducted on natural language questions and the results showed that the proposed answer generation method is effective and can satisfy user's demand.