The invention provides an optimization method and system for a knowledge base of a man-machine interactive system. The method comprises the steps of obtaining multiple pieces of corresponding candidate answer information according to question information input by a user, and forming a plurality of first result question vectors by the candidate answer information and the question information input by the user; judging whether checked data information matched with the question information input by the user exists in the knowledge base or not; if yes, forming a second result question vector by the checked data information, matched with the question information input by the user, in the knowledge base; and calculating the similarity between each first result question vector and the second result question vector, and if the similarity is greater than a first preset threshold, adding the corresponding candidate answer information and the question information input by the user into the knowledge base. According to the method and the system, the data in the knowledge base is automatically optimized, so that the knowledge base optimization efficiency is improved and the knowledge base optimization time is shortened.