MKL,an algorithm for meta-knowledge learning,is presented in NDKAS,which is a knowledge acquisition system.On the basis of classification and abstraction, it can induces meta-knowledge which is used to effectively organize rules in knowledge base in binary-tree structure,which satisfies the essential properties of meta-knowledge.The structural representation of meta-knowledge,the algorithm description of MKL and the proof of satisfaction of the essential properties are given in the paper.An example of MKL application is also described.