关键词: 语义标注, 条件信息熵, 核函数, 支持向量机

Abstract: A novel method of software semantic annotation based on conditional information entropy reduction and multi-kernel SVM is proposed.Compared with traditional ontology semantic annotation，this method has some good theoretical characteristics，e.g. it achieves automatic software semantic annotation by using machine learning;positive and negative samples are balanced by resampling;conditional information entropy is used to reduct the sample demensions and a conversion method is given out to algebraic reduction;linear combination of several kernel functions considers both learning and generalization abilities.Annotation test results show that the CIEMK-SVM method has high classification accuracy，which proves its effectiveness and usefulness.

Key words: semantic annotation, conditional information entropy, kernel function, support vector machine