Abstract: The paper proposes a new semantic similarity calculation method which is based on tree structure.It combines factors that affect semantic similarity such as the semantic relation between concept nodes, semantic distance, the depth of concept nodes and node density.It uses the tree hierarchy to express the semantic information between concept nodes, and improve the traditional method to compute the density of concept nodes, also joined the adjustable parameters to adapt to different scenarios.At last it is verified this algorithm has a strong advantage by experiments.