Semantic similarity is the essential part of automatic question answering system, and the computation of semantic similarity based on ontology can provide more various semantic information compared with the traditional method. In this paper, we firstly propose a semantic similarity algorithm used in the sentence similarity computation, combining relations of semantics, hierarchical structure and depth. Next, based on conventional sentence similarity algorithm, we develop a weighted sentence similarity algorithm. Finally, a question answering system is implemented with the two proposed algorithms. The experimental results show that our proposed method can achieve higher accuracy than others.