The domain knowledge of forest pests and diseases refers to many different disciplines with large numbers of knowledge .It is a significant and challenging project to share and reuse this knowledge and to retrieve what you need intelligently and efficiently. In this paper, ontology as the modeling approach is introduced into the field of forest protection .Through studying the characteristics of the knowledge of forest pests and diseases, this paper constructs a domain ontology to obtain concept semantic tree as the foundation of computing semantic similarity. The novel approach computes semantic similarity according to hierarchical structures and inheritance relationships in the concept semantic tree. The automatic question answering system from this model is able to accept natural language description, and return relevant answers automatically after word segmentation, semantic similarity computing and statements similarity computing. Experimental results show high accuracy and integrality achieved by using this retrieval system.