Question answering system in the weapon field not only enables users to obtain information on weapons quickly and accurately, but also provides smarter question answering. With military weapons as the research direction, an SVM question classification method based on Chinese character algorithm is proposed, and a question answering system over knowledge graph of weapons is established. The domain word segmentation is used in this system to analyze user questions, extract question features for classification, intelligently generate SPARQL, query RDF to get answers, and return them to the user. Experiments show that the intelligent question answering system with the multi-question classification is effective and highly accurate in answering questions in the field of weapon.