To support users to quickly access information they need from the agricultural library’s vast information and to improve the low intelligence query service, a model for intelligent library information retrieval was constructed. The semantic web mode was introduced and the information retrieval framework was designed. The model structure consisted of three parts: Information data integration, user interface and information retrieval match. The key method supporting retrieval was designed. The traditional semantic similarity algorithm was improved according to its shortages. An algorithm based on semantic distance was designed and tested. The results can improve the recall ratio and precision of information retrieval, improving information retrieval performance.