An Intelligent Question-Answering System for Agricultural Pests and Diseases Based on DeepSeek and Knowledge Graph

Agricultural pests and diseases severely affect the yield and quality of crops. Existing pest and disease data exhibit issues such as multi-source heterogeneity, dispersion, and chaos, making it difficult to acquire knowledge. To address these problems, this study constructs an intelligent question-answering system for agricultural pests and diseases based on DeepSeek and knowledge graph technology. The system obtains data through web crawlers, constructs a domain-specific data subset, and utilizes named entity recognition and relation extraction techniques to build a knowledge graph. Models such as PNAttention + BERT + CRF are employed to enhance the performance of entity and relation extraction. Moreover, it achieves multimodal question-answering through YOLOv8 and DeepSeek, supporting natural language queries. After testing, the system performs well in entity recognition and relation extraction, and all functional modules operate stably.