Food safety is an important part of protecting the lives of ordinary people. Food safety incidents have caused great concern around us. Due to lacking of technical assistance to monitor the food health hazards, universal testing cannot detect each kind of food, this paper studies and analyzes the complaint documents in the field of food. It extracts food hazards information from food complaint documents, for example, the extraction of food hazard information time, key words of food hazard information, and food hazard information center words. In the classification of food hazard information, a template for food hazards is established. By keyword matching and Hownet similarity calculation, the information center words with time labeling are extracted and classified into five different food hazard categories. It will provide support for subsequent research on food safety early warning mechanism.