In the online environment, public sentiment events on food safety are characterized by their rapid disseminations and high impacts. If public sentiment issues were not taken seriously and addressed in a timely manner, they are highly likely to lead to sudden mass incidents. Constructing a robust online public sentiment management index system for food safety helps identify, evaluate, predict and respond to public opinion. This paper firstly explained the mechanism of food safety online public sentiment dissemination. Secondly, the technical architecture of food safety online public sentiment analysis system based on big data was used to build a food safety online public sentiment index system and establishes a food safety online public sentiment early warning model. Finally, solutions to early warning of online public sentiment on the topic of food safety in the era of big data from the government, media and enterprises aspects were proposed. The use of natural language processing and information retrieval (NLPIR) and other technical means to collect, understand, analyze, extract and ultimately dive deeper into information on public sentiment on the internet was explained, so that public sentiment could be detected, warned and dealt with early, and public concerns could be addressed. With the public's preferences for expressing emotions and making personal statements via the internet, the government, the media, the public and companies should give priority to early warnings of public sentiment on food safety. How to detect, warn, handle and mitigate online public sentiment events early is of great significance to the government and relevant departments.