Natural language processing in food safety research: A systematic review

Natural language processing (NLP) is a subfield of artificial intelligence that uses and processes textual data. Especially thanks to the recent developments, this subfield holds promise as a valuable toolkit for food safety research and policy. However, there is no recent review of the studies at the intersection of natural language processing and food safety. This review aspires to be a reference source for researchers who wish to leverage natural language processing for food safety research. First, we introduce different modelling approaches in natural language processing. Then, we provide a taxonomy of the studies with respect to different dimensions. We also describe the labelled datasets for the studies. Finally, we set a future agenda for research. We examined 102 studies published between 2010–2024. The most common purpose expressed in the studies was assisting authorities in general monitoring of hazards. Text classification was the most frequently applied natural language processing task. Social media was the most prominent data source. Limited sharing of the labelled datasets was considered as a significant challenge of this line of research. Potential future work ranges from prioritization of hazards to monitoring of social media for dangerous food “challenges”.