Researchers from San Diego State University, Virginia Tech and Loyola Marymount University have found a way to use artificial intelligence (AI) to enable consumers to monitor food safety before they buy a product.

According to the CDC, 80% of food poisoning cases are of unknown origin. The team’s new food safety monitoring system (FSMS) utilizes customer comments on websites to identify products associated with food-related illnesses.

To create the system, the team used an AI technology called text mining to analyze comments and reviews on two sites, Amazon.com and iwaspoisioned.com. Amazon is a major retailer for many items, including food, and iwaspoisioned.com is a site where customers alert others to cases of food poisoning.

Source: Pixabay

To test their text-mining algorithm, the team used two datasets. The first was 11,190 randomly selected Amazon reviews of grocery items purchased between 2000 to 2018. The second was 8,596 reviews of food products posted in iwaspoisioned.com. The computers were programmed to recognize words associated with foodborne illnesses, including “sick”, “vomiting”, “fever” and “nausea”. Using this information, the computer found a list of flagged products, from protein bars and powder to herbal teas. Two of the items flagged had already been recalled.

The final step was for a panel of 21 experts to manually review the computer’s findings. They verified the risk level of a product and suggest remediation strategies for the manufacturers.

In the future, the team hopes to create a way to alert consumers to food product risks while they are shopping online. This would make it easier for consumers to avoid products that may make them sick.

A paper on this research was published in Risk Analysis.