For the last few years, Amazon has been cracking down on people who are paid money for posting positive reviews on items. However, for some people, negative reviews on Amazon have proven to be far more interesting than the positive ones.
Don't Miss: Win a FREE Nintendo Switch in our Giveaway
The researchers, who hailed from the University of Washington's Data Science for Social Good program have looked toward Amazon food reviews to predict recalls. A machine learning platform was developed so that they could mine the text of Amazon reviews so that they could make safety predictions. The program is still in progress, it has the potential to save people from the headache of food recalls and food poisoning.
Elaine Nsoesie, the lead professor on the project at UW's Institute for Health Metrics and Evaluation said that the process will be a slow one, according to Geekwire.
“Some of [the cases] were as long as one year, from the time we saw someone write a review saying ‘there is something wrong with this product’ to the time the FDA actually issued a recall,” she said.
This delay happens because recalls are only official after they have been investigated - and only that happens after hospitals have reported that patients had food poisoning. This will hopefully shorten the process by using information from the different Amazon reviews to trigger real-time investigations.
It is a fairly simple concept, but the technology that is required is complex. The idea is to take the text and compare it to previously recalled products so that they can “learn what in the text is actually indicating that this thing should be recalled, or this thing shouldn’t be recalled,” explained the project’s lead data scientist, Valentina Staneva.
This form of analysis is challenging and the team of Data Science Fellows from the UW's Data Science for Social Good program worked with the team, testing out different approaches to the problem. They flagged specific words in reviews to discover which would be recalled. These words included: "sick," "mold," and "vomit."
It also takes into account the number of ratings for a product over time. The red bar allows the readers to compare reviews to the FDA's recall. The volume of reviews has been challenging because for some items, there are thousands.
“We’re dealing with something that occurs very rarely,” Staneva said. If they were to use it to identify products that weren't recalled, it would be easier. However, because that isn't much help to society, it probably wouldn't be as beneficial. Over the next year, they will improve the program using UW students; they hope to expand it toward the use of social media.
In the meantime, they will be looking toward potential partnerships as well.
The Washington State Department of Health has shown an interest into the program, Nsoesie said, and the team hopes to build a real-time dashboard so that they could get calls to the health officials as quickly as possible.
Staneva said that the program would have to be tested rigorously if a partnership is pursued, considering that there are health implications involved. “The cost of missed recalled product could be somebody’s death,” she said.
The team is hoping to work with Amazon, as well. “I think it would be amazing if we could have a collaboration with them and they could help us get data and get it faster,” Nsoesie said. “So if we’re going to develop something that is real-time, then we need to have the data in real time.”