Newswise — KINGSTON, R.I. – March 25, 2014 -- Online retail sales in the United States are expected to reach $370 billion annually by 2017, a number that eclipses the government budgets of more than 40 nations. Yet as consumers grow comfortable with e-commerce, a University of Rhode Island engineering professor worries that it brings risks of sellers manipulating markets for profit.
Yan Sun, associate professor of computer engineering, wants to bring order to a digital world where it is unclear which online customer reviews can be trusted. And she’s doing it in a novel way, by using errors to identify fake reviews of everything from hotel rooms to socks.
“In engineering we can’t avoid errors, so instead we figure out if we can use them,” she said.
Sun’s research found that there should be no identifiable pattern to the reviews for a particular product. With that knowledge, she developed computer models that can analyze hundreds of reviews looking for patterns that are difficult to spot with the human eye. If the models find very few “outliers” to the reviews, the system raises a red flag.
Currently, the process is possible only in a lab setting. But Sun is working on a consumer version that would allow web shoppers to paste the address of a review page into a website that automatically analyzes the reviews for honesty. Such a system would require cooperation from major web retailers, which may be leery of how consumers would perceive the results.
Working with colleagues, Sun is also applying game theory to design a ratings system that encourages honest reviews. A seller that fakes reviews typically pays for inflated rankings, which increases the product’s price as the seller seeks to cover expenses. Sun wants to know at what point paying for reviews is no longer profitable.
“We want to make the whole e-commerce market healthier,” she said. “Instead of sellers spending money manipulating product reviews, have them spend money making their products better.”
The URI professor became interested in social computing when she realized the connections between the field and her expertise in signal processing. That field requires pattern recognition, which is directly translatable to analyzing hundreds or thousands of product reviews. The topic also matches her passion for using engineering to make a tangible difference in an average person’s life.
“I can talk to anyone on the street as long as they are older than eight and under 80 and they would probably understand and appreciate what I do,” she said.
Sun never set out to study social computing. As a child, she tagged along to her mother’s research lab in China, where her mother worked as a physicist. The particle accelerator and the taste of innovation captivated her. When Sun headed to college, her mother bemoaned how her lab needed more computer and electrical engineers to design better research equipment. With that on her mind, Sun entered college pursuing those fields.
After earning a doctorate at the University of Maryland, Sun set her sights on joining academia where she could research and inspire the next generation of engineers. She turned down three other job offers to come to URI in 2004, attracted by the University’s environment of fostering research and collaboration among disciplines.
# # #