Around the Quad
Online Reviews Too Good to Be True?
Online retailers depend on reviews as a sales tool but can you believe them? Increasingly reviews are tainted by “opinion spam”—phony positive comments to encourage sales or negative remarks to downgrade competitors.
To help identify the real deal, Cornell researchers are developing computer software to pick up the clues hidden in false postings. In a test on 800 reviews of Chicago hotels, a computer picked out deceptive positive opinions with almost 90 percent accuracy.
The work was conducted by associate professor of communication Jeff Hancock, professor of computer science Claire Cardie, and graduate students Myle Ott, MEN ’07 and Yejin Choi, MS ’09.
The researchers asked a group of people to deliberately write false positive reviews of 20 Chicago hotels. These were submitted with an equal number of truthful examples to three human judges, who scored no better than chance in identifying deception. Next, the researchers applied computer analysis based on subtle features of the text. Truthful hotel reviews, for example, are more likely to use concrete words about the hotel like “bathroom,” “check-in,” or “price.” Deceptive entries more often include things that set the scene like “vacation,” “business trip,” or “my husband.” Also, truth-tellers use more nouns, while deceivers use more verbs.
Ott cautions that the work so far is only validated for reviews of Chicago hotels. Next, they’ll look
at other categories and at negative commentary.