Blog: Cornell Computers Spot 'Opinion Spam'

Cornell Computers Spot 'Opinion Spam'
Cornell Chronicle (07/25/11) Bill Steele

Cornell University researchers have developed software that can identify opinion spam, which are phony positive reviews created by sellers to help sell their products, or negative reviews meant to downgrade competitors. In a test of 800 reviews of Chicago-area hotels, the program was able to identify deceptive reviews with almost 90 percent accuracy. The researchers, led by professors Claire Cardie and Jeff Hancock, found that truthful hotel reviews were more likely to contain concrete words that had to do with the hotel, such as "bathroom," "check-in," or "price," while deceptive reviews contained scene-setting words, such as "vacation," "business trip," and "my husband." In general, deceivers use more verbs and honest reviewers use more nouns. The researchers found that the best results came from combining keyword analysis with the ways certain words are combined in pairs. The next step will be to see if the system can be extended to other categories, such as restaurants and consumer products, says Cornell graduate student Myle Ott.

View Full Article