Sign In

Communications of the ACM

ACM TechNews

Rumor-Detector Software Ids Disputed Claims on Twitter

View as: Print Mobile App Share:
One minute after a hacked Twitter account sent out a rumor of explosions at the White House, users were already inquiring about its accuracy. Blue nodes show inquiry tweets. Yellow represent correcting tweets.

Researchers at the University of Michigan have developed software that can detect rumors in tweets on Twitter.

Credit: Zhe Zhao

University of Michigan professor Qiaozhu Mei and colleagues recently unveiled software that can quickly detect rumors on Twitter.

In one demonstration involving 30 million tweets related to the Boston Marathon bombing in April 2013, the software successfully detected 110 rumors, with an average accuracy of more than 50 percent. Methods that use trending topic detection and hashtags as signals detect rumors with less than 10 percent accuracy. Moreover, the software is faster at finding controversial statements that can be fact-checked.

The software is designed to listen in on Twitter traffic for words and phrases that suggest users are "questioning the truth value of information," including "unconfirmed," "Is this true?," and "Really?" The goal is to correct erroneous claims and help limit the negative effects when they spread in these hyperconnected times.

The software looks for more tweets about a controversial statement to determine how far the rumor has spread, but the team relies on humans for fact-checking.

The researchers plan to launch a Web service called Rumor Lens in the next few months.

From University of Michigan
View Full Article


Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA


No entries found