Having Twitter followers, Facebook likes, YouTube views and subscribers has become an indispensable part of an organization’s online reputation nowadays. Whenever we come across a new company or product, we Google it and check out its website, its Facebook and Twitter and see the number of likes and followers respectively to assess whether the company is genuine or not. This has now become a general trend.
People who want to have a good online reputation are now ready to hire freelancers to boost their popularity on e-commerce websites and in social media. However, Facebook, Google and Twitter make their revenue based on the reliability of these sites and in order to take the control of the situation, Google has released findings on how you can automatically identify these fake reputation builders or crowdturfers (reputation salesmen).
These findings were published in a paper presented by a team of researchers and was partly also supported by a Google Faculty Research Award.
Lets understand what is Crowdturfing, but before that first understand what is Astroturfing, it is the practice of masking the sponsors of a message or organization (e.g. political, advertising, religious or public relations) to make it appear as though it originates from and is supported by grassroots participant(s). So Crowdturfing is - Crowdsourcing + Astroturfing.
According to the paper, detecting crowdturfing gigs is important as it enables them to remove these gigs before it can attract buyers, and eventually, it allows them to prohibit these sellers from posting such gigs. To identify crowdturfing gigs, the team has built machine-learned models by making use of the manually labeled 1,550 gig dataset.
Freelancer, Odesk and Fiverr are some of the examples of freelancing micro-task sites which have been successful in building a cottage industry out of part-time workers who are globally distributed. Nonprofit Samasource makes use of similar systems to help poverty stricken youth and women find work. Though these websites offer excellent services but are less trustworthy due to their fake reputations.
The researchers are hopeful that the technology developed by them will help companies like Twitter identify and automatically ban bots and fake followers.
The researcher’s database also reveals that the machine-learning detection system has a good rate of accuracy. Their experimental results clearly show that their models can detect crowdturfing gigs with an accuracy rate of 97.35 %. By using these models, they were able to spot 19,904 crowdturfing gigs in Fiverr and out of that 27.3 % were search engine targeting gigs, 70.7 % were social media targeting gigs and 2 % were user traffic targeting gigs.
You can read full research paper here - http://digital.cs.usu.edu/~kyumin/pubs/lee14icwsm.pdf .
Advertisements