18 December 2018

Research pick: Fake news and fake reviews - "Online social networking services and spam detection approaches in opinion mining – a review"

Consumers have the opportunity to express their views about the products and services they use in ways that were simply not technologically possible a decade ago. Social media and social networks allow them to opine wildly to fellow users and also directly and in public to the companies that provide those products and services.

Of course, with any system involving subjectivity and more critically, money, there is likely to be gaming of that system on both sides. An unscrupulous company may attempt to spam the system and improve its ratings artificially. Conversely, an individual or pressure group with a particular grudge might wish to sabotage that company’s ranking.

Writing in the International Journal of Web Based Communities, Meesala Shobha Rani and S. Sumathy of the School of Computer Science and Engineering, at Vellore Institute of Technology, in Vellore, India, have looked to nature for inspiration to find the best way to root out opinion spam. They have reviewed algorithms that use the notions of “a moth to a flame”, “grey wolf hunting”, or “flower pollination by insects”. Their review looks at how well different approaches are able to detect fake opinions on social media. The same approaches might also be useful in spotting fake political opinion and even fake news.

Customers depend on e-commerce sites to save them shopping time and they require accurate and honest reviews on those sites to help them in their decision making, the team reports. However, spammers, often in exchange for payment can post a fake opinion, good or bad and so degrade the quality of the reviews. In their assessment of the different approaches, the team found that “grey wolf” is the more effective and might well be adopted by organizations and companies hoping to detect and delete fake opinion on their systems.

Rani, M.S. and Sumathy, S. (2018) ‘Online social networking services and spam detection approaches in opinion mining – a review‘, Int. J. Web Based Communities, Vol. 14, No. 4, pp.353-378.

No comments: