The Fake Review Menace
In the age of the new internet, the freedom to advertise yourself and your product lie solely with you. You don’t need to pay a hefty fee for your product to reach millions of people with a shabbily written script and clumsy actors. You can just go online, make a video, blog or even a review under where you intend to sell the product.
From makeup to cleaning mops, and everything in between, you can write a review under where you intend to sell it and let people know your experience with it. This has taken the monopoly from the successful giving everyone a chance at success. The downside to this though is the manipulative purchase of the reviews itself.
The trend of selling products through famous online influencers where the reach is millions of people in one post has seen rapid growth. You have platforms that promote these that openly link the influencers with the producers. This leaves the larger audience at a disadvantage. Looking up reviews of products online is no longer helpful. Now, if you google ‘fake reviews’ you find an endless amount of material on how to spot and avoid them. But as the technology to spot them is developed the reviewers themselves get cleverer like self mutating viruses. There is growing popularity of sentiment analysis and opinion mining and an increasing number of publications on how to spot fake reviews. A wide range of algorithms is utilized to detect spammers. They in turn, learn from their mistakes and change their techniques to avoid these spam indicators making the algorithm redundant.
Research on how to automate the detection of these fake reviews is still in progress. Supervised algorithms have been developed and evaluated to do the job using tools of data analytics and opinion mining or sentiment analysis. Sentiment analysis is a part of Natural Language Processing that analyses the opinion, sentiment or the emotions of the writer towards the entity or product. Words like ‘beautiful’, amazing, lovely etc. are identified along with the semantic variations with which they are used. Different algorithms like apriori, FP (frequent pattern) growth, ECLAT, RElim etc are used to first identify the pattern and then make a distinction between genuine and fake reviews.
But what exactly happens with these algorithms?
Datasets of reviews are collected, both genuine and fake. The text is then classified into whether it was a good or a bad experience. The sample data is split into the positive and the negative part. The text is cleaned by removing unnecessary words, adding a POS (Parts of speech) tagging etc. Machine learning models are trained by sampling good, bad, real and fake reviews. Its time of upload, frequency with the user ID, language, length of the review, usage of words and quality of writing etc., are determined. For example, fake reviews tend to be shorter using vague language to describe the product which is most definitely not experienced.
Though there is no definite way to identify fake reviews there are promising results to predict and curb them. Let’s not forget that it is a joint effort of the consumers and corporations.
Nu-Pie provides Data Analytics Services & the best solutions to your business needs. Know more on Data Analytics services at https://www.nu-pie.com/ or connect to us for a free consultation on firstname.lastname@example.org.