A proposed system for opinion mining using machine learning, NLP and classifiers

Abstract

In today’s life consumer reviews are the part of everyday life. User read the reviews before purchase, or stores it for finding the best product through comparison of the product review. From customers view point the reviews play vital role to make a decision regarding an online purchase as well as spammers to write the fake reviews which can increase or defame the reputation of any product. Spammers are using these platforms illegally for financial benefits/incentives are involved in writing fake reviews and they are trying to achieve their motive in terms of financial or to defeat the competitor which causes an explosive growth of sentiment/opinion spamming of writing forged/fake reviews. The present studies and research are used to analyse and categorize the opinion spamming into three different detection targets opinion spam, spammers, and to find the collusive opinion spammer groups so that false opinions can be avoided. Opinion spamming further divided into three different types based on textual and linguistic, behavioral, and relational features. The motivation behind this work is to study the dynamics of spam diffusion and extract the latent features that fuel the diffusion process. The user-based features and content-based features have been used for the categorization of spam/non-spam content. The contributions of this work are building the datasetwhich assists as the ground-truth for classifying/analyzing the variation of fraud/genuine and non-spam/spam information diffusion and to analyze the effects of topics over the diffusibility of non-spam and spam evidences/information. The paper, carried out an in-depth analysis of Twitter Spam diffusion.