Depth based social media analytics

Abstract

The post-modern times has witnessed the progress from information age to an era where so people virtually exist online, doing everything from socializing to finishing business transactions in virtual space. So companies are finding more and more reasons to tap into this space to do business with their consumers, right from making virtual world a point of sale to getting feedback on product and process performance. Social media monitoring is automatically summarizing opinions and user reviews and accumulating it over the sum of information that a system can have access to. Some of the commonly used tools for web analytics today are Google Analytics, MetaTraffic, PostRank etc. These analytic engines can do tasks like market intelligence, monitoring reputation online, brand management, financial sentiment analysis, managing customer service managing political campaigns, finding traffic to a website etc. The engines typically make use of NLP (short for Natural Language Processing), that forms the basis for doing tasks like sentiment analysis, supervised learning or unsupervised learning. All these typically involve handling volumes of information, aggregated over individual reviews from online users. We propose a modification to the contemporary techniques for doing a depth based social media analysis, called the Feature Based Social Media Analysis. Instead of focusing just on the volume of information that can be compiled this system first identifies what are the possible features that can be associated with a product or a service. Once such features are in place then for each feature it builds a map of sentiments by running queries across various social media sites like Twitter or Facebook. Once the rating for individual features are obtained (called the Opinion Orientation Identification) then the features can be compiled together to get a holistic review on the product or the service. Such a system would be more up to date and more flexible for addition or removal of certain product features, with a new review attained by a simple recompilation with the new list of features. As of now the system will focus only on publicly available information, for instance if a tweet or a scrap or a post (in Twitter, Orkut and Facebook respectively) is made private then such instances cannot be accessed by this system. Other challenges include authenticity of the available information and managing multiple languages in the same platform.