GrapeVine: Tracking the Pulse of Businesses using Twitter
Twitter is no longer simply a social network but a "social information network" used for sharing information, ideas and opinions of matters of mass importance. This is especially true in the consumer-business sector where companies are leveraging it to reach out and "talk" to their consumer base at an unprecedented scale and speed. However, the low signal to noise ratio and the lack of structured data in Twitter's data stream makes it much harder to "listen" to what consumers are saying about the company: their opinions, feedback and sentiment. Almost half of all tweets are personal conversations, users' self promotion, random observations or spam, which are not of any use for a business-oriented use case. Moreover, Twitter does not formally attempt to organize tweets based on any contextual or categorical information, nor does it collect any detailed information about its users beyond an optimal, short textual description. We propose the design of a Twitter-based consumer-business tool which can help companies bridge this communication gap and better monitor and analyze in near real-time, the opinion, sentiment, and feedback that exists about them amongst the Twitter consumer base. We implement and evaluate the following basic features of this tool: 1) a classifier which distinguishes between "relevant" and "irrelevant" tweets, from a business perspective, and filters the data stream, 2) an algorithm to discover the "top tweeters" mentioning a company and rank them based on their expertise in the company's domain and 3) an algorithm to detect trending stories pertaining to a company.