Data Mining & Information System Lab   Database & Information Systems Group
Insight Finding with Sentiment Analysis of Large-scale Social Media

Background & Motivation

  • Current search engines are good on "Fact Finding" but not capable of "Insight Finding".

  • Fact Fining : What is the system requirements of Windows 7?

  • Insight Finding : Would the people buy it? If so, why?

  • Web 2.0 enables users to generate media and share them with others through internet

  • Information flow from one-directional to bi-directional

  • Ability to search for sentiment and insight become important

Goal of Project

  • Develop a noble insight finding method that can collect user generated media, expand the ontology automatically for broad domains, analyze and summarize the sentiments.
Method & Results
  • Develop insight crawler for collecting the consumer generated media.

  • Develop technologies for extracting sentiments using combination of statistical learning and rule-based methods; techniques for removing spam postings and weighting method for improving quality; classification models for classifying sentiment orientations and for determining the relevant domains.

  • Develop ontology expansion model for automatic discovery of new domains and for automatic expansion of predefined domains