Module 9: Determining Polarizing Entities and Opinions in Online Networks

Faculty Contact: Xifeng Yan

Concepts: Opinion mining, sentiment analysis, topic modeling, neural networks, social network analysis, NLP

Research Areas:

Datasets: Amazon reviews, Twitter social network

Abstract: The Web contains a wealth of opinions about products and people. While some demonstrate consensus with regards to a particular entity, others may show the opposite. The goal of this module is to explore methods to determine which entities are the subject of the most divergent opinions in a network. We will use online product reviews in order to determine which features of a given product (if any) cause the most disparate opinions amongst reviewers. Different methods for topic modeling as well as neural networks will be explored in this task. We also hope, time allowing, to apply our methods to social networks, such as Twitter, in order to determine which public figures are the most polarizing. Such analyses would enable us to ask further questions about online social networks. For example, are opinions shared amongst followers to a higher degree than would be expected in a random graph? Can we trace the spread of these polarizing opinions throughout the network?

Proposed by Tegan Brennan, CS Trainee, in March 2015

Active Quarters: 

  • Spring 2015, Tegan Brennan