|Group 1||Qualitative Approaches||Diane Rasmussen Neal (University of Western Ontario, Canada)|
|Group 2||Social Network Analysis||Bernie Hogan (University of Oxford, UK)|
1. To what extent does influence in one social media platform translate into influence on another. That is, do my recommendations, tastes and advice have as much sway on Twitter as on linked in?
2. Do differences in platform design make a difference to network structure and if so how?
3. When is it useful to think about configurations or motifs for influence? That is, when do we think not just about centrality/position and not about community structure but the triads and cycles that lead to larger network structures.
4. Will there ever be a network that consolidates all the different social network sites and if so how? If not, what are the barriers both practical and technical?
|Group 3||Political Mobilization & Engagement||Greg Elmer (Ryerson University, Canada)|
|Group 4||Social Media and Marketing||Ramesh Venkat (Saint Mary’s University, Canada)|
|Group 5||Opinion Mining and Sentiment Analysis||Nancy McCracken (Syracuse University, USA)|
|Our overall research question is: How do we use opinion mining and sentiment analysis for studying influence on social media? As a result of our discussion, we would like to set up some future directions for this research area. What are the important aspects we can explore and contribute to the current body of knowledge? To help answer this, we suggest the following discussion questions.Discussion Questions1. How is opinion mining or sentiment analysis related to studying influence on social media, in other words, what is the motivation of studying influence using opinion mining and sentiment analysis? Are there any theories of how one’s opinions or sentiment, or how they express them, affects their influence on social media?|
2. What specific questions of opinion mining or sentiment analysis and in what types of social media would be useful? Within the broad outlines of detecting sentiment or opinion, researchers often investigate more specific questions, such as recognizing stance or disagreement in online debates or detecting sarcasm on Twitter. Which ones of these do you think are important for determining influence (you may want to think of specific influence scenarios) or what other ones do you know of or would you propose?
3. Current methods for sentiment analysis and opinion mining usually fall into the unsupervised camp, with methods such as counting the number of positive or negative words in a sentiment lexicon, or into the supervised camp, with machine learning methods that required data annotated either by humans or taken from situations where the results are known. Can we apply these existing opinion mining/sentiment analysis methods for the study of influence on social media? What are the available tools? How do we use these tools for this specific purpose? What are the challenges?
4. Accuracy of opinion mining and sentiment analysis generally ranges in the 70-80% in the most successful cases, where accuracy is defined by comparing each document or utterance. (And in many cases, human inter-annotator agreement is about 80% at best.) Is this type of accuracy good enough for use in influence applications? Are there other types of evaluations that would be useful?