Document Type


Date of Degree

Fall 2012

Degree Name

PhD (Doctor of Philosophy)

Degree In

Computer Science

First Advisor

Srinivasan, Padmini

Second Advisor

Pant, Gautam

First Committee Member

Srinivasan, Padmini

Second Committee Member

Pant, Gautam

Third Committee Member

Segre, Alberto

Fourth Committee Member

Cremer, James

Fifth Committee Member

Hourcade, Juan Pablo


The unprecedented amount of user generated content from emerging social media platforms like Facebook and Twitter make them invaluable sources of information for research. Twitter in particular has about 500 million registered accounts globally who are generating approximately 340 million messages daily containing personal updates, general life observations, opinions, moods, etc. Twitter's vast amount of data, which is generally available, offers an ideal source for mining entities' behaviors. This thesis explores two research streams involving mining Twitter data. In the first work, we seek to understand the Twitter-based stakeholder communication strategies of firms. We analyze tweets posted by firms to build a system that can automatically predict target stakeholder groups of a given tweet. We also examine and incorporate firm characteristics into the system for performance improvement. The result will potentially provide valuable business intelligence to market analysts who would like to discover social media strategies and behaviors of firms. In the second work, we investigate how readers from different parts of the world react to news headlines through their Twitter messages. We design a framework for data collection, statistical analysis, sentiment analysis, and language model comparison to understand the interests and reactions of Twitter users towards news headlines. The results from this work can possibly help news organizations have better understanding of their audience for better services. Though the two research directions may seem distinct, there are points of connection. In both cases, we are interested in the impact of companies (firms and news organizations). Moreover the methods used are similar. Our results illustrate that just by gathering Twitter data stream and developing a framework to examine them, we are able to discover many interesting insights about news readers and firms.


xii, 107 pages


Includes bibliographical references (pages 102-107).


This thesis has been optimized for improved web viewing. If you require the original version, contact the University Archives at the University of Iowa:


Copyright © 2012 Hung Viet Tran