Ph.D. Thesis Proposal Announcement

Title: Computational Approaches to Understanding Stylistic Variation in Online Writing

Umashanthi Pavalanathan

Ph.D. Student 

School of Interactive Computing 

College of Computing 

Georgia Institute of Technology

Date: Wednesday, April 19, 2017 

Time: 11:00 AM-1:00PM EDT (8:00 AM-10:00AM PDT)

Location: TSRB 223

Committee:

Dr. Jacob Eisenstein (Advisor), School of Interactive Computing, Georgia Institute of Technology 

Dr. Eric Gilbert, School of Interactive Computing, Georgia Institute of Technology 

Dr. Munmun De Choudhury, School of Interactive Computing, Georgia Institute of Technology

Dr. Michael Gamon, Knowledge Technologies Group, Microsoft Research

Abstract:

Language use in online interactions varies from community to community, from individual to individual, and even for individuals in different contexts. While prior work has identified these differences, far less is understood about why these differences have arisen in online writing. My thesis focuses on this 'why' question. The reasons for linguistic diversity in online writing could be multifold. As more and more interpersonal social interactions are conducted through mediated channels, there is an increasing need to express multiple social meanings in varied social situations through linguistic means. In the absence of non-verbal cues, the technology-mediated channels provide several affordances to conduct interpersonal interactions. How do factors such as the need to convey varied social meanings in online interpersonal interactions and the affordances in technology- mediated channels shape online writing? My thesis work investigates this interplay through a series of large-scale computational studies of linguistic style variation in online writing. 

Using data-driven  methods to extract linguistic variables and employing causal statistical analysis, I have investigated the social meaning of varied non-standard language usage in social media and the effects of new technological affordances in online social platforms on individuals' writing style. Further, I have developed a multi-dimensional style lexicon using unsupervised techniques to quantitatively study stylistic variation in online communities. My proposed work will further explore the social dimensions of variation in online writing by focusing on (1) linguistic style accommodation and community reception in online multi-community environments such as Reddit and (2) the effects of norm enforcement in online writing style. My thesis work will advance our understanding of how individuals utilize the affordances in online social platforms and shift style to achieve varied social goals.  Understanding the social dimensions of linguistic style variation in online writing has important consequences for the design of language technology and social computing systems, and beyond.