Degree Type

Dissertation

Date of Award

2015

Degree Name

Doctor of Philosophy

Department

Greenlee School of Journalism and Communication

Major

Human Computer Interaction

First Advisor

Eric Abbott

Abstract

Opinions are valuable, and with the advent of social media, plentiful. Opinions are not always intelligible, however. Therefore, many of the views of social media users are ignored. This dissertation seeks to confront the challenges associated with opinion mining and sentiment analysis by investigating three aspects of opinion expression and consumption in social media. The universality of opinion itself is explored through an innovative application of social science research in survey construction, semantic distance analysis, and corpus linguistics. Results include a universal taxonomy of 18 sentiment types shown to be portable across 15 languages. The universality of opinion processing is explored through a qualitative meta-synthesis (QMS) analysis of social psychology, opinion mining and sentiment analysis, and voting systems scholarship. Results include a comprehensive theoretical model of opinion processing: the States, Processes, Effects, and Quality (SPEQ) model for opinion mining and sentiment analysis. SPEQ defines seven states of opinion, six processes which govern the transitions between those states and five quality and integrity measures for the evaluation of those processes. Lastly, the concept of a structured opinion syntax is explored. Despite strong resentment to symbolic representations of meaning by subjects, learning and priming effects for both the encoding and decoding of structured opinion support the contention that such a syntax could be developed and used. Many future directions for research are presented for each aspect of opinion investigated.

DOI

https://doi.org/10.31274/etd-180810-4842

Copyright Owner

Erin Mikel Phillips

Language

en

File Format

application/pdf

File Size

221 pages

Share

COinS