Electronic Theses and Dissertations

Identifier

1195

Jun XieFollow

Date

2014

Document Type

Thesis

Degree Name

Master of Science

Major

Psychology

Concentration

General Psychology

Committee Chair

XIANGEN HU

Committee Member

Arthur C Graesser

Committee Member

Vasile Rus

Abstract

Online text streams, such as those found in blogs and Twitter, play a key role in public opinion and trend detection. This paper aims to propose a simple and intuitive semantic-based method to track online topic evolutions as a function of domains. The method decomposes online text streams into chronological and continuous Smallest Independent Corpora (SIC) via moving windows. The method was implemented in a discussion about a serious car accident found on a Chinese online social media site known as Sina Weibo. The study adopted irrelevant corpora, irrelevant topics, and human ratings to validate the method’s performance. The results indicated that the method can detect dominant domains as accurately as human raters. But it needs more improvement to detect the same trajectories of topic evolutions as human ratings. Furthermore the method is sensitive to the change of topics and corpora and its performance is not affected by window sizes.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.

Download

Share

COinS