Using latent semantic analysis to detect non-cognitive variables of academic performance

This thesis explores the possibilities of using latent semantic analysis to detect evidence of intrapersonal personality variables in post-secondary student essays. Determining student achievement based on non-cognitive variables is a complex process. Automated essay scoring tools are already in use today in grading and evaluating student texts based on cognitive domain traits, but at this time are not utilized to analyze non-cognitive domains such as personality. Could such tools be configured to detect non-cognitive variables in student essays? Key concepts in this proposal-personality traits, latent semantic analysis, automated essay evaluation, and online cinema reviews-are explored followed by a literature review to justify the research. As a proof of concept study, 43 writing samples written to a constructed response task are collected and analyzed by a test model specifically designed to evaluate sentiment in a movie review constructed response format. A test model is created using LightSIDE, a software tool for text assessment, to predict the sentiment of these essays with highly encouraging results. The thesis concludes with a path for future research in the largely unexplored area of automated assessment of non-cognitive variables.

Name thesis:degree_name

Master of Science in Professional and Technical Communication - (M.S.)

Discipline thesis:degree_discipline

Humanities

Year

2014

Author dc:creator

• Aalderks, Daniel Richard

Contributors dc:contributor

• Norbert Elliot
• Andrew Klobucar
• Judith D. Redling

• Latent semantic analysis
• Student achievement
• Non-cognitive variables
• Technical and Professional Writing

Repository record dc:identifier

https://digitalcommons.njit.edu/theses/181

OAI identifier oai:identifier

oai:digitalcommons.njit.edu:theses-1180