Topic Modeling: Latent Semantic Analysis for the Social Sciences
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
2018 by the Southwestern Social Science Association Objective: Topic modeling (TM) refers to a group of methods for mathematically identifying latent topics in large corpora of data. Although TM shows promise as a tool for social science research, most researchers lack awareness of the tool's utility. Therefore, this article provides a brief overview of TM's logic and processes, offers a simple example, and suggests several possible uses in social sciences. Methods: Using latent semantic analysis in our example, we analyzed transcripts of the 2016 U.S. presidential debates between Hillary Clinton and Donald Trump. Results: Resulting topics paralleled the most frequent policy-related Internet searches at the time. When divided by candidate, changes in emergent topics reflected individual policy stances, with nuanced differences between the two. Conclusion: Findings underscored the utility of TM to identify thematic patterns embedded in large quantities of text. TM, therefore, represents a valuable addition to the social scientist's methodological tool set.