A Structural Topic Sentiment Model for Text Analysis

47 Pages Posted: 6 Apr 2022

See all articles by Li Chen

Li Chen

Cornell University - Samuel Curtis Johnson Graduate School of Management

Shawn Mankad

North Carolina State University - Department of Business Management

Date Written: January 29, 2022

Abstract

We consider the common setting where one observes a large number of opinionated text documents and related covariates, such as the text of online reviews along with the date of the review and the author demographic information. In this setting it can be of interest to understand how the covariates determine the text composition as well as the prevalence and sentiment of various discussion themes. Yet, most topic modeling methods are designed to summarize the text for the purpose of exploratory analysis, not to perform this type of formal statistical inference. Further, topic modeling methods generally do not try to estimate the sentiment of discussion along separate topics which can be critical in business applications (e.g., for summarizing service or product quality). We develop a topic model called the Structural Topic Sentiment (STS) model that introduces a new document-level latent sentiment variable for each topic, which modulates the word frequency within a topic. These latent topic sentiment variables are controlled by document-level covariates to allow for experimental control and regression analysis. We also introduce new computational methods to resolve scalability issues that have forced previous models to restrict to a small number of categorical covariates. We benchmark the STS model on three real-world datasets from surveys, blogs, and Yelp restaurant reviews around the COVID-19 pandemic. Our model recovers meaningful results including rich insights about how COVID-19 affects online reviews, demonstrating that the STS model can be useful for regression analysis and potentially causal inference with text data in addition to topic modeling's traditional use of descriptive analysis.

Keywords: Causal inference, topic prevalence, topic sentiment, text analysis, Laplace approximation, Poisson regression, nonnegative matrix factorization

Suggested Citation

Chen, Li and Mankad, Shawn, A Structural Topic Sentiment Model for Text Analysis (January 29, 2022). Available at SSRN: https://ssrn.com/abstract=4020651 or http://dx.doi.org/10.2139/ssrn.4020651

Li Chen (Contact Author)

Cornell University - Samuel Curtis Johnson Graduate School of Management ( email )

Ithaca, NY 14853
United States

Shawn Mankad

North Carolina State University - Department of Business Management ( email )

Raleigh, NC 27695
United States

HOME PAGE: http://mankad-research.github.io/

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
55
Abstract Views
173
rank
504,013
PlumX Metrics