Home // International Journal On Advances in Systems and Measurements, volume 13, numbers 3 and 4, 2020 // View article

Structural Equation Modeling with Sentiment Information and Hierarchical Topic Modeling

Authors:
Takurou Ogawa
Ryosuke Saga

Keywords: sentiment analysis, service analysis, structural equation modeling, hierarchical Latent Dirichlet Allocation, causal analysis

Abstract:
Service evaluation depends on various factors, such as assurance, responsiveness, and tangibles. Given that emotional satisfaction affects service satisfaction, analyzing both the evaluation and sentiments is important in improving service. Previous studies have identified the evaluation factor and determined the degree of influence on the resulting evaluation. However, there is little effective analysis that reflects the influence of such a factor on sentiment. In this study, we use hierarchal Latent Dirichlet Allocation and structural equation modeling (SEM) to express the causality relationships of service evaluation visually and quantitatively. Sentiment obtained quantitatively by using sentiment analysis is newly applied to SEM to obtain knowledge reflecting the influence of sentiment. As a result of the experiment, we can identify the causality of service and determine the influence of the evaluation factor and sentiment quantitatively. Furthermore, we conduct an experiment that compares a causal model with and without sentiment information and improve the model interpretability.

Pages: 230 to 239

Copyright: Copyright (c) to authors, 2020. Used with permission.

Publication date: December 30, 2020

Published in: journal

ISSN: 1942-261x