Text-Mining Open-Ended Survey Responses Using Structural Topic Modeling: A Practical Demonstration to Understand Parents’ Coping Methods During the COVID-19 Pandemic in Singapore

Abstract

Objective: Open-ended survey questions crucially contribute to researchers' understandings of respondents' experiences. However, analyzing open-ended responses using human coders is labor-intensive and prone to inconsistencies. Structural topic modeling (STM) is a text mining method that discover topics from textual data. We demonstrate the use of STM to analyze openended survey responses to understand how parents cope during COVID-19 lock-down in Singapore. Method: We administered online surveys to 199 parents in Singapore during the COVID-19 lock-down. To show a STM analysis, we demonstrated a workflow that includes steps in data preprocessing, model estimation, model selection, and model interpretation.Results: An 18-topic model best fitted the data based on model diagnostics and researchers' expertise. Prevalent coping methods described by respondents include "Spousal Support", "Routines/Schedules" and "Managing Expectations". Topic prevalence for some topics varies with respondents' levels of parenting stress and whether parents were fathers or mothers. Conclusion: STM offers an efficient, valid, and replicable way to analyze textual data such asopen-ended survey responses and case notes that can complement researchers' knowledge and skills. STM can be used as part of a multistage research process or to support other analyses such as clarifying quantitative findings and identifying preliminary themes from qualitative data.

All Lives Matter discussions on Twitter: Varied use, prevalence, and interpretive repertoires

“…For example, when words like ‘mouse’ ‘bird’ and ‘dog’ co‐occur in a dataset, they might get computed as one topic, which could be labelled ‘urban animals’. STM is also a mixed membership model because it allows each document (here a tweet) to consist of a range of topics (Chung et al, 2022; Roberts et al, 2014). STM can ‘account for polysemy (i.e., a word's different contextually based meanings)’ (Chung et al, 2022, p. 6).…”

“…STM is also a mixed membership model because it allows each document (here a tweet) to consist of a range of topics (Chung et al, 2022; Roberts et al, 2014). STM can ‘account for polysemy (i.e., a word's different contextually based meanings)’ (Chung et al, 2022, p. 6). An example is how the word ‘mouse’ could co‐occur with other animals and words connected to technology (computer mouse).…”

Latent Dirichlet allocation topic modeling of free‐text responses exploring the negative impact of the early COVID‐19 pandemic on research in nursing

“…Today, as digitized data are more available and text‐mining software becomes more accessible than before, many researchers have come to use topic modeling (Buenano‐Fernandez et al, 2020; Chung et al, 2022; Pietsch & Lessmann, 2019; Vijayan, 2021). Topic modeling is one of the natural language processing (NLP) techniques in the field of machine learning.…”