Computational Text Analysis Methods for Developmental Research

Session Type: Paper Symposium

Abstract

Natural language processing (NLP) methods have become increasingly popular in social science research because these methods can reduce dimensionality of text in order to handle large amounts of data and better visualize latent relationships in the text. In addition, NLP methods may complement qualitative methods to provide more reliable and robust analyses of textual data (e.g., Baumer, Mimno, Guha, Quan, & Gay, 2017). Despite the utility and increasing popularity of NLP methods, it may be unclear to how to integrate NLP methods into developmental research. This symposium will leverage NLP methods to explore how these tools can be used to answer relevant questions in developmental research.

The first paper will examine how Latent Dirichlet Allocation (LDA) and qualitative methodologies complement each other in the analysis of a textual dataset of journal articles on exploratory behavior in children. The second paper will use LDA to analyze textual data from parenting subreddits and highlight potential applications to parenting intervention design. The third paper will use textual comparison methods like cosine similarity to compare how two textual datasets represent the experience of puberty. An expert in NLP will serve as discussant and contextualize what these methods may be used in broader child development research. Collectively, this series of papers will provide insights into how novel text analysis methods can benefit developmental research.

Sub Unit

• Methods, History, Theory

Chair

• Mary Kate Koch, University of Florida

Discussant

• David Mimno, Cornell University

Individual Presentations

• Comparing Traditional Qualitative Methods to Latent Dirichlet Allocation in the Analysis of Exploratory Behaviour in Children - Presenting Author: Lauryn Hagg, Deakin University; Non-Presenting Author: Anna Klas, Deakin University; Non-Presenting Author: Subhadra Evans, Deakin University; Non-Presenting Author: Christopher Greenwood, Deakin University; Non-Presenting Author: Elizabeth Mary Westrupp, Deakin University; Non-Presenting Author: Matthew Fuller-Tyszkiewicz, Deakin University; Non-Presenting Author: Stephanie Merkouris, Deakin University; Non-Presenting Author: George Youssef, Deakin University

• Text Mining of Reddit Posts: Using Latent Dirichlet Allocation to Identify Common Parenting Issues - Presenting Author: Elizabeth Mary Westrupp, Deakin University; Non-Presenting Author: Christopher Greenwood, Deakin University; Non-Presenting Author: Matthew Fuller-Tyszkiewicz, Deakin University; Non-Presenting Author: Tomer Berkowitz, Deakin University; Non-Presenting Author: Lauryn Hagg, Deakin University; Non-Presenting Author: George Youssef, Deakin University

• Text Analysis of Puberty Experiences From Retrospective and Concurrent Perspectives - Presenting Author: Mary Kate Koch, University of Florida; Non-Presenting Author: Jane E Mendle, Cornell University