ASC Workshop: Using Topic Models to Qualitatively Code Large Amounts of Text

Session Submission Type: Workshop

Abstract/Description

Given advances in computing and storage, qualitative researchers are now having to grapple
with massive amounts of text data, which may be too difficult or invite too many errors to code
by hand. In these circumstances, the data science technique of topic modeling may be useful.
Topic modeling is machine learning technique that aids in qualitative coding of different forms
of text documents by examining the documents for patterns in words or phrases, then clustering those words and phrases into “topics” or “themes.” Two types of topic models are most commonly used in criminology: structured topic modeling and biterm topic modeling.
This workshop will expose attendees to the basics of topic modeling and give attendees hands
on practice with real text data available on ICSPR. During the first part of the workshop,
attendees will receive an overview of topic modeling generally, with a focus on structured
topic modeling and biterm topic modeling. Then using RStudio, during the second part of the
workshop, attendees will be directly working with two sets of text data to thematically code the
data using both structured topic and biterm topic models. To participate in the hands-on
activities, attendees should have some exposure to and knowledge of RStudio and the most recent version of RStudio on their computer to use in the workshop. All code and data will be provided.

Sub Unit

• Area XVII. Workshops / Qualitative Methods

Instructors

• Danielle Wallace, Arizona State University

• Connor Stewart, Arizona State University