An Introduction to Topic Modeling Using the Latent Dirichlet Allocation in Educational Research: Potential Applications and Limitations

Topic modeling using Latent Dirichlet Allocation (LDA) is a type of text mining approach. Text mining encompasses a range of techniques and processes for extracting information and knowledge from large collections of textual data. LDA, as a specific method within text mining, helps achieve this goal by uncovering underlying topics or themes present in a corpus of text documents. There is emerging interest in the use of approaches such as topic modeling in the field of educational research for such tasks which may require analyzing vast amounts of unstructured textual data, including open-ended responses to surveys or assessments, student essays, or other unstructured forms of text data. Topic modeling provides a means to identify themes (i.e., topics) within such data, helping researchers identify underlying structures and patterns that may be used to complement or serve the function of more time-consuming conventional methods of qualitative analysis. Particularly when studying novel phenomena at scale, topic modeling could be an especially useful method. At the same time, there are limitations to using topic modeling in educational research. In the present chapter, we introduce topic modeling by describing past educational research that has used this method, provide general steps for conducting topic modeling, and discuss the limitations of using topic modeling in educational research.