The promise of computational text analysis for augmenting qualitative research in demography

Margaret Frye, Princeton University

This paper explores the advantages of applying computational text analysis to qualitative data in demography. I begin by examining some challenges that demographers are particularly likely to face in analyzing qualitative data—large amounts of data and the difficulty of comparing themes across external categories—and discussing ways that new tools from machine learning and computer science might help to address these challenges. I then describe three applications of text analysis using a set of conversational journals about HIV/AIDS from Malawi. These applications vary in the extent to which computational techniques either supplement or supplant more traditional methods of qualitative data analysis. In the first example, computational techniques are used only for sample selection; in the second, to analyze particular themes over time, and in the third, a computer algorithm is trained to identify latent themes in the text as well as reveal how they vary over time and across individuals.