A Lightweight Interactive Regular Expression Generator for Qualitative Coding in Quantitative Ethnography.

Quantitative ethnography approaches are often used to analyze large scale qualitative data. Manually coding such data is expensive and time consuming, if not impractical or impossible. In contrast, machine learning algorithms can code virtually unlimited amounts of data once a model has been created. However, machine learning approaches lack transparency and rely on large amount of training data. An alternative automated coding approach using regular expressions has the advantage of minimizing required training data while providing transparency. However, manually creating regular expressions during the coding process can be a very challenging task for many researchers. One potential solution to this challenge is automatic regular expression generation. Unfortunately, existing algorithms are all based on large pre-coded training data which is often unavailable in quantitative ethnography tasks. In this paper, we present a lightweight and interactive algorithm that actively constructs regular expression-based coding classifiers with the researcher. We use a simulation on an education data to show that the proposed algorithm is promising.