Text Themer: An AI solution for extracting themes from qualitative data

The Problem

Text Themer was developed to solve the problem of analyzing open-ended qualitative data, such as survey responses. Manually theming and categorizing open-ended responses is labour-, time- and cost-intensive, prone to human error, and meaningful insights from qualitative data are often lost or ignored entirely.

The Solution

The Open Text Themer solution leverages the power of word embeddings to analyze and extract themes from open-ended qualitative data in an automated and exploratory manner. Text Themer was built using open source tools within existing budget.

The Impact

Text Themer transformed a labour-intensive qualitative data analysis into a highly efficient process, saving significant time and resources. By leveraging Natural Language Processing and word embeddings, it reduces human error and uncovers hidden thematic patterns that traditional methods often miss. Ultimately, this ensures valuable insights are captured and empowers staff to focus entirely on data interpretation rather than manual sorting. Text Themer is a democratized solution -- available to all City of Edmonton employees.

Lessons Learned

The primary takeaway from the Text Themer project is that transitioning from exploratory tools to production-ready applications requires aggressive management of technical debt and a commitment to model modernization.

Launch year: 2022