Making Sense of Messy Survey Responses with R
Regular talk, 3:50 - 4:50 PM
The Governor’s office requested that thousands of open-ended survey responses be analyzed and translated into actionable goals within a short timeframe. Faced with this challenge, we asked a practical question: could R help us analyze a large volume of survey text in a way that was reproducible, transparent, and driven by themes emerging from the data rather than pre-defined (and potentially biased) categories?
This talk is designed for data enthusiasts of varying skill levels. I will walk through the fundamentals of text mining in R, including cleaning and tokenizing text, exploring term frequencies and n-grams. I will also provide a high-level introduction to sentiment analysis and topic modeling as methods for surfacing themes directly from the data.
The focus is not on building a perfect model, but on what it actually takes to get started under real-world constraints. Attendees will see practical workflows for organizing messy qualitative data and communicating insights to non-technical stakeholders. The goal is to demystify text mining and demonstrate how an approachable, reproducible process in R can turn open-ended survey responses into meaningful insight.
 | Pronouns: she/herPortland, OR, USA |