Over the past year, Halvor Tyseng, Markus Kreutzer, and Jonas Timmann Mjaaland have been working on how text embeddings can be applied to develop innovative methods aimed at qualitative research. Text embedding models are a group of AI models that represent texts as vectors in a semantic "meaning space." This space has the property that texts similar in content receive embedding vectors that are spatially close.
Representing text numerically makes it easier to process using more traditional machine learning methods. The work on text embeddings, in collaboration with Associate Professor Tor Ole Odden and Professor Anders Malthe-Sørensen, has led to the publication "Using Text Embeddings for Deductive Qualitative Research at Scale in Physics Education."
This publication was the basis for the presentation held at the Rochester Institute of Technology (RIT) in Rochester, NY. The presentation was hosted at the Center for Teaching in STEM, equivalent to CCSE at UiO.
Later in the week, a workshop was also held where doctoral candidates and postdocs tested the qualitative research method developed at CCSE. Both the presentation and the workshop received positive responses. The trip to the USA has helped to strengthen collaboration across the Atlantic and led to a concrete collaborative project using solutions developed by the team at CCSE to analyze data collected at RIT.
Professor Ben Zwickl from RIT and CCSE was a fantastic host, who, among other things, took Markus, Halvor, and Jonas to see RIT's hockey team play on Friday night. Between meetings and lectures, the research group also had the opportunity to visit the mighty Niagara Falls. "An amazing experience," says Markus. "Agreed!" Jonas and Halvor echo, "Rochester is a great place, and we have a terrific collaboration going that we look forward to continuing!" the young researchers conclude.
