Supervised Categorization of Open Response Feedback in Higher Education

Ville Antero Kivimäki, Thomas Bergström, Jiri Lallimo, Alexander Jung

Forskningsoutput: Kapitel i bok/rapport/konferenshandlingKonferensbidragVetenskapligPeer review


Efficient use of student feedback is mandatory for the continuous development of curricula and
teaching skills. Student feedback can be collected in various ways such as using Likert scale questions
or open responses to particular questions. While Likert-type data is rather easy to process using
standard analytic tools, open response feedback is more challenging. However, studies suggest that
open response items written in natural language provide qualitative and situated insights for practical
course development. In large-scale courses, processing open responses through human labour becomes
unfeasible in volume, too complex by content and might lead to serious bias. In an age of quality
assurance, tenure track systems and large-scale courses on digital platforms, teachers need new tools
to process natural language data collected on their courses. In this paper, we study the application of
machine learning methods for classifying open response student feedback data in a higher education
institution (HEI) context to support course development based on summative course feedback. We
develop a model for data processing and apply it in Microsoft Azure Machine Learning Studio (AMLS).
For model validation, we use human-processed training data consisting of 1580 feedback items. We
end up suggesting semi-structured formulation for collecting open response items as a part of
summative course feedback, based on our findings and related literature.
Titel på gästpublikationEUNIS 2019 Congress Proceedings
Antal sidor4
StatusPublicerad - 2019
MoE-publikationstypA4 Artikel i en konferenspublikation

Citera det här