The Development of an Automatic Text Classifier Enhanced Dashboard in Supporting Teachers' Facilitation of Virtual Problem-Based Learning Activities

• In Event: Technical Issues in Data Mining and Learning Analytics

Abstract

Virtual problem-based learning (PBL) environment can generate large amounts of textual usage data. However, it is challenging for instructors to understand the large amount of data and translate them into useful information during PBL activities. This study proposes a natural language processing (NLP) model using bidirectional encoder representations from transformers (BERT). BERT was trained to classify large amounts of textual data such as students’ written justifications. The results from applying this model was tested from both theoretical perspectives and statistical analysis, and a teachers’ dashboard prototype is created. This study has shown BERT model can be incorporated in learning analytic dashboards to translate the large amount of usage data into interpretable formats to assist teachers in tracking and facilitating PBL.

Authors

• Zilong Pan, The University of Texas at Austin

• Chenglu Li, The University of Texas at Austin

• Wenting Zou, The University of Texas at Austin

• Min Liu, The University of Texas at Austin