# Learning Analytics from Spoken Discussion Dialogs in Flipped Classroom

**Authors:** Hang Su, Borislav Dzodzo, Changlun Li, Danyang Zhao, Hao Geng, Yunxiang Li, Sidharth Jaggi, Helen Meng  
**arXiv:** [2301.12399](https://arxiv.org/abs/2301.12399)  
**Date:** 2023-01-29  
**Summary page:** https://www.summarizepaper.com/en/paper/2301.12399v1/

## Key points

- The flipped classroom is a popular pedagogical approach
- It involves watching video-recorded lectures at home and group-based problem-solving discussions in class
- Spoken discussion dialogues are common and provide valuable information about learning processes and outcomes
- Learning analytics can be used to collect and analyze spoken discussion dialogues
- Researchers recorded in-class group discussions, manually transcribed them, and extracted features using multiple tools and customized processing techniques
- Statistical analyses were performed to explore indicators related to group learning outcomes from face-to-face discussion dialogues in the flipped classroom
- Machine learning algorithms were applied to these indicators to predict group learning outcomes as High, Mid, or Low with 78.9% accuracy
- Analyzing spoken discussion dialogues can help educators understand how groups learn and identify areas for improvement
- This approach could potentially be scaled up for use across different classrooms and educational settings

## Summary

The flipped classroom is a pedagogical approach that has gained popularity in recent years. It involves students watching video-recorded lectures at home before attending group-based problem-solving discussions in class. Spoken discussion dialogues are common in the flipped classroom and provide valuable information about the learning processes and outcomes of students. This study focuses on using learning analytics to collect and analyze spoken discussion dialogues in order to gain insights into group learning processes and outcomes. The researchers recorded in-class group discussions throughout the semester and manually transcribed them. They then extracted features from the dialogues using multiple tools and customized processing techniques. Statistical analyses were performed to explore indicators related to group learning outcomes from face-to-face discussion dialogues in the flipped classroom. Machine learning algorithms were applied to these indicators to predict group learning outcomes as High, Mid, or Low. The best prediction accuracy achieved was 78.9%, demonstrating the feasibility of automatic learning outcome prediction from group discussion dialogues in the flipped classroom. This study provides evidence that analyzing spoken discussion dialogues can help educators understand how groups learn and identify areas for improvement. By leveraging machine learning algorithms, this approach could potentially be scaled up for use across different classrooms and educational settings, providing valuable insights into student learning outcomes.