Classifying Different Types of Talk During Collaboration

Abstract

Pair programmers utilizing more Exploratory (critical, constructive) talk has been shown to help students achieve a better mutual understanding of problems they are solving. In this paper, we investigate the promise of fine tuning a pretrained transformer-based machine learning model to classify utterances into Exploratory, Cumulative, and Disputational talk. The task of classifying utterances into different types of collaborative talk was approached as a multi-label text classification problem. This is the first successful automatic classification of utterances into the different types of collaborative talk.