Short Description

When analysing human-2-human interaction and collaboration in physical spaces, digital systems are challenged to understand this interaction. Similarly, researchers who analyse (many) human-2-human interactions are challenged to do this manually. In this master thesis, the goal is to test and adapt existing approaches for computational conversation analysis for the specific case of audio recordings of conversations led between two individuals while doing an Alternative Uses Test with or without the help of ChatGPT.

Your Task

Speech-to-text conversion

Automatic identification of turns (who is speaking)

Temporal analysis of turn taking (who speaks for how long, pauses)

Option 1: Do individuals who work together well adapt their speaking towards each other too? This requires a computational estimation of the degree of adaptation of the speakers towards each other (i.e., entrainment) and an analysis of its relationship to task-perfor- mance measures.

Option 2: Is the fluency of the spoken turn-taking an indicator for the collaborative task per- formance? This requires a computational identification of the communication flow and an analysis of its relationship to task performance measures.

Your Profile (recommended)

Python, Machine Learning, principles of human speech communication

Contact:

Barbara Schuppler (b.schuppler@tugraz.at) and Viktoria Pammer-Schindler (viktoria.pammer-schindler@tugraz.at)