Tracking conflict and emotions with a computational qualitative discourse analytic support approach

Area of Science:

• Computational Linguistics
• Psychology
• Artificial Intelligence

Background:

• Manual labeling of emotional states in discourse analysis is subjective and time-consuming.
• Current methods lack objectivity and may miss dynamic emotional shifts.
• Accurate emotional state inference is crucial for understanding conversational dynamics.

Purpose of the Study:

• To introduce a novel qualitative discourse analytic support method for labeling emotional states over time.
• To enhance the objectivity and repeatability of emotional state analysis in conversations.
• To specifically identify and analyze instances of inter-speaker conflict in discourse.

Main Methods:

• A two-step machine learning approach was developed.
• The method supports the temporal mapping of emotional states of conversational participants.
• The technique was validated using well-studied broadcast interviews.

Main Results:

• The machine learning approach effectively decodes the emergence, sustainment, and resolution of conflict.
• The method provides evidence of emotional state changes in interlocutors over time.
• Findings demonstrate the utility of the technique in supporting in-depth discourse studies.

Conclusions:

• The proposed method offers a more objective and efficient way to analyze emotional dynamics in conversations.
• This tool can aid researchers in identifying and understanding moments of conflict and emotional shifts.
• The approach has the potential to significantly advance qualitative discourse analysis.