N"urnberg NLP at PsyDefDetect: Multi-Axis Voter Ensembles for Psychological Defence Mechanism Classification
Researchers developed a multi-axis voter ensemble to classify psychological defence mechanisms in supportive conversations. By combining different class granularities, training methods, and base models across nine voters, the system achieved an F1 score of .420. This demonstrates that leveraging independent voter errors is key to overcoming ambiguity in detecting overlapping defence boundaries.