Nürnberg NLP at PsyDefDetect: Multi-Axis Voter Ensembles for Psychological Defence Mechanism Classification
- Detecting levels of psychological defence mechanisms in supportive conversations is inherently ambiguous. In the PsyDefDetect shared task at BioNLP 2026 the eight positive defence categories share surface language and differ only in pragmatic function and trained raters reach only moderate inter-annotator agreement. On such a task the decisive lever is not a stronger single model but error independence, since any single representation will waver on the overlapping defence boundaries. We translate this insight into a 9-voter ensemble spanning three orthogonal axes: class granularity (all nine classes for the gatekeeper, only the eight defence classes for the specialists), training method (generative and discriminative) and base model. The system reaches F1 = .420 on the hidden test set, placing first among 21 registered teams.
| Author: | Philipp SteigerwaldORCiD, Eric RudolphORCiD, Jens AlbrechtORCiD |
|---|---|
| DOI: | https://doi.org/10.18653/v1/2026.bionlp-2.9 |
| Parent Title (English): | Proceedings of the BioNLP 2026 (Shared Tasks) |
| Publisher: | Association for Computational Linguistics |
| Place of publication: | Stroudsburg, PA, USA |
| Document Type: | conference proceeding (article) |
| Language: | English |
| Reviewed: | Begutachtet/Reviewed |
| Release Date: | 2026/08/04 |
| Tag: | DMRS; Emotional Support Conversations; Ensemble Classification; Large Language Models; Psychological Defence Mechanisms; Shared Task |
| Pagenumber: | 7 |
| First Page: | 59 |
| Last Page: | 65 |
| institutes: | Institut für e-Beratung |
| Fakultät Informatik | |
| Fakultät Sozialwissenschaften | |
| Zentrum für Künstliche Intelligenz (KIZ) | |
| Research Themes: | Digitalisierung & Künstliche Intelligenz |
