Downloads · 30 days
0
HarmoniCA_apathy
julia-pfarr/HarmoniCA_apathy
HarmoniCA_apathy is a text classification model from julia-pfarr. Use it when you need a label for a piece of text. It is set up for sentence-transformers. The card lists the license as cc-by-4.0.
Part of the HarmoniCA collection for harmonising psychiatric questionnaire items across studies. This model assigns items from apathy questionnaires to one of two theoretically motivated symptom dimensions.
Access
Public
Updated Jun 8, 2026
Repo size
1.1 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors1.1 GB · 100%
At a glance
- Task
- Text Classification
- Library
- sentence-transformers
- License
- cc-by-4.0
- Access
- Public
- Created
- Jun 8, 2026
- Updated
- Jun 8, 2026
- SHA
- e2ec4354
Try a prompt
Base models
Spaces
Model card
From the Hugging Face model README
HarmoniCA — Apathy Model
Part of the HarmoniCA collection for harmonising psychiatric questionnaire items across studies. This model assigns items from apathy questionnaires to one of two theoretically motivated symptom dimensions.
Dimensions
| ID | Label | Description |
|---|---|---|
| 1 | Behavioral-Cognitive Apathy | goal-directed cognition and behavior, initiative, planning, curiosity, effort, productivity |
| 2 | Affective Apathy | emotional responsivity, emotional blunting, indifference to positive or negative events |
Items rated as not belonging to the apathy construct are assigned dimension -1 (Does not fit).
Model Architecture
- Type: ContrastiveModel (contrastive fine-tuning of a sentence transformer)
- Base encoder: BAAI/bge-large-en-v1.5
Training Data
- Training items: 127 expert-labeled questionnaire items
- Questionnaires: AES, AMI, LARS, SAS, MDS-UPDRS (apathy items), NMSSQuest, NPI, DAS, SHAPS, TEPS, SCL-90 apathy subscale, plus data augmentation
- Labels: Expert consensus from a survey of 40 psychiatrists/neurologists (round 1)
Evaluation
Evaluated on 68 held-out items from from unseen questionnaires against expert labels.
| Cohen's κ | Accuracy |
|---|---|
| 0.55 | 0.78 |
Cross-validation (leave-one-questionnaire-out, 11 folds): mean accuracy 77.9% ± 25.0%.
Limitations
- Trained and validated on Parkinson's disease cohort questionnaires; performance may differ in other populations.
- Dimension structure reflects theoretical consensus among a specific expert panel and may not match all theoretical frameworks.
Citation
in preparation
- Task
- Text Classification
- Library
- sentence-transformers
- License
- cc-by-4.0
- Languages
- en
- Created
- Jun 8, 2026
- Updated
- Jun 8, 2026