deberta-v3 - ECB hawkish/dovish/neutral classifier

Fine-tuned from microsoft/deberta-v3-base on annotated ECB communication (press conference statements, speeches and monetary-policy statements). Labels: dovish, neutral, hawkish.

Test-set performance

loss accuracy f1 macro f1 weighted
2.1243 0.4868 0.3933 0.3991

Usage

from transformers import pipeline
clf = pipeline("text-classification", model="LorenzoAleCon29/deberta-v3-ecb-hawkish-dovish")
clf("Inflationary pressures warrant a tightening of the policy stance.")

Training curves, trainer states and full reports: https://github.com/LorenzoAlessandro/BdF-Project

Downloads last month
9
Safetensors
Model size
0.2B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for LorenzoAleCon29/deberta-v3-ecb-hawkish-dovish

Finetuned
(666)
this model