raphael-r/bright-gliner-demographics

BRIGHT NER: GLiNER2 fine-tuned for demographics

Description

This is a GLiNER2 architecture fine-tuned to extract clinical neuro-oncology entities related to the demographics semantic group. It was trained on a synthetic dataset generated for the properly de-identified BRIGHT project dataset (see the generated_data folder in the primary repository).

This model repository was specifically designed to fit within the bright_db overarching namespace.

Fields

It extracts the following fields (described in French):

• sexe: Sexe du patient (M ou F)
• annee_de_naissance: Année de naissance (entier)
• activite_professionnelle: Profession ou métier du patient
• antecedent_tumoral: Antécédent personnel de tumeur cérébrale
• ik_clinique: Score Karnofsky (KPS 0-100) ou score OMS (PS 0-4)
• dominance_cerebrale: Dominance hémisphérique (droitier, gaucher)
• neuroncologue: Nom du neuro-oncologue
• neurochirurgien: Nom du neurochirurgien
• radiotherapeute: Nom du radiothérapeute
• anatomo_pathologiste: Nom de l'anatomopathologiste

Performance on Validation Set

Aggregates:

• Macro F1: 0.5720 (Precision: 0.5270, Recall: 0.7990)
• Micro F1: 0.6623 (Precision: 0.5677, Recall: 0.7949)

Per-Label Breakdowns:

Label	Precision	Recall	F1
sexe	0.9677	0.2875	0.4433
annee_de_naissance	0.8633	0.9557	0.9072
activite_professionnelle	0.4524	1.0000	0.6230
antecedent_tumoral	0.2222	0.1818	0.2000
ik_clinique	0.9767	1.0000	0.9882
dominance_cerebrale	0.3215	1.0000	0.4866
neuroncologue	0.4108	1.0000	0.5823
neurochirurgien	0.4291	0.9912	0.5989
radiotherapeute	0.3793	0.8148	0.5176
anatomo_pathologiste	0.2470	0.7593	0.3727

Usage

# Inference Code
from gliner2 import GLiNER2

model = GLiNER2.from_pretrained("raphael-r/bright-gliner-demographics")
text = "Patient presenting with epileptic seizures..."
entities = model.extract_entities(text)

for entity in entities:
    print(entity["text"], "=>", entity["label"])