raphael-r/bright-gliner-dates_outcomes

BRIGHT NER: GLiNER2 fine-tuned for dates_outcomes

Description

This is a GLiNER2 architecture fine-tuned to extract clinical neuro-oncology entities related to the dates_outcomes 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):

• date_chir: Date intervention neurochirurgicale ou résection
• date_rcp: Date réunion concertation pluridisciplinaire
• dn_date: Date dernières nouvelles ou dernier suivi
• date_deces: Date décès patient (seulement si décédé)
• date_1er_symptome: Date apparition premiers symptômes
• exam_radio_date_decouverte: Date premier examen découvrant la tumeur
• date_progression: Date récidive/progression
• survie_globale: Durée survie en mois
• infos_deces: Circonstances décès

Performance on Validation Set

Aggregates:

• Macro F1: 0.2458 (Precision: 0.2195, Recall: 0.6817)
• Micro F1: 0.3032 (Precision: 0.1861, Recall: 0.8171)

Per-Label Breakdowns:

Label	Precision	Recall	F1
date_chir	0.0615	1.0000	0.1159
date_rcp	0.7347	0.9231	0.8182
dn_date	0.0000	0.0000	0.0000
date_deces	0.0000	0.0000	0.0000
date_1er_symptome	0.0615	1.0000	0.1159
exam_radio_date_decouverte	0.0462	1.0000	0.0882
date_progression	0.0333	1.0000	0.0645
survie_globale	0.3714	0.8125	0.5098
infos_deces	0.6667	0.4000	0.5000

Usage

# Inference Code
from gliner2 import GLiNER2

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

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