raphael-r/bright-eds-tumor_location

BRIGHT NER: EDS-NLP (CamemBERT + CRF) fine-tuned for tumor_location

Description

This is a EDS-NLP (CamemBERT + CRF) architecture fine-tuned to extract clinical neuro-oncology entities related to the tumor_location 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):

• tumeur_lateralite: Latéralité (gauche, droite ou bilatérale)
• tumeur_position: Lobe ou structure cérébrale
• localisation_chir: Région cérébrale ciblée par la chirurgie

Performance on Validation Set

Aggregates:

• Macro F1: 0.9065 (Precision: 0.8307, Recall: 1.0000)
• Micro F1: 0.9075 (Precision: 0.8307, Recall: 1.0000)

Per-Label Breakdowns:

Label	Precision	Recall	F1
tumeur_lateralite	0.8816	1.0000	0.9371
tumeur_position	0.8598	1.0000	0.9246
localisation_chir	0.7508	1.0000	0.8577

Usage

# Inference Code
import edsnlp

nlp = edsnlp.load("raphael-r/bright-eds-tumor_location")
doc = nlp("Patient presenting with epileptic seizures...")

for ent in doc.ents:
    print(ent.text, "=>", ent.label_)