A Fine-grained NER model for extracting benefit information from French benefit booklets

Koverhoop

FG-NER

Fine-Grained Named Entity Recognition, a stack of models to infer benefit details from lengthy Insurance Contracts and Booklets. The models are based on the French language, achieving a Coarse-Grained Entity Accuracy of 100%, and the following scores for Fine-Grained Entity Recognition:

F-score for FG models for each CG type:

  1. Amounts: 94%

  2. Percentages: 83%

  3. Ages: 81%

  4. Duration: 82%