Arabic Named Entity Recognition Model

Pretrained BERT-based (arabic-bert-base) Named Entity Recognition model for Arabic.

The pre-trained model can recognize the following entities:

  1. PERSON
  1. ORGANIZATION
  1. LOCATION
  1. DATE
  1. PRODUCT
  1. COMPETITION
  1. PRIZE
  1. EVENT
  1. DISEASE

Example

Find here a complete example to use this model

Training Corpus

The training corpus is made of 378.000 tokens (14.000 sentences) collected from the Web and annotated manually. It can be obtained from here: https://www.textorch.com/datasets/arabic-named-entity-recognition-corpus

Results

The results on a valid corpus made of 30.000 tokens shows an F-measure of ~87%.