Named Entity Recognition API with Flair

You can use it as a rest engine on Tock. More info :

https://github.com/theopenconversationkit/tock

This API has this endpoint :

POST /api/v1/parse

{
  "language":"fr",
  "text": "Mon nom est Jo Hisahi"
}

The 200 response would answer :

{
  "entities": [
    {
      "entity": {
        "start": 12,
        "end": 21,
        "type": {
          "name": "flair:person"
        }
      },
      "probability": 0.9251298308372498
  ]
}

GET /api/v1/entities

The 200 response would answer :

[
  "flair:person",
  "flair:location",
  "flair:organization"
]

Using Flair library on v1 for NER.

Contribute

First, use virtualenv venv then source venv/bin/activate and pip install -r requirements/dev.txt for development.

Download the models with the script in build/download_models.sh.

Start the WSGI with FLASK_ENV=development python application.py in development mode.

Docker

Dockerfile provided for deployment.

Build : docker build -t name .

Run local : docker run -p 8080:5000 name

Amazon

See https://aws.amazon.com/fr/blogs/devops/dockerizing-a-python-web-app/ for a FLask deployment on EB.