wanyun1903 / NER

REST API for Named Entity Recognition (NER)

Pre-requisites

Python3

Framework:

• Flask
• Docker

Pretrained NER model

• Spacy

Database:

• PostgresSQL (docker image)

Running the flask application by building a new image with the docker file on the local machine

#The directory 
cd NER

# To build the image
docker-compose up -d --build

# To initiate the database
docker-compose exec web python manage.py create_db

The applications will be running on docker_IP_Address:5000.

Example:

http://192.168.99.100:5000/

Application Routes

• (/): index.html. The main page with the relevant buttons to upload the json file and to retrieve the count of the different entities
• (/uploader): Post request - that will take in a file (.json), of which the file can consist of either a single or multiple news articles that have been already processed and identify the entities within each of the news article.

Sample input: can be found in NER/sample_input/sample_news_10.json. 
Sample output:  {"0": "[{"row":0,"text":"Meyer Handelman Co. Increases Position","label":"ORG"}]}. row refers to the index of the news. 

• (/retrieveCount): Get request - that will retrieve count of the different entities of each unique news article from a PostgreSQL database (docker image).

Sample output:
[{"label":"CARDINAL","count":22,"row":0},{"label":"DATE","count":34,"row":0}]. row refers to the index of the news. 

Credit

• Susan Li - https://towardsdatascience.com/building-a-flask-api-to-automatically-extract-named-entities-using-spacy-2fd3f54ebbc6
• Michael Herman - https://testdriven.io/blog/dockerizing-flask-with-postgres-gunicorn-and-nginx/