Skip to content

Latest commit

 

History

History
68 lines (46 loc) · 1.58 KB

README.md

File metadata and controls

68 lines (46 loc) · 1.58 KB
Raw

Python Coreference Resolver Service

Our service-based alternative to Coreference Resolution.

Python Interpreter

Python 3.6

Python <=3.6 don't support variable type hinting. The code relies on Python 3.6.x.

The docker container is using 3.6.1.

Running locally

Install dependencies, best inside a virtual environment.

pip install -r requirements.txt
pip install uwsgi

This will start the service locally bound to 0.0.0.0:5128.

uwsgi --ini uwsgi.ini

You must have a CoreNLP server running somewhere. You can configure it with the environmental variable PYCOBALT_CORENLP=http://localhost:9000. If you are not using our docker image which includes the required models, you must have the following models in the server's classpath:

  • stanford-corenlp-models-current.jar
  • stanford-english-corenlp-models-current.jar
  • stanford-english-kbp-corenlp-models-current.jar

It is using CoreNLP version 3.7.0.

Running within a Docker Container

The service is wrapped in the docker-compose.yml file which also includes an instance of CoreNLP.

docker-compose up

Testing

curl -X POST \
 -H "Content-Type: application/json" \
 -H "Accept: application/json" \
 -d '{"text": "Donald Trump is the president of USA. He is a business man."}' \
 "http://localhost:5128/resolve/text"

Expected output:

{
  "text": "Donald Trump is the president of USA. Donal Trump is a business man.",
}

Contributors (alphabetical order)

  • Andre Freitas
  • Bernhard Bermeitinger
  • Siegfried Handschuh