antoinebriand/corenlp-server repository overview

⁠Core NLP Service

CoreNLP Service is a Docker image that runs the CoreNLP Server.

⁠Getting started

To use this image, you need to have Docker⁠ installed. Once you have a running docker instance, you can use one of the prebuilt images from the docker hub⁠.

For exemple you can use the latest version with this command :

docker run -p 9000:9000 antoinebriand/corenlp-server:latest

This will start the server on the local 9000 port of your machine. For production use, you should avoid using the latest tag.

⁠Settings

⁠RAM

The server will be deployed with 1Gb of RAM by default. To update this value, you can pass an environment variable on start:

#Start the sever with 4Gb of RAM
docker run -e CORENLP_MAX_MEMORY=4g -it antoinebriand/corenlp-server:latest

⁠Annotators

By default, the server will be started with the following annotators : tokenize,ssplit,ner.You can override this behavior by initializing the CORENLP_ANNOTATORS env var :

docker run -e CORENLP_MAX_MEMORY=4g -e CORENLP_ANNOTATORS="tokenize,ssplit,pos,ner" -it antoinebriand/corenlp-server:latest

⁠Deploy the image

Since this image is stateless (no databases), it can be ran in a cluster as a microservice. If you have a running
Docker Swarm cluster you can create a service this way :

docker service create --name corenlp \
--publish 9000:9000 \
--env CORENLP_MAX_MEMORY=4g \
--env CORENLP_ANNOTATORS="tokenize,ssplit,pos,ner"\
antoinebriand/corenlp-server:latest

This will start a service on your cluster that expose Stanford on port 9000 with 4Gb of Ram. Since you are using Swarm you can also scale this service in order to take advantage of the built in load-balancing of Swarm :

#Start 10 instances of the server
docker service scale corenlp=10

Stopping the service is done with this command :

docker service rm corenlp