Closed wfaithfull closed 5 years ago
Dockerfile for NLP here: https://github.com/NLPbox/stanford-corenlp-docker
Excellent, thought there would be one but that saves me looking around.
NLP no longer required
Yep, coreNLP has been removed now as of #5 (PR #6). This should be good as a standalone docker image now.
I've setup a repo on docker hub that builds an image from master branch. I'm thinking of setting up a really light http server on the linode that docker hub can ping every time a new image is built.
@jon1010hill sounds sensible. We could even just run an NGINX to delegate to a shell script like this.
or circleci
I'm making good progress with CircleCI. Seems to do what we need, should finish it this weekend.
We need to wrap this application into a dockerfile. In my opinion, this dockerfile should exclude stanford coreNLP server. What we need then is:
That way, we can run a single Stanford CoreNLP image on our server (which is a very heavy application), and we can potentially run lots of SiteGeist instances all pointing at the same NLP server. Since each SiteGeist application is only concerned with one geofence, if we want to cover multiple areas we need to run many of them.
This limitation is by design - the task of dynamically setting the geofence requires us to have knowledge of all tweets, because we can't know their geographic location in advance. It is a task that would require a lot of Twitter API access and a lot of heavy lifting.