Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
What happened to displacy? #426
@nathanathan @lx5491 @arvindsg @aliabbasjp Facing the same issue :( so decided to write an alternative for until the full Displacy comes back. At the moment it's a simple Flask application (https://github.com/sderygithub/NLTree) you'll have to run locally.. I may just push it on my server in a few days and host it. You can see on my website what it currently looks like (http://dery.xyz/projects/nltree/overview.html)
The displaCy and sense2vec demos should now be back online: https://spacy.io/demos
They're both running on the same server, and there may be some short term instability. Fingers crossed! The demos aren't hugely resource intensive, so in theory everything should be fine; but it's still a slightly rickety arrangement.
The good news is that both demos will be open-sourced. I hope to flip the repos to public tonight. Documentation will be a bit sparse, but they should be reasonably easy to host. If anyone is able to donate some compute resources, please get in touch: firstname.lastname@example.org.
Unfortunately I'm away for the next week (timing...), but I'll do my best to keep things up. After I get back, the updates should be coming quite quickly.
Yay, so nice to see displaCy back online!
A little note on the displaCy front-end code: I developed the core of this over a year ago with a pretty specific (and now fairly outdated) concept in mind. It also hasn't been maintained for months, so there's definitely room for improvement.
I've already started re-writing the basic visualiser to switch it over to SVGs and use a more logical JSON structure to describe the words, arcs and relations (because to be honest, the current one is kinda messy). Using SVGs will also make easier to export the graphics and embed them in documents and blog posts. At the moment, I'm thinking about moving more towards a cleaner and more reliable visualiser and not focusing too much on the rather abstract manual annotation mode.
But I'm curious to hear your suggestions and feature requests!