Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

#

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Jump to bottom

Better Universal Dependencies support #10

Open
dmcc opened this issue May 5, 2015 · 0 comments
Open

Better Universal Dependencies support #10

dmcc opened this issue May 5, 2015 · 0 comments
Labels
enhancement

Comments

@dmcc
Copy link
Owner

dmcc commented May 5, 2015

This would involve at least the following:

  1. Add the -originalDependencies option for both backends.
  2. Find a way to download the feature mapping and include it in the classpath. It's included in the giant models jar files, so we could include those, but it seems overkill to download these if we can avoid it.
  3. Populate the features field with features from universal dependencies (requires 2.)
  4. Map the POS tags to their Universal counterparts.
@dmcc dmcc mentioned this issue Jul 29, 2015
Closed
dmcc added a commit that referenced this issue Jul 30, 2015
@dmcc
Improve Universal Dependencies support
54c84be 
Addresses (most of) issue #10
(does not handle the feature mappings currently)

StanfordDependencies: new "universal" parameter to convert_tree()
which outputs UD if available (otherwise outputs SD with potentially no
warnings since SubprocessBackend can't know from an arbitrary jar
whether it is new enough to support UD). To match CoreNLP behavior,
universal defaults to True. Added and updated documentation.

JPypeBackend: checks for UD support and enables it if requested and available (unlike SubprocessBackend, this gives a warning if not available)

SubprocessBackend: minimal updates to support UD

Testing code: String data was (finally) pulled out of
test_stanforddependencies and moved to test/data.py.
The strings needed to be generalized to include UD variants
of the output, so a class structure is used. We now test
SD and UD versions of both backends by extending the
DefaultBackendTest hierarchy.
# for free to join this conversation on GitHub. Already have an account? # to comment
Assignees
No one assigned
Labels
enhancement
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@dmcc