Bad grouping of relations in relations.html? about docs HOT 11 CLOSED

The table is intended to replicate (excepting layout details) Table 1 from de Marneffe et al. Universal Stanford Dependencies: A cross-linguistic typology LREC'14 (http://www.lrec-conf.org/proceedings/lrec2014/pdf/1062_Paper.pdf).

nmod is found (also) under "Non-core dependents of clausal predicates" in the paper as well as on http://universaldependencies.github.io/docs/relations.html . The other place where nmod occurs (on both) is under "Noun dependents".

As I haven't studied this structuring of the relation types much, I don't think I can help explain the grouping. @mcdm : any comments / clarification regarding this categorization would be much appreciated!

from docs.

Oh, now I see the other occurrence of nmod. Without making this clear, the table is a little confusing for someone looking at it for the first time (just tested on myself), but we should probably discuss this in the e-mail group instead of here.

Thanks
Dan

----- Original Message -----

From: "spyysalo" [email protected]
To: "UniversalDependencies/docs" [email protected]
Cc: "Dan Zeman" [email protected]
Sent: Wednesday, 10 September, 2014 8:52:01 AM
Subject: Re: [docs] Bad grouping of relations in relations.html? (#24)

The table is intended to replicate (excepting layout details) Table 1 from de
Marneffe et al. Universal Stanford Dependencies: A cross-linguistic typology
LREC'14 ( http://www.lrec-conf.org/proceedings/lrec2014/pdf/1062_Paper.pdf
).

nmod is found (also) under "Non-core dependents of clausal predicates" in the
paper as well as on
http://universaldependencies.github.io/docs/relations.html . The other place
where nmod occurs (on both) is under "Noun dependents".

As I haven't studied this structuring of the relation types much, I don't
think I can help explain the grouping. @mcdm : any comments / clarification
regarding this categorization would be much appreciated!

—
Reply to this email directly or view it on GitHub .

from docs.

The text "Note: nmod, ncmod, nfincl, and neg appear in two places." does appear below the table (again, copied from the paper), but I agree that it's not all that noticeable.

(I'm actually much in favor of using this system rather than email to discuss detailed aspects of the documentation whenever possible. There are many potential improvements that don't really merit writing 10+ people in each case, and there's no way to keep track of many small(er) issues over email.)

from docs.

The text "Note: nmod, ncmod, nfincl, and neg appear in two places." does
appear below the table (again, copied from the paper), but I agree that it's
not all that noticeable.
I would try to put it above the table instead, although I can no more test the noticeability on myself :-) But it is probably in the mysterious file ud-table.html that I failed to find.

(I'm actually much in favor of using this system rather than email to discuss
detailed aspects of the documentation whenever possible. There are many
potential improvements that don't really merit writing 10+ people in each
case, and there's no way to keep track of many small(er) issues over email.)
True, except that I thought it would involve redesigning the table (which I do not think any more now) and then it should be also discussed with Chris, Marie and others. And I suspect that the only person regularly monitoring the issue tracker is you :-)

from docs.

it is probably in the mysterious file ud-table.html that I failed to find.

You're right. The source is _includes/ud-table.html. (Anything appearing in an {% include ... %} will be found in _includes/.) We used the include mechanism for this as the table was needed in multiple documents (#12).

I suspect that the only person regularly monitoring the issue tracker is you :-)

I suspect you're right about that :-) This is why I wrote @mcdm above; this should trigger notifications on this issue for Marie (unless she has switched this feature off). We could also try pinging at least Chris using this.

Not sure how many of the contributors have (and use) GitHub accounts at the moment, but I think with a bit of encouragement this might well work out. (It's required anyway to be able to edit.)

from docs.

I have modified both the table and the description pages of the four labels affected. In my view it is not confusing now.

from docs.

Agreed, this is an improvement! (An asterisk on the affected labels might have done the job also.)

from docs.

Looks good to me too! And yes I do get the emails ;-) But all this conversation happened between 2 am and 5am my time when I was sleeping!

from docs.

@mcdm : great, thanks! (Didn't mean in any way to imply you'd need to be on call for this project 24/7 ;-))

from docs.

The grouping is intended to emphasize

• the kind of governors (so you have dependents of clausal predicates, of nouns, etc.)
• the kind of dependents (either nominal dependent, predicate dependent, etc.)
  Does that help?

from docs.

Oh yes, it perfectly makes sense. It was just that I first encountered nmod in the "dependents of clausal predicates" group, then clicked on it and saw example phrase that did not belong to this group. I think it is fixed now, at least from my perspective.

from docs.