search

UniversalDependencies / UD_Irish-IDT

Irish data
Other
6 stars 7 forks source link

4400+ corrections to Form features #64

Closed kscanne closed 3 years ago

kscanne commented 3 years ago

Everything should be correct now. Note I have "mb'" tagged as Form=Ecl,VF. Also scanned everything for emphatic forms and those are all marked.

tlynn747 commented 3 years ago

I spotted a change of compound to nmod in the training file: "droim an domain". The guys would have labelled it as such because it's not fully compositional. So in terms of our work on MWE identification (Abigail's work in particular), I'd prefer to capture this MWE in some way, if you oppose compound (I know the dev can throw it off). I'm happy with fixed as an alternative, capturing the idiomatic sense. (ie it's not literally the back of the world).

It's the only compound - nmod change I can see as the diff file is so big! If there are any more, can you flag them?

I'll merge it now and the update can go into the next change push from either of us.

kscanne commented 3 years ago

Thanks for merging. I'll switch that one example back in my next PR and we can discuss the bigger issues elsewhere.

tlynn747 commented 3 years ago

Super. This was a massive help, thanks so much. Those noun features were a major stress.