UniversalDependencies / UD_Irish-IDT

Irish data
Other
6 stars 7 forks source link

xcomp:pred vs obl #153

Open ThFransen84 opened 1 year ago

ThFransen84 commented 1 year ago

I queried the issues but the following issue didn't come up; I hope I'm not duplicating anything. I'm also relatively new to UD ...

I'm unsure how to label prepositional phrases with lemma as the head. According to the documentation, instances like Níl fhios agam are obl:prep (https://universaldependencies.org/ga/dep/obl-prep.html), while e.g. Bhí saol ainnis aici ... (IUDT sent_id: 545) has xcomp:pred, which is, admittedly, quite rare with prep. pronouns, apart from ann "there" (which is explicitly listed in the documentation as being an example of xcomp:pred, see https://universaldependencies.org/ga/dep/xcomp-pred.html).

Instances with prep. phrases other than prep. pronouns seem more arbitrarily labeled with either dependency relation:

obl:

go raibh siad os coinne an phoill chearnaigh... (sent_id: 458), Ach beo bocht a bheadh ar an té... (sent_id: 460), cailín a bhíodh mar amhranaí... (sent_id: 461), bhí an-tóir ag na Francaigh ar chuile shórt... (2x) (sent_id: 468), Ní raibh eolas dá laghad ag an tiománaí (sent_id: 917), Tá stádas faoi leith ag an UDA sa tSeanchill (2x) (sent_id: 923), ceann de na buntáistí is luachmhaire a bhíonn ag pobal neamhspléach... (sent_id: 970), etc.

xcomp:pred:

agus atá siad faoi láthair... (sent_id: 918), atá ar siúl... (sent_id: 922), go bhfuil géarghá le Comhairleoir Gaeilge lánaimseartha... (sent_id: 947), Bhí an fear ag an doras. (sent_id: 1037), Tá an teach a bhí ag an mhinistéir... (sent_id: 1209), Tá caighdeáin dhif-riúla imeartha ag na hÍsiltírigh ná mar atá againn. (sent_id: 1796), etc. In the last example (which I obviously had to include here ;-)), we have coordinated arguments, although the first is xcomp:pred and the second obl:prep, which feels inconsistent.

There are instances with multiple arguments; in those cases I can see how one of the arguments of should be xcomp:pred, and the other obl(:prep).

Any help greatly appreciated.

tlynn747 commented 1 year ago

Will look into this next week... thanks for using the treebank and the opportunity to dig into the data :)

tlynn747 commented 1 year ago

Thanks for pointing these few inconsistencies out Dorus. Most of the issues you point out occur between sentences 1-1020, which was the original IDT that was converted to UD - and there's a higher chance that those issues arose during the conversion/ reviewing stage.

But overall the decisions aren't abritary.

Bí + preposition for ownership is obl If that ownership is denoted by a prepositonal pronoun, then the language specific label obl:prep is used. Bí + predicate in the form of verbal noun (progressive aspectual phrases) = xcomp Bí + predicate in the form of prepositional phrase (e.g. beidh an chóip sin ina fianaise) = xcomp:pred

Out of the examples you gave, the following are incorrect and:

Prepositions attached to nouns and verbal nouns are usually nmod

The labelling of ann is probably the most vague. We went with treating it as a predicate, but could possible be labelled as advmod.

ThFransen84 commented 1 year ago

Thanks, Teresa. A follow-up question: is + preposition for ownership restricted to ag? I'm thinking of cases like tá gruaig dhubh ar an gcailín / uirthi etc.