home edit page issue tracker

Uppsala Group on Clitics and Reflexive Pronouns

(Kiril Simov, Petya Osenova, Simonetta Montemagni, Mojgan Seraji, Nizar Habash, Marie-Catherine de Marneffe)

Clitic doubling

In languages where real clitic doubling can appear (such as Bulgarian or Greek), the clitic elements should be tagged as expl and depend on the predicate:

edit

  

    
input
    
brat
    
info
  
I will him it read the book to the boy
expl(read, him)
expl(read, it)
dobj(read, book)
nmod(read, boy)
{
    "text": "I will him it read the book to the boy ",
    "entities": [
        [ "embedded-1-T1", "token", [ [ 0, 1 ] ] ],
        [ "embedded-1-T2", "token", [ [ 2, 6 ] ] ],
        [ "embedded-1-T3", "token", [ [ 7, 10 ] ] ],
        [ "embedded-1-T4", "token", [ [ 11, 13 ] ] ],
        [ "embedded-1-T5", "token", [ [ 14, 18 ] ] ],
        [ "embedded-1-T6", "token", [ [ 19, 22 ] ] ],
        [ "embedded-1-T7", "token", [ [ 23, 27 ] ] ],
        [ "embedded-1-T8", "token", [ [ 28, 30 ] ] ],
        [ "embedded-1-T9", "token", [ [ 31, 34 ] ] ],
        [ "embedded-1-T10", "token", [ [ 35, 38 ] ] ]
    ],
    "relations": [
        [ "embedded-1-R1", "expl", [ [ "arg1", "embedded-1-T5" ], [ "arg2", "embedded-1-T3" ] ] ],
        [ "embedded-1-R2", "expl", [ [ "arg1", "embedded-1-T5" ], [ "arg2", "embedded-1-T4" ] ] ],
        [ "embedded-1-R3", "dobj", [ [ "arg1", "embedded-1-T5" ], [ "arg2", "embedded-1-T7" ] ] ],
        [ "embedded-1-R4", "nmod", [ [ "arg1", "embedded-1-T5" ], [ "arg2", "embedded-1-T10" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "1" ]
}

When the direct or indirect objects are realized with a clitic only, there are tagged as dobj or iobj.

We suggest that if needed, languages can have subtypes indicating the relation (dobj, iobj), though it is unclear whether that is really the best idea.

For languages with dislocation, such as French, we use the dislocated relation, consistent with the existing guidelines.

Clitic acting as part of the verbal inflection

If the pronoun has a real grammatical function (dobj, iobj, advmod), we use that function as relation. If not, we use expl, possibly with the following subtypes.

edit

  

    
input
    
brat
    
info
  
Maria si lava \n Maria washes herself
dobj(lava, si)
{
    "text": "Maria si lava \n Maria washes herself ",
    "entities": [
        [ "embedded-2-T1", "token", [ [ 0, 5 ] ] ],
        [ "embedded-2-T2", "token", [ [ 6, 8 ] ] ],
        [ "embedded-2-T3", "token", [ [ 9, 13 ] ] ],
        [ "embedded-2-T5", "token", [ [ 16, 21 ] ] ],
        [ "embedded-2-T6", "token", [ [ 22, 28 ] ] ],
        [ "embedded-2-T7", "token", [ [ 29, 36 ] ] ]
    ],
    "relations": [
        [ "embedded-2-R1", "dobj", [ [ "arg1", "embedded-2-T3" ], [ "arg2", "embedded-2-T2" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "2" ]
}

“pv”: pronominal verb

edit

  

    
input
    
brat
    
info
  
la finestra si è rotta \n the window broke
expl:pv(rotta, si)
{
    "text": "la finestra si è rotta \n the window broke ",
    "entities": [
        [ "embedded-3-T1", "token", [ [ 0, 2 ] ] ],
        [ "embedded-3-T2", "token", [ [ 3, 11 ] ] ],
        [ "embedded-3-T3", "token", [ [ 12, 14 ] ] ],
        [ "embedded-3-T4", "token", [ [ 15, 16 ] ] ],
        [ "embedded-3-T5", "token", [ [ 17, 22 ] ] ],
        [ "embedded-3-T7", "token", [ [ 25, 28 ] ] ],
        [ "embedded-3-T8", "token", [ [ 29, 35 ] ] ],
        [ "embedded-3-T9", "token", [ [ 36, 41 ] ] ]
    ],
    "relations": [
        [ "embedded-3-R1", "expl:pv", [ [ "arg1", "embedded-3-T5" ], [ "arg2", "embedded-3-T3" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "3" ]
}

“pass”: passivazing uses

edit

  

    
input
    
brat
    
info
  
la neve si vede dal balcone \n Snow can be seen from the balcony
expl:pass(vede, si)
{
    "text": "la neve si vede dal balcone \n Snow can be seen from the balcony ",
    "entities": [
        [ "embedded-4-T1", "token", [ [ 0, 2 ] ] ],
        [ "embedded-4-T2", "token", [ [ 3, 7 ] ] ],
        [ "embedded-4-T3", "token", [ [ 8, 10 ] ] ],
        [ "embedded-4-T4", "token", [ [ 11, 15 ] ] ],
        [ "embedded-4-T5", "token", [ [ 16, 19 ] ] ],
        [ "embedded-4-T6", "token", [ [ 20, 27 ] ] ],
        [ "embedded-4-T8", "token", [ [ 30, 34 ] ] ],
        [ "embedded-4-T9", "token", [ [ 35, 38 ] ] ],
        [ "embedded-4-T10", "token", [ [ 39, 41 ] ] ],
        [ "embedded-4-T11", "token", [ [ 42, 46 ] ] ],
        [ "embedded-4-T12", "token", [ [ 47, 51 ] ] ],
        [ "embedded-4-T13", "token", [ [ 52, 55 ] ] ],
        [ "embedded-4-T14", "token", [ [ 56, 63 ] ] ]
    ],
    "relations": [
        [ "embedded-4-R1", "expl:pass", [ [ "arg1", "embedded-4-T4" ], [ "arg2", "embedded-4-T3" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "4" ]
}

“impers”: impersonal construction

edit

  

    
input
    
brat
    
info
  
qui si parla francese  \n here French is spoken
expl:impers(parla, si)
{
    "text": "qui si parla francese \n here French is spoken ",
    "entities": [
        [ "embedded-5-T1", "token", [ [ 0, 3 ] ] ],
        [ "embedded-5-T2", "token", [ [ 4, 6 ] ] ],
        [ "embedded-5-T3", "token", [ [ 7, 12 ] ] ],
        [ "embedded-5-T4", "token", [ [ 13, 21 ] ] ],
        [ "embedded-5-T6", "token", [ [ 24, 28 ] ] ],
        [ "embedded-5-T7", "token", [ [ 29, 35 ] ] ],
        [ "embedded-5-T8", "token", [ [ 36, 38 ] ] ],
        [ "embedded-5-T9", "token", [ [ 39, 45 ] ] ]
    ],
    "relations": [
        [ "embedded-5-R1", "expl:impers", [ [ "arg1", "embedded-5-T3" ], [ "arg2", "embedded-5-T2" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "5" ]
}

Existential constructions

For existential constructions, expl is used, consistent with the existing guidelines.

© 2014–2024 Universal Dependencies contributors. Site powered by Annodoc and brat

.