Dependencies
Note: This document is a draft. Not all relations have been converted to USD-compatible description yet.
Note: nmod, neg and punct appear in two places.
|
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||
acl: clausal modifier of noun
acl stands for finite and non-finite clauses that modify a noun, in
contrast to the advcl relation which is used for adverbial clauses
that modify a predicate. The head of the acl relation is the noun
that is modified, and the dependent is the head of the clause that
modifies the noun.
These modifiers include infinitive and participial modifiers (correspond to infmod and partmod in the original Turku Dependency Treebank). Instead, the third possible type of clausal modifiers of nouns, relative clause modifier (acl:relcl), is defined as a subtype of acl.
Minulla oli lupa mennä ulos . \n I had permission to_go out .
nmod:own(oli-2, Minulla-1)
nsubj(oli-2, lupa-3)
acl(lupa-3, mennä-4)
advmod(mennä-4, ulos-5)
punct(oli-2, .-6)
The participial modifier is a participle verb which modifies a noun phrase. Note that the participle can take arguments, for instance a subject, just as any verb. (Also the MA-derivation is treated as a participle in UD Finnish.)
Äidin leipoma kakku oli menestys . \n Mother baked_by cake was success .
nsubj(leipoma-2, Äidin-1)
acl(kakku-3, leipoma-2)
nsubj:cop(menestys-5, kakku-3)
cop(menestys-5, oli-4)
punct(menestys-5, .-6)
Saadut lahjat ilahduttivat lapsia . \n Received presents made_happy children .
acl(lahjat-2, Saadut-1)
nsubj(ilahduttivat-3, lahjat-2)
dobj(ilahduttivat-3, lapsia-4)
punct(ilahduttivat-3, .-5)
Occasionally, participial verb forms can modify a verb as well. These uses include cases that are clearly modifiers, as well as some more complement-like situations. In the complement-like situations, one of the clausal complement types (ccomp, xcomp or xcomp:ds) should be used, whereas modifiers are marked as adverbial clause modifiers (advcl).
Huolestuneena juoksin hänen luokseen . \n Worried I_ran him to .
advcl(juoksin-2, Huolestuneena-1)
nmod(juoksin-2, hänen-3)
case(hänen-3, luokseen-4)
punct(juoksin-2, .-5)
Ksylitoli osoittautui kariesta ehkäiseväksi . \n Xylitol turned_out karies preventing .
nsubj(osoittautui-2, Ksylitoli-1)
xcomp(osoittautui-2, ehkäiseväksi-4)
dobj(ehkäiseväksi-4, kariesta-3)
punct(osoittautui-2, .-5)
Diffs
Turku Dependency Treebank
We do not attempt to distinguish modifiers including secondary predication from other type of modifiers. Therefore, optional predicative like modifiers are attached to the main verb with one of the modifier relations.
acl:relcl: relative clause modifier
A relative clause modifier (acl:relcl) marks relative clauses. The
governor is the phrase or clause modified. Usually, it is a noun, but in Finnish it can also be a verb, when the dependent refers to the entire action described. The dependent is the main
predicate of the relative clause.
Mies , jonka hän oli nähnyt eilen , oli taas ovella . \n The_man , whom he had seen yesterday , was again at_the_door .
acl:relcl(Mies-1, nähnyt-6)
punct(nähnyt-6, ,-2)
dobj(nähnyt-6, jonka-3)
nsubj(nähnyt-6, hän-4)
aux(nähnyt-6, oli-5)
advmod(nähnyt-6, eilen-7)
punct(nähnyt-6, ,-8)
nsubj(oli-9, Mies-1)
advmod(oli-9, taas-10)
nmod(oli-9, ovella-11)
punct(oli-9, .-12)
Ovi kolahti auki , mikä säikäytti lapsen . \n The_door clanked open , which scared the_child .
nsubj(kolahti-2, Ovi-1)
advmod(kolahti-2, auki-3)
acl:relcl(kolahti-2, säikäytti-6)
punct(säikäytti-6, ,-4)
nsubj(säikäytti-6, mikä-5)
dobj(säikäytti-6, lapsen-7)
punct(kolahti-2, .-8)
Diffs
Turku Dependency Treebank
The governor can also be a verb, when the dependent refers to the entire action described in the main sentence.
FinnTreeBank
FI_FTB applies the universal relation acl instead
of the language-specific relation acl:relcl.
advcl: adverbial clause modifier
Adverbial clause modifiers (advcl) are subordinate clauses that
are not complements. Also non-complement infinitival or temporal clauses
(lauseenvastike, see for instance
(see ISK §876) and non-complement participles modifying verbs are
marked as advcl. If there is a subordinating conjunction present, it
is marked with the dependency type mark.
Kun äiti tuli kotiin , isä keitti kahvia . \n When mother came home , father made coffee .
mark(tuli-3, Kun-1)
nsubj(tuli-3, äiti-2)
nmod(tuli-3, kotiin-4)
punct(tuli-3, ,-5)
nsubj(keitti-7, isä-6)
advcl(keitti-7, tuli-3)
dobj(keitti-7, kahvia-8)
punct(keitti-7, .-9)
Äidin tullessa kotiin isä keitti kahvia . \n Mother when_came home father made coffee .
nsubj(tullessa-2, Äidin-1)
nmod(tullessa-2, kotiin-3)
nsubj(keitti-5, isä-4)
advcl(keitti-5, tullessa-2)
dobj(keitti-5, kahvia-6)
punct(keitti-5, .-7)
Huolestuneena juoksin hänen luokseen . \n Worried I_ran him to .
advcl(juoksin-2, Huolestuneena-1)
nmod(juoksin-2, hänen-3)
case(hänen-3, luokseen-4)
punct(juoksin-2, .-5)
The dependency type advcl is used also in comparative
constructions, most often involving adjectives in the comparative
form. The head of the advcl dependency is the comparative
wordform, and the dependent is the compared element.
The annotation of comparative and superlative structures is described in Comparatives and superlatives.
parempi kuin eilinen näytelmä \n better than yesterday's(adj.) play
advcl(parempi-1, näytelmä-4)
mark(näytelmä-4, kuin-2)
amod(näytelmä-4, eilinen-3)
References
advmod: adverb modifier
The dependency type advmod is used for adverb modifiers of verbs,
nominals and adverbs alike.
Hän käveli kotiin hitaasti . \n He walked home slowly .
nsubj(käveli-2, Hän-1)
nmod(käveli-2, kotiin-3)
advmod(käveli-2, hitaasti-4)
punct(käveli-2, .-5)
Minä otin kaapista myös vasaran . \n I took from_closet also hammer .
nsubj(otin-2, Minä-1)
nmod(otin-2, kaapista-3)
dobj(otin-2, vasaran-5)
advmod(vasaran-5, myös-4)
punct(otin-2, .-6)
Also quantification modifiers are annotated as adverb modifiers in UD Finnish (correspond to quantmod in the original Stanford Dependencies and the Turku Dependency Treebank). Quantification modifiers are quantifiers that modify a numerical expression. Typically quantifiers are adverbs, but also few adjectives are allowed as quantifiers.
Alue oli suuruudeltaan noin kymmenen neliökilometriä . \n The_area was of_its_size about ten square_kilometres .
nsubj:cop(neliökilometriä-6, Alue-1)
cop(neliökilometriä-6, oli-2)
nmod(neliökilometriä-6, suuruudeltaan-3)
advmod(kymmenen-5, noin-4)
nummod(neliökilometriä-6, kymmenen-5)
punct(neliökilometriä-6, .-7)
amod: adjectival modifier
Nouns may take adjectival modifiers, which are marked with the
dependency type amod. It is also possible for an adjective to take
another adjective as a modifier. (These adjectival modifiers are
generally expressed with -ly adverbs in English.)
Tien vieressä on suuri kivi . \n Road next_to is large rock .
nmod(on-3, Tien-1)
adpos(Tien-1, vieressä-2)
nsubj(on-3, kivi-5)
amod(kivi-5, suuri-4)
punct(on-3, .-6)
poikkeuksellisen suuri kivi \n exceptional(ly) large rock
amod(suuri-2, poikkeuksellisen-1)
amod(kivi-3, suuri-2)
appos: apposition
An apposition (appos) is a grammaticalized, paradigmatic addition
(usually a noun phrase), which has the same referent as its head word,
and the same grammatical function
(see ISK §1059).
Also structures with a supporting noun
(see ISK §567) are
considered appositional. Appositional structures and the closely
related appellation modifiers are discussed in detail in
the document on
specific syntactic constructions in Finnish.
Professori , Matti Tamminen , luennoi tänään . \n The_professor , Matti Tamminen , lectures today .
appos(Professori-1, Tamminen-4)
punct(Tamminen-4, ,-2)
punct(Tamminen-4, ,-5)
name(Tamminen-4, Matti-3)
nsubj(luennoi-6, Professori-1)
advmod(luennoi-6, tänään-7)
punct(luennoi-6, .-8)
Kirjassa Putkinotko on mielenkiintoinen juoni . \n In_the_book Putkinotko is interesting plot .
appos(Kirjassa-1, Putkinotko-2)
nmod(on-3, Kirjassa-1)
nsubj(on-3, juoni-5)
amod(juoni-5, mielenkiintoinen-4)
punct(on-3, .-6)
aux: auxiliary
In UD Finnish, only a closed list of verbs can act as auxiliaries, including the main modal verbs (see ISK §1562) and in addition the verbs olla “to be” and aikoa “to be going to”.
The full list of auxiliaries is as follows:
- täytyä “must”
- pitää “have to”
- tarvita “need”
- joutua “have to”
- voida “be able to, can”
- saattaa “may”
- taitaa “be+probably, may”
- mahtaa “be+probably, may”
- olla “be”
- aikoa “be going to”
Examples
Hän saattoi lähteä jo . \n He may(impf.) leave already .
nsubj(lähteä-3, Hän-1)
aux(lähteä-3, saattoi-2)
advmod(lähteä-3, jo-4)
punct(lähteä-3, .-5)
References
- http://scripta.kotus.fi/visk/sisallys.php?p=1562 (in Finnish)
Diffs
In FinnTreeBank (FI_FTB), a wide list of verbs which are
modal or otherwise abstract by nature have been counted
among the auxiliaries. Also a limited set of verbal
compound constructions has been annotated using the
relation aux. See the full lists in
FTB Annotation Manual: 16.2 Auxiliary verb.
auxpass: passive auxiliary
The only passive auxiliary (auxpass) in Finnish is olla (to
be). An auxiliary is only considered a passive auxiliary if the main
verb is in passive, not if only the auxiliary is in passive. In the
latter case the auxiliary is marked as a non-passive auxiliary,
aux.
Suunnitelmaan on tehty muutoksia . \n Into_the_plan have_been made changes .
nmod(tehty-3, Suunnitelmaan-1)
auxpass(tehty-3, on-2)
dobj(tehty-3, muutoksia-4)
punct(tehty-3, .-5)
Suunnitelmaan voidaan tehdä muutoksia . \n Into_the_plan can_be made(1st_inf.) changes .
nmod(tehdä-3, Suunnitelmaan-1)
aux(tehdä-3, voidaan-2)
dobj(tehdä-3, muutoksia-4)
punct(tehdä-3, .-5)
Diffs
FinnTreeBank (FI_FTB) does not use the auxpass relation,
but uses aux instead.
case: case marking
The dependency type case is used for the adposition in pre- and
postpositional phrases. The head of
an adpositional phrase is the nominal, not the adposition, so as to
analyze adpositional phrases similarly to nominal modifiers without an
adposition. (Such nominal modifiers are frequent in Finnish, as cases
are often used for the same purpose as adpositions.) To the same end,
the type case is used in combination with the type nmod, which is
also used for nominal modifiers when no adposition is present (see
nmod).
Talo sijaitsee mäen takana . \n House is_located hill behind .
nsubj(sijaitsee-2, Talo-1)
nmod(sijaitsee-2, mäen-3)
case(mäen-3, takana-4)
punct(sijaitsee-2, .-5)
cc: coordinating conjunction
Coordinating conjunctions are marked as dependents of the first
coordinated element, and the dependency type used is cc.
kirjat , kynät ja viivottimet \n books , pencils and rulers
conj(kirjat-1, kynät-3)
punct(kirjat-1, ,-2)
cc(kirjat-1, ja-4)
conj(kirjat-1, viivottimet-5)
cc:preconj: preconjunct
The preconjunct (cc:preconj) dependency relation marks the
first part of those two-part coordinating conjunctions where the
two parts are separated by coordinated elements.
See also CONJ.
sekä talo että piha \n both house and yard
cc:preconj(talo-2, sekä-1)
cc(talo-2, että-3)
conj(talo-2, piha-4)
Such two-part coordinating conjunctions are:
- joko … tai “either … or”
- milloin … milloin “when … when”
- mitä … sitä “the … the”
- niin … kuin “as well as”
- paitsi … myös “not only … but also”
- sekä … että “both … and”
- sitä … mitä “the … the”
- sitä mukaa … mitä “the … the” (alternative wording)
- toisaalta … ja toisaalta “on the one hand … and on the other hand”
- toisaalta … mutta toisaalta “on the one hand … but on the other hand”
- toisaalta … toisaalta “on the one hand … on the other hand”
- vuoroin … vuoroin “in turn … in turn”
- yhtä lailla … kuin (+kin) “as much … as also”
ccomp: clausal complement
Clausal complement is an object like clausal dependent. The governor is most commonly, although not always, the main verb or predicative of the main clause, and the dependent is the main verb or predicative of the dependent clause. The clausal complement can also modify a word other than a verb, most often a noun or pronoun. Most commonly clausal complements are että-clauses.
If the subject of the clausal complement is controlled (that is, must be the same as the higher subject or object, with no other possible interpretation) the appropriate relation is either xcomp or xcomp:ds.
Sanoin hänelle , että vesi kiehuu . \n I_said to_him , that water boils .
nmod(Sanoin-1, hänelle-2)
ccomp(Sanoin-1, kiehuu-6)
punct(kiehuu-6, ,-3)
mark(kiehuu-6, että-4)
nsubj(kiehuu-6, vesi-5)
punct(Sanoin-1, .-7)
Se , että naapurin koira haukkui kovasti , suututti äitiä . \n It , that neighbor's dog barked loudly , made_angry mother .
ccomp(Se-1, haukkui-6)
punct(haukkui-6, ,-2)
punct(haukkui-6, ,-8)
mark(haukkui-6, että-3)
nmod:poss(koira-5, naapurin-4)
nsubj(haukkui-6, koira-5)
advmod(haukkui-6, kovasti-7)
nsubj(suututti-9, Se-1)
dobj(suututti-9, äitiä-10)
punct(suututti-9, .-11)
Diffs
In FinnTreeBank (FI_FTB), if a clausal complement modifies
a word other than a verb, the reason is most likely
a verb-derived noun
(eg. ei käy kieltäminen, etteikö olisi mennyt(ccomp)
“no denying that ∅ wouldn’t have gone”).
Clauses modifying regular nouns get marked as acl:
semmonen kampaus et se kestää(ccomp)
lit. “the kind of hairstyle that it lasts”
(according to Universal dependency relations: acl).
compound: compound
compound is one of the three UD relations UD for compounding,
together with mwe and name.
UD Finnish additionally defines two subtypes of compound,
compound:nn for noun compounds and compound:prt
for phrasal verb particles. As the other UD types and the
UD Finnish subtypes cover most cases of compounding, only
few cases are annotated specifically as compound.
Numerical expressions consisting of multiple tokens are annotated
using the compound dependency type. The last word of the numerical
expression is the governor, and the number dependencies are chained.
(These relations are annotated number in
TDT.)
Poikasia on yleensä 3 - 5 . \n Youngsters are usually 3 to 5 .
nsubj:cop(5-6, Poikasia-1)
cop(5-6, on-2)
advmod(5-6, yleensä-3)
compound(--5, 3-4)
compound(5-6, --5)
punct(5-6, .-7)
compound:nn: noun compound modifier
The dependency type compound:nn, which stands for noun compound
modifier has two basic uses in the Finnish UD scheme.
First, In Finnish, compounds are generally written as a single word,
but for instance some compounds involving foreign words or proper
names are written separately using a dash. These are annotated using
compound:nn.
Da Vinci -merkkinen luomiväri \n Da Vinci -make eyeshadow
name(Vinci-2, Da-1)
compound:nn(-merkkinen-3, Vinci-2)
amod(luomiväri-4, -merkkinen-3)
The second use of the type compound:nn is to mark appellation
modifiers, which are modifying, non-inflecting noun phrases that
generally express profession, rank, position, assignment or other such
classifiable property (see ISK
§1062). The
phenomenon is closely related to that of apposition, and the
distinction between the two is described in
Appositions and appellation modifiers.
Professori Matti Tamminen pitää puheen . \n Professor Matti Tamminen gives a_speech .
compound:nn(Tamminen-3, Professori-1)
name(Tamminen-3, Matti-2)
nsubj(pitää-4, Tamminen-3)
dobj(pitää-4, puheen-5)
punct(pitää-4, .-6)
In written Finnish, erroneously writing compounds as two words is a
common mistake (e.g. ruoka pöytä). While these cases were annotated
as nn in the TDT corpus, in
UD Finnish goeswith is used instead.
compound:prt: phrasal particle
Phrasal particles (compound:prt) are used in connection with phrasal verbs, (partikkeliverbi, “particle verb” in Finnish grammar) where the particle is considered an integral part of the verb expression. The governor of the dependency is the verb, and the dependent is the phrasal particle.
Kävi ilmi , että ehdotus oli hyvä . \n Turned out , that the_suggestion was good .
compound:prt(Kävi-1, ilmi-2)
csubj(Kävi-1, hyvä-7)
punct(hyvä-7, ,-3)
mark(hyvä-7, että-4)
nsubj:cop(hyvä-7, ehdotus-5)
cop(hyvä-7, oli-6)
punct(hyvä-7, .-8)
Verb particles (see ISK §455 and §658) are the only case where particles are distinguished from adverbs in UD Finnish. This distinction can be made by the following rough rules. A word is a verbal particle if it, together with the verb, forms an expression that has a meaning that differs from the meaning of the verb alone, and if the word cannot be modified by an adverb.
For instance, laittaa kiinni (make closed, close) is not a phrasal verb, as kiinni can be modified.
Laitoin oven kokonaan kiinni. (I closed the door entirely.)
In contrast, ottaa kiinni (catch) is a phrasal verb, as it has a meaning distinct from the verb ottaa (take), and kiinni cannot be modified.
*Poliisi otti rosvon kokonaan kiinni. (*The police caught the robber entirely.)
See detailsconj: coordinated element
The SD scheme marks coordination so that the first coordinated element acts as the head, and the rest of the elements in the coordination, as well as the coordinating conjunction, depend on it. Coordinated elements are marked with the dependency type conj.
maitoa , leipää ja voita \n milk , bread and butter
punct(maitoa-1, ,-2)
conj(maitoa-1, leipää-3)
cc(maitoa-1, ja-4)
conj(maitoa-1, voita-5)
Conj is in Finnish used also in implicit coordinations without a coordinating conjunction, when the coordinated elements are not separated by a colon, semicolon or dash. When one of these is present, parataxis is used.
Kenkäkyllästyminen on kamala tauti , haluan kevään jo tänne. \n To\_get\_bored\_with\_shoes is an\_awful disease , I\_want spring already here.
conj(tauti-4, haluan-6)
Diffs
Turku Dependency Treebank
Conj is used also for implicit coordinations without a coordinating conjunction when the coordinated elements are not separated by a colon, semicolon or dash.
cop: copula
Copular clauses receive a special treatment in the SD scheme. The
predicative acts as the head word of the clause, and the copular verb
depends on it using a cop (copula) dependency. The only copular
verb in Finnish is olla “be” (see ISK
§891).
Huivi oli punainen . \n The_scarf was red .
nsubj:cop(punainen-3, Huivi-1)
cop(punainen-3, oli-2)
punct(punainen-3, .-4)
References
- http://scripta.kotus.fi/visk/sisallys.php?p=891 (in Finnish)
csubj: clausal subject
A clausal subject (csubj) is a clause that acts as the subject of
another clause.
Hänen aikomuksenaan oli mennä ulos . \n His intention(essive) was to_go out .
poss(aikomuksenaan-2, Hänen-1)
nmod(oli-3, aikomuksenaan-2)
csubj(oli-3, mennä-4)
advmod(mennä-4, ulos-5)
punct(oli-3, .-6)
csubj:cop: clausal copular subject
A clausal copular subject (csubj:cop) is a clause that acts as the
subject of another, copular clause. As in all copular clauses, the
predicative acts as the head of the clause and hence it is also the
governor of the copular subject.
Oli hauskaa käydä teatterissa . \n Was fun to_go to_theatre .
cop(hauskaa-2, Oli-1)
csubj:cop(hauskaa-2, käydä-3)
nmod(käydä-3, teatterissa-4)
punct(hauskaa-2, .-5)
dep: dependent
The dependency type dep is used when no other, more specific dependency
type applies, for example due to software limitation or error.
In the original TDT corpus, the type dep is additionally used to
annotate a number of idiomatic multi-word-expressions. In UD Finnish,
these are annotated as mwe.
det: determiner
There are no definite or indefinite articles in the Finnish language, but there are other determiners (see ISK §1409). In TDT, mostly pronouns are marked as determiners (det), because numerals, which can also be analyzed as determiner-like, are marked as numeral modifiers nummod, and genitive modifiers, also determiner-like, are marked with nmod:poss.
Kaikki miehet pudistivat päätään . \n All men shook their_head .
det(miehet-2, Kaikki-1)
nsubj(pudistivat-3, miehet-2)
dobj(pudistivat-3, päätään-4)
punct(pudistivat-3, .-5)
discourse: discourse element
Discourse elements are interjections, exclamations or emoticons. They are attached to the main verb or predicative of the
sentence with the discourse dependency type. In Finnish, these do not include other discourse markers.
Hei , tule auttamaan ! \n Hey , come to_help !
discourse(tule-3, Hei-1)
punct(Hei-1, ,-2)
xcomp(tule-3, auttamaan-4)
punct(tule-3, !-5)
Hmm ... Mitähän tuohon sanoisi ? \n Umm ... What to_that to_say ?
discourse(sanoisi-5, Hmm-1)
punct(Hmm-1, ...-2)
dobj(sanoisi-5, Mitähän-3)
nmod(sanoisi-5, tuohon-4)
punct(sanoisi-5, ?-6)
Iguazu on Argentiinassa :] \n Iguazu is in Argentina :]
discourse(on-2, :]-4)
Diffs
Turku Dependency Treebank
These do not include other discourse markers than interjections or exclamations.
dislocated: dislocated elements
The dislocated relation is used for fronted or postposed elements
that do not fulfill the usual core grammatical relations of a
sentence.
(TODO: Finnish dislocated example)
Diffs
Turku Dependency Treebank
dislocated is not annotated in TDT and is not used in
the current version of the UD Finnish corpus.
dobj: direct object
The dependency type dobj is used for (nominal) direct objects of the
verb.
Hän sulki oven . \n He closed the_door .
nsubj(sulki-2, Hän-1)
dobj(sulki-2, oven-3)
punct(sulki-2, .-4)
As in Finnish a passive clause does not have a subject, and what in
English would be considered the passive subject, is in Finnish the
direct object, the type nsubjpass is not used in TDT, but the type
dobj is used instead.
Päätös tehtiin marraskuussa . \n The_decision was_made in_November .
dobj(tehtiin-2, Päätös-1)
nmod(tehtiin-2, marraskuussa-3)
punct(tehtiin-2, .-4)
expl: expletive
This relation captures pleonastic nominals. An expletive or pleonastic nominal is one where the nominal does not satisfy a semantic role of the predicate, such as the existential there in English. The main predicate of the clause (the verb or predicate adjective or noun) is the governor.
(TODO example)
Diffs
Turku Dependency Treebank
expl is not annotated in TDT and is not used in the current version of the UD Finnish corpus.
foreign: foreign
The dependency relation foreign is used to label sequences
of unanalyzed foreign words. These are annotated without regard
to internal structure, marking all words in the sequence direct
dependents of the head. The head is the first word of the foreign phrase.
Note that foreign does not apply to foreign names, which
are annotated using name, or to sequences of analyzed words
of foreign origin (such as established loanwords), which are
annotated with their full syntactic structure.
Opimme fyysikoiden Let's assume a spherical cow -lähestymistavan .
dobj(Opimme-1, -lähestymistavan-8)
nmod:poss(-lähestymistavan-8, fyysikoiden-2)
compound:nn(-lähestymistavan-8, Let's-3)
foreign(Let's-3, assume-4)
foreign(Let's-3, a-5)
foreign(Let's-3, spherical-6)
foreign(Let's-3, cow-7)
punct(Opimme-1, .-9)
goeswith: goeswith
This goeswith dependency relation links two parts of a compound word
that are erroneously separated in text that is not well edited. The
head is in some sense the “main” part, often the second part.
Ruoka pöytä notkui herkkuja . \n Food table was_full_of goodies .
goeswith(pöytä-2, Ruoka-1)
nsubj(notkui-3, pöytä-2)
dobj(notkui-3, herkkuja-4)
punct(notkui-3, .-5)
Note that the original TDT
corpus used the nn relation
(compound:nn in UD Finnish) to annotate these relations.
list: list
The list relation is used for chains of comparable items.
Web text often contains passages which are meant to be interpreted
as lists but are parsed as single sentences. Email signatures in
particular contain these structures, in the form of contact
information: the different contact information items are labeled
as list; the key-value pair relations are labeled as appos.
In lists with more than two items, all items of the list should modify the first one.
Maria Korhonen Puhelin: 555-9814 Sähköposti: mkorhonen@example.com
name(Maria-1, Korhonen-2)
list(Maria-1, Puhelin:-3)
list(Maria-1, Sähköposti:-5)
appos(Puhelin:-3, 555-9814-4)
appos(Sähköposti:-5, mkorhonen@example.com-6)
Diffs
Turku Dependency Treebank
list is not annotated in TDT and is not used in
the current version of the UD Finnish corpus.
mark: marker
A marker (mark) is the subordinating conjunction in a
non-complement subordinate clause.
Kun tulin kotiin , jätin avaimen pöydälle . \n When I_came home , I_left key on_table .
mark(tulin-2, Kun-1)
nmod(tulin-2, kotiin-3)
punct(tulin-2, ,-4)
advcl(jätin-5, tulin-2)
dobj(jätin-5, avaimen-6)
nmod(jätin-5, pöydälle-7)
punct(jätin-5, .-8)
The dependency relation mark is also used to identify the
complementizer appearing in most clausal complements, where the head
of the dependency is the main verb of the subordinate clause. The only
complementizer in Finnish is että “that”. (These relations are
annotated complm in TDT.)
Sanoin , että hän voi tulla . \n I_said , that he can come .
ccomp(Sanoin-1, tulla-6)
punct(tulla-6, ,-2)
mark(tulla-6, että-3)
nsubj(tulla-6, hän-4)
aux(tulla-6, voi-5)
punct(Sanoin-1, .-7)
mwe: multi-word expression
The multi-word expression (modifier) relation is used for certain fixed expressions that behave as a unit.
Multiword expressions are annotated in a flat, head-initial structure, in which all words in the expression modify the first one using the mwe label.
Itse asiassa hän tuli jo eilen . \n As_a matter_of_fact he came already yesterday .
mwe(Itse-1, asiassa-2)
advmod(tuli-4, Itse-1)
nsubj(tuli-4, hän-3)
advmod(tuli-4, eilen-6)
advmod(eilen-6, jo-5)
punct(tuli-4, .-7)
name: name
Proper names constituted of more than one word are annotated using the dependency type name. The first (leftmost) word is the head, and the other words are direct dependents of the head.
The name dependency relation is used in cases where the multi-word name does not have an obvious internal syntactic structure, as is the case with for instance names of people (Matti Virtanen) or cities (New York).
If a name has an obvious internal structure, as is often the case in names of books and movies for instance, this structure is marked instead, and the name relation is marked as a secondary relation (DEPS field) spanning over the whole name phrase (from first token to the last).
Jumalat juhlivat öisin on Donna Tarttin esikoisteos . \n Gods celebrate by_night is Donna Tartt's first_work .
nsubj(juhlivat-2, Jumalat-1)
nmod(juhlivat-2, öisin-3)
name(Jumalat-1, öisin-3)
nsubj:cop(esikoisteos-7, juhlivat-2)
cop(esikoisteos-7, on-4)
punct(esikoisteos-7, .-8)
name(Donna-5, Tarttin-6)
nmod:poss(esikoisteos-7, Donna-5)
neg: negation modifier
In Finnish, negation is marked using the verb ei, which is used as an auxiliary and assigned the dependency relation neg. The most commonly negated elements are verbs and verb phrases, but occasional exceptions in verbless constructions are allowed.
If a conjunction or adverb has been merged together with ei, as in for instance ettei (että+ei “that+not”) or miksei (miksi+ei “why+not”), then the word is marked as a conjunction or an adverb rather than a negation verb. However, eikä “and+not”, when it appears alone and not coordinating another clause or phrase, is still marked as neg.
Hän ei sanonut mitään . \n He didn't say anything .
nsubj(sanonut-3, Hän-1)
neg(sanonut-3, ei-2)
dobj(sanonut-3, mitään-4)
punct(sanonut-3, .-5)
Diffs
FinnTreeBank (FI_FTB) is tokenized differently:
it treats ettei the same as että ei (“that not”),
miksei as miksi ei (“why not”) and so on,
as if they were two separate words.
Consequently, ei (“not”) represents a normal negation verb
in these combinations and gets annotated neg.
On the other hand, eikä (“and+not”) is treated as a single
token with a clitic particle (and also marked as neg).
nmod: nominal modifier
Nominal modifiers are inflected nominals which modify most commonly a verb or a noun phrase. They can occur alone or together with an adposition in an adpositional phrase. Both cases are analyzed similarly, as semantically nominal modifiers and adpositional phrases are similar.
Maljakko oli pöydällä . \n The_vase was on_the_table .
nsubj(oli-2, Maljakko-1)
nmod(oli-2, pöydällä-3)
punct(oli-2, .-4)
Maljakko oli pöydän päällä . \n The_vase was table on_top_of .
nsubj(oli-2, Maljakko-1)
nmod(oli-2, pöydän-3)
case(pöydän-3, päällä-4)
punct(oli-2, .-5)
nmod:gobj: genitive object
Certain nouns, those which have been directly derived from a verb or otherwise have a verb counterpart, can take an object in Finnish. These objects closely resemble more general genitive modifiers nmod:poss.
Perunan viljely oli tuolloin harvinaista . \n Potato(N+genitive) growing(N+nominative) was then rare .
nmod:gobj(viljely-2, Perunan-1)
nsubj:cop(harvinaista-5, viljely-2)
cop(harvinaista-5, oli-3)
advmod(harvinaista-5, tuolloin-4)
punct(harvinaista-5, .-6)
nmod:gsubj: genitive subject
Genitive subjects are subject-like arguments taken by a noun. This is in parallel to genitive objects nmod:gobj.
maljakon särkyminen \n vase(gen.) breaking
nmod:gsubj(särkyminen-2, maljakon-1)
nmod:own: haver
In Finnish, there is no direct equivalent to the verb have. Rather, having is expressed using the verb olla, to be. For instance, the meaning of the sentence I have a pen would be expressed in Finnish by Minulla on kynä, literally “At me is a pen”, with “me” in the adessive and “pen” in the nominative case. In UD Finnish, these so called possessive clauses (omistuslause) are analyzed as a subtype to existential clauses, (eksistentiaalilause) making the thing had (kynä in the previous example) the subject.
This kind of an analysis would naturally result in the haver being
marked as a nominal modifier, nmod. However, as nmod is a very
frequent dependency type that encodes many different meanings, the
information that the clause is about having or owning would be
lost. Therefore, the UD Finnish scheme applies the nmod:own
dependency type for nominal modifiers that encode owning,
following the approach of TDT.
The governor of the dependency is the verb olla, and the dependent
is the haver or owner, which is required to be in the adessive case.
The haver must also be an animate being or a group of animate beings.
Matilla on uusi auto . \n At_Matti is new car .
nmod:own(on-2, Matilla-1)
nsubj(on-2, auto-4)
amod(auto-4, uusi-3)
punct(on-2, .-5)
nmod:poss: genitive modifier
The dependency type poss stands for possessive in the original SD
scheme, but in UD Finnish, the corresponding type nmod:poss is used
for genitive modifiers in general, which in Finnish often but not nearly
always imply possession. There are two kinds of genitive modifiers that
are not annotated using the general genitive modifier type: the genitive
object, nmod:gobj and the genitive subject, nmod:gsubj.
Matin penaali jäi kouluun . \n Matti's pencilcase was_left at_school .
nmod:poss(penaali-2, Matin-1)
nsubj(jäi-3, penaali-2)
nmod(jäi-3, kouluun-4)
punct(jäi-3, .-5)
Autossa on kahden litran moottori . \n In_the_car is two litre's engine .
nmod(on-2, Autossa-1)
nsubj(on-2, moottori-5)
punct(on-2, .-6)
nmod:poss(moottori-5, litran-4)
nummod(litran-4, kahden-3)
Diffs
FinnTreeBank (FI_FTB) applies
the universal relation nmod
instead of the language-specific nmod:poss.
nsubj: nominal subject
The dependency type nsubj marks nominal subjects of the non-copular
clause.
Eeva tuli töistä . \n Eeva came from_work .
nsubj(tuli-2, Eeva-1)
nmod(tuli-2, töistä-3)
punct(tuli-2, .-4)
nsubj:cop: nominal copular subject
The dependency type nsubj:cop is used for the nominal subject
of a copular clause. The predicative is the head of the copular
clause, and also the governor of the nsubj:cop dependency.
Matto on jo kuiva . \n The_mat is already dry .
nsubj:cop(kuiva-4, Matto-1)
cop(kuiva-4, on-2)
advmod(kuiva-4, jo-3)
punct(kuiva-4, .-5)
nummod: numeric modifier
Numeric modifiers of a noun or NP, including both cardinal and
ordinal numbers, are marked with the nummod dependency type. This
dependency type is used also with for instance years and program
versions.
Laukku painoi 20 kiloa . \n Bag weighed 20 kilograms .
nsubj(painoi-2, Laukku-1)
dobj(painoi-2, kiloa-4)
nummod(kiloa-4, 20-3)
punct(painoi-2, .-5)
Tapasin hänet vuonna 1972 . \n I_met him in_the_year 1972 .
dobj(Tapasin-1, hänet-2)
nmod(Tapasin-1, vuonna-3)
nummod(vuonna-3, 1972-4)
punct(Tapasin-1, .-5)
parataxis: parataxis
parataxis dependencies mark two different phenomena. Firstly, they are used with certain implicit coordinations. These coordinations are recognized by two factors: there is no coordinating conjunction, and the independent clauses are separated by a colon, semicolon or a dash. As with explicit coordinations, the first element is the governor. Also parenthetical and interjected clauses can receive the parataxis dependency. If there is a coordinating conjunction present (regardless of punctuation) or if the clauses are separated by merely a comma, the coordination type conj is used (differs from the UD).
Lapset leipoivat piparit ; äiti paistoi ne . \n The_children made the_cookies ; the_mother baked them .
nsubj(leipoivat-2, Lapset-1)
dobj(leipoivat-2, piparit-3)
punct(leipoivat-2, ;-4)
punct(leipoivat-2, .-8)
parataxis(leipoivat-2, paistoi-6)
nsubj(paistoi-6, äiti-5)
dobj(paistoi-6, ne-7)
Ilmoittakaa suoraan sähköpostilla ( palautelomakekin käy ) .
nmod(Ilmoittakaa-1, sähköpostilla-3)
advmod(sähköpostilla-3, suoraan-2)
punct(Ilmoittakaa-1, .-8)
parataxis(Ilmoittakaa-1, käy-6)
nsubj(käy-6, palautelomakekin-5)
punct(käy-6, (-4)
punct(käy-6, )-7)
parataxis is also used for direct speech. If the verb of saying (or thinking, etc.) precedes the utterance, it acts as the governor, and the main verb or predicative of the utterance is the dependent. However, if the utterance precedes the speech verb, the relation is reversed.
Mies sanoi : Tulen kotiin vasta myöhään . \n The_man said : I_will_come home only late .
nsubj(Mies-1, sanoi-2)
punct(sanoi-2, :-3)
nmod(Tulen-4, kotiin-5)
advmod(Tulen-4, myöhään-7)
advmod(myöhään-7, vasta-6)
parataxis(sanoi-2, Tulen-4)
punct(sanoi-2, .-8)
Tulen kotiin vasta myöhään , mies sanoi . \n I_will_come home only late , the_man said .
nmod(Tulen-1, kotiin-2)
advmod(Tulen-1, myöhään-4)
advmod(myöhään-4, vasta-3)
punct(Tulen-1, ,-5)
nsubj(sanoi-7, mies-6)
parataxis(Tulen-1, sanoi-7)
punct(Tulen-1, .-8)
Diffs
Turku Dependency Treebank
If the clauses are separated by a comma and there is no coordinating conjunction, TDT does not use parataxis but conj.
FinnTreeBank
FI_FTB does not apply the parataxis relation.
Instead, the phenomena of implicit coordination,
parenthetical clauses, interjected clauses and direct speech
have been described using some other, most applicable
relation available (usually conj or dobj).
punct: punctuation
The dependency type punct is used to mark punctuation. The dependent is the punctuation symbol, and the governor is the element which the punctuation symbol delimits. For instance, with coordination, the first coordinated element is the head of all punct dependencies in the coordination, and with subordinate clauses, the head of the subordinate clause is the governor of the punct.
kerrostaloja , rivitaloja ja omakotitaloja \n blockhouses , rowhouses and one_family_houses
punct(kerrostaloja-1, ,-2)
conj(kerrostaloja-1, rivitaloja-3)
cc(kerrostaloja-1, ja-4)
conj(kerrostaloja-1, omakotitaloja-5)
Kotona oli hiljaista , kun hän palasi . \n At_home was quiet , when he returned .
advmod(oli-2, Kotona-1)
xcomp(oli-2, hiljaista-3)
advcl(oli-2, palasi-7)
punct(palasi-7, ,-4)
punct(oli-2, .-8)
mark(palasi-7, kun-5)
nsubj(palasi-7, hän-6)
Diffs
By the current release of FI_FTB (FinnTreeBank), the manual annotation of punctuation marks has not been completed. Instead the automatic annotation links the punctuation marks to the closest token available (usually the previous one).
remnant: remnant in ellipsis
The remnant relation is used to provide a treatment of ellipsis (in
the case of gapping and stripping, where a predicational or verbal
head gets elided). In particular, remnant aims to provide analyses
that do not postulate empty nodes.
Maija meni Pariisiin ja Mirja Prahaan \n Maija went to Paris and Mirja to Prague
remnant(Maija, Mirja)
remnant(Pariisiin, Prahaan)
Diffs
FinnTreeBank (FI_FTB) does not apply the relation remnant.
Instead, in elliptical cases the relation conj has been
chosen to correspond to its correlate in a preceding clause.
reparandum: overridden disfluency
We use reparandum to indicate disfluencies overridden in a speech
repair. The disfluency is the dependent of the repair.
(TODO: Finnish reparandum example)
Diffs
Turku Dependency Treebank
reparandum is not annotated in TDT and is not used in
the current version of the UD Finnish corpus.
FinnTreeBank
The relation reparandum is used in FI_FTB
for the clearest cases of speech repair.
root: root
The root grammatical relation points to the root of the sentence. A
fake node ROOT is used as the governor.
ROOT Minä pidän oluesta .
root(ROOT, pidän)
vocative: vocative modifier
The dependency type vocative is used for vocatives, that is,
expressions where someone is being addressed. The governor of the
dependency is the main predicate of the clause where the addressing
occurs.
Pekka , tulisitko tänne ? \n Pekka , would_you_come here ?
vocative(tulisitko-3, Pekka-1)
punct(Pekka-1, ,-2)
advmod(tulisitko-3, tänne-4)
punct(tulisitko-3, ?-5)
xcomp: open clausal complement
The dependency type xcomp is reserved for clausal complements which have an external subject, that is, whose subject is shared with the complemented verb (a phenomenon also known as subject control). Note that the subject of the complementing clause must be the subject of the complemented verb, not any other sentence element (in other cases the dependency type xcomp:ds should be used instead). These complements are always non-finite.
Hän alkoi hakata halkoja . \n He started chopping the_wood .
nsubj(alkoi-2, Hän-1)
xcomp(alkoi-2, hakata-3)
dobj(hakata-3, halkoja-4)
punct(alkoi-2, .-5)
Many of the complements with an external subject resemble cases that are analyzed as main verbs with auxiliaries. Both auxiliaries and xcomp complements share their subject with another verb, but only a closed list of verbs are analyzed as auxiliaries (see aux). Note also that in auxiliary cases the second verb is the governor, whereas with xcomp the first verb becomes governor (unless the word order is inverse).
Secondary predicates
The xcomp relation is also used in constructions that are known as secondary predicates or predicatives, which are core arguments of the predicate. Most commonly these are adjetives, but also participial and nominal arguments are possible. For example in the sentence Hän teki siitä hyvin vaikeaa there are two mixed predicates: 1) hän teki jotain, and 2) se on vaikeaa. The secondary predication can be provided to any higher clause element.
Hän teki siitä hyvin vaikeaa . \n He made (from_)it very difficult .
nsubj(teki, Hän)
nmod(teki, siitä)
advmod(vaikeaa, hyvin)
xcomp(teki, vaikeaa)
punct(teki, .)
Minulla oli mukavaa .
xcomp(oli, mukavaa)
nmod(oli, Minulla)
punct(oli, .)
Note: Adjectival complements correspond to acomp (adjectival complement) in
the original Stanford Dependencies and the Turku Dependency
Treebank.
Diffs
Secondary predication is not recognized in FinnTreeBank (FI_FTB).
Therefore xcomp is always a non-finite verb
(or a predicative of a non-finite copular clause) in FI_FTB.
As the examples above represent types of result and state clauses
(ISK § 891),
in FI_FTB they have been annotated using predicatives
(see chapter Copulas in Specific constructions).
xcomp:ds: clausal complement with different subject
The dependency type xcomp:ds, which stands for clausal
complement with different subject, is a subtype of xcomp (open clausal complement).
It is used for clausal complements where the subject of the complement clause is inherited from the higher clause, but it’s not a subject of the governing clause (if it is a subject, xcomp is used instead). Basically, this means that the subject of the complementing clause must be the object of the complemented clause. These complements are always non-finite.
Note that this subtype represents a departure from the general UD taxonomy approach in differentiating between inherited subject and object (see Syntax: General Principles).
Sain hänet itkemään . \n I_made him cry .
xcomp:ds(Sain-1, itkemään-3)
nsubj(itkemään-3, hänet-2)
punct(Sain-1, .-4)