Kaingang/Universal Dependencies

From LING073
Jump to: navigation, search

kgp.annotated2.ud.conllu and kgp.annotated2.nomorph.conllu are the files we chose to train the parser on (not kgp.annotated.ud.conllu or kgp.annotated.nomorph.conllu).


Number of forms, UAS, LAS scores for kgp.annotated2.ud.conllu

$ udpipe --accuracy --parse kgp.withmorph.udpipe ../kgp.annotated2.ud.conllu
Loading UDPipe model: done.
Parsing from gold tokenization with gold tags - forms: 160, UAS: 70.62%, LAS: 55.00%
  • Number of sentences in corpus: 18

Number of forms, UAS, LAS scores for kgp.annotated2.nomorph.conllu

$ udpipe --accuracy --parse kgp.nomorph.udpipe ../kgp.annotated2.nomorph.conllu
Loading UDPipe model: done.
Parsing from gold tokenization with gold tags - forms: 160, UAS: 70.62%, LAS: 55.00%
  • Number of sentences in corpus: 18

the performance looks the same? —Jwashin1 (talk) 11:37, 21 May 2019 (EDT)


(for each relation: a description of the relation, noting various ways it might be used in the language; and two examples of the relation from your corpus, preferably illustrating what you described.)

Five dependency relations that we used at least twice in our annotation


Kaingang uses a set of postpositions (or what our source dictionary calls "circumstance markers" in the same way that English and Portuguese uses prepositions. We connected the objects of said prepositions with their respective postpositions using the CASE.

Inh mỹ tóg tỹ, ĩn nĩ ve nĩ.

me to SUB TOP, house similar looks ASP

To me, it looks like a house.

In the above example, we would connect the subject and topic Inh "I/me" to its dependent postposition mỹ to with CASE dependency.

Kaingang also uses a large set of different subject and topic markers that indicate what the subject and focus of the sentence is. We connected all of the nouns with their respective subject markers using CASE.

Kyfé tỹ tóg, ti nĩgé gynh.

knife with SUB his finger cut

(He) cut his finger with a knife (lit: the knife cut his finger)

In the above example, we would connect Kyfé "knife" with its respective subject marker (which always goes AFTER the subject noun phrase), in this case, tóg. (NOTE: whole phrases can function as a subject. We see this in this example because the subject of this sentence is a post-positional phrase: "with a knife").


We used the NSUBJ relation to relate subjects to non-verbal predicates. Where Portuguese and English use a copula "to be" verb, Kaingang instead uses a particular aspect marker "vẽ".

"Ka fej kusũg vẽ."

tree flower red ASP

"The tree's flower is red"

In the above example, we would connect the nonverbal predicate kusũg to its NSUBJ dependent fej.

"Inh mỹnh fi vẽ."

my mother her ASP

"She is my mother" (lit: my mother is her)

In the above example, we would connect the nonverbal predicate fi with its NSUBJ dependent mỹnh.


Kaingang does not conjugate its verbs. Instead, Kaingang speakers use mode markers, aspect markers, and opinion markers to modify verbs, define verb aspect, and apply changes to modality, respectively.

We used ADVMOD to connect a verb to each of its individual mode, aspect, and opinion markers.

this looks more like aux than advmodJwashin1 (talk) 11:39, 21 May 2019 (EDT)

Rán sór inh mũ.

write want I do

"I want to write" (lit: I am wanting to write)

Some verb indicators of mode function as would verbs in English. In this example, sór "to want" is a verb modifier (it can NOT stand alone as a verb), and is modifying rán "to write". We connect rán to its dependent sór using the ADVMOD dependency.

Ti ki kanhró inh nĩ ha

him in know I ASP now

"I know him now."

In the above example, the word ha "now" is an indicator of mode that modifies the verb. We connected the main verb kanhró "to know (a person)" to its dependent modifier using a ADVMOD.


Kaingang expresses noun possession and ownership by juxtaposing nouns with one another. They place the possessor in front of the possessed to make a genitive phrase. We connected a possessed noun to its preceding possessor using NMOD:POSS.

Porko fe kãnhvy ko mũ sóg nĩ.

Pig lung eat seated 1ps.SUB ASP

"I eat (while sitting) pig's lung"

In the above example, fe kãnhvy "lung" is an object being possessed by porko "pig". Hence, "pig's lung". We connected the possessed lung to its dependent possessor via a NMOD:POSS dependency.

Inh pãnĩ vỹ, kaga nĩ.

my back TOP hurt ASP

My back is hurting.

In the above example, the word pãnĩ is directly behind a personal pronoun, and is therefore acting as a possessed noun possessed by the personal pronoun inh "I". We connect pãnĩ to its dependent possessor with NMOD:POSS.


We used OBJ to connect nouns to their dependent direct objects. Direct objects are not marked, unlike their subject counterparts, and objects normally appear directly in front of their corresponding verb.

Pã’i ag vĩ pãtén ã huri.

authority their orders violate you already

"You have already violated the authority's orders."

In the above example, the word "orders, laws, edicts, words" is a direct object of the verb pãtén "violate". We connect the verb to its direct object dependent with OBJ.

Inh kẽj jẽnfĩ inh huri.

My basket finish I already/have

"I already finished my basket"

In this example, inh kẽj "my basket" is the direct object of jẽnfĩ "finish". We connect jẽnfĩ to its dependent direct object kẽj using OBJ.