UD for Old Guaraní

Tokenization and Word Segmentation

• Words are delimited by whitespace characters
• According to typographical rules, many punctuation marks are attached to a neighboring word. They are given as separate tokens (words);
• There are no adjectives in Old Guaraní. Modification is made by composition, juxtaposing lexical roots, so when a lexical root is modified by another a new word appears as in kuɲãporaŋ(a) ‘beautiful woman’ (kuɲã ‘woman’ + poraŋ-(a) ‘beauty’). Such words are treated sometimes as multiword tokens, sometimes as a single word.
• Some compound words from Portuguese are written as one word, e.g. santaCruz ‘holy cross’, espíritoSanto ‘holy spirit’.

Morphology

Tags

• Tupinambá uses 16 of the 17 universal POS categories. ADJ is not used since there is no separate class of adjectives. Stative-verbs and possessed nouns behave alike, in a way that is not possible to distinguish them morphologically (I am ugly / I have uglyness; ugly boy / boy with uglyness /(the) boy has uglyness). The fundamental distinction in Tupinambá is that between predicates and arguments (potentially referring expressions), reflected in the NOUNor VERB tags. Lexical roots in Tupinambá are predicates, which require function indicating morphology in order to function as arguments (Croft 2021)

Mapping UPOS to XPOS Old Guaraní

Nominal Features

• Old Guaraní nouns are not marked for gender. Number is optionally marked by the lexical root etá `(be) many’.

kunumĩ ‘boy / boys’
kunumĩ-etá ‘boys’

• Person indexes and pronouns are given below:

Possessor indexes, as the name suggests, only index possessors. They are marked not marked with PronType, but they are marked as Poss=Yes. Argument indexes are used with verbal predicates, as also are the portmanteau indexes (see below). Switch-reference indexes are used in dependent clauses with subjects coreferential with the subjects of the main clauses. Person indexes distinguish Number(Singular or Plural). They also distinguish Clusivity in the 1^st person plural.

• Nouns can take the following Cases: Tra, Loc, Per, Dat.

• The relational markers Rel, which indicate contiguity or non-contiguity between a head and its dependent, take respectively the following features: Rel=Cont and Rel=NCont. A third type or relational, Rel=HUM, indicates that obligatorily possessed noun is given without reference to a possessor at the same time marking the possessor as human. The reflexive/correferential morpheme o. which is often referred to as ‘relational₃’ is associated with the feature-value Reflex=Yes. Another relational Rel=Humindicates that the possessor is human. These are shown below:

• Nouns may also be reduplicated in both ways denoting: plurality, collectivity, superlativity, and other semantic nuances. Numerals may also be reduplicated in order to indicate distribution.
• Nouns are also marked for tense.
• As an omnipredicative language, lexical roots in Tupinambá are existential predicates. In order to function as arguments, the referential marker (a ̴ ∅), is required (marked as Case=Ref) despite its function being nothing like that of nominal cases.

Verbal Features

• The lexical root in the gerund (VerbForm=Ger) is marked as VERB even when combining with a relational.
• Verbs are marked for aspect: Compl (completive), Iter (Iterative), Suc Successive.
• Verbs are also marked for mood: Perm (permissive) Imp (imperative).
• Lexical roots may be reduplicated in two differentways: monosylabic reduplication (Red=Mo), disylabic reduplication (Red=Di). The modify the aspect of the verb in different ways: disylabic reduplication indicate the repetition or duration of an action; monosylabic reduplication indicates iteration of the action.

Nominalizers

Tupinambá has many nominalizers with different functions. All but

Syntax

• As a head-marking language, core arguments, are indexed on the predicate, in the order SOV as in the example below:

aɲan
a-ɲan
1.SG-run
'I ran/run'

osepjak 
o-s-epjak
3.SG(S)-3(O)-see
He/she/it/they see her/him/it/them

• Nominal phrases (NPs) semantically related to the core-arguments can appear in any order in relation to the predicate (where the core arguments are indexed). This is exemplified below through the sentence John sees Mary:

John_i Mary_j o_i-s_j-epjak

Mary_j John_i o_i-s_j-epjak

o_i-s_j-epjak John_i Mary_j

o_i-s_j-epjak Mary_j John_i

Mary_j o_i-s_j-epjak John_i

John_i o_i-s_j-epjak Mary_j

• The dependency linking the core arguments with the NPs semantically related to the core-arguments are obl (oblique).

• The object of transitive verbs is always indexed by Rel=Ncont. Tupinambá has transitive verbs only when objects are third person. In the other cases a predicative possession is used. When the object is third person, the feature Person takes values combining both arguments (S and O): Person=13, Person=23, and Person=33.
• Other combinations of person indexes occur in the case of 1 → 2. Two portmanteau pronouns are available: oro- 1 → 2.SG, and opo- 1 → 2.PL:

orojuká 
oro-juká
1.2SG(O)-kill
'I/We kill(ed) you'

opojuká 
opo-juká
1.2PL(O)-kill
'I/we kill(ed) you'

• What has been traditionally called circunstantial mood or indicative II in some Tupí-Guaraní languages referes to a nominalization accompanied by the fronting of an adverbial expression: adverbs, adverbial expressions, postpositional phrases (oblique topicalization). The nominalized form in this case is marked by the feature-value OblTop=Yes.

Arguments (or potentially refering expressions)

Since all lexical roots in Tupinambá are predicates, their use as potentially refering expressions or arguments require aditional morphology. Compare both examples below:

nerub
ne=r-ub
2.SG=Cont-father
'I have a father'

neruβa
ne=r-ub-a
2.SG=Cont-father-Ref
'My father'

Treebanks

There is 1 Tupinamba UD treebank:

• Tupinamba-A

Instruction: Treebank-specific pages are generated automatically from the README file in the treebank repository and from the data in the latest release. Link to the respective *-index.html page in the treebanks folder, using the language code and the treebank code in the file name.

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