| layout | title | udver |
|---|---|---|
base |
Korean UD |
2 |
- The tokenization of the Korean UD treebanks follows the tokenization of the Korean data distributed by the SPMRL 2013 shared task, which is a straightforward whitespace-based tokenization with conventional separation of punctuation.
- There are no words with spaces.
- There are currently no multi-word tokens. This may change in the future, as some words have no
space between them, and instead of indicating this by
SpaceAfter=Noin MISC, multi-word tokens may be preferable.
- At present, the lemma column in the GSD and Kaist treebanks violates the UD guidelines. Instead of showing a selected surface form as the citation form for the lexeme, it shows the morphemes delimited by the plus (+) character. This should be fixed in future version and a real lemma should be provided.
- All 17 universal POS categories are relevant in Korean, including particles (PART). At present, hundreds of word types are tagged PART. This is a legacy of an existing Korean morphological analyzer and many of these words should probably belong to another category in UD; however, the exact list has yet to be worked out.
- The following words are treated as auxiliaries (AUX):
- The affirmative copula 이 i (“to be”) is written as a suffix to the nominal predicate but it is treated as a separate auxiliary verb in UD.
- 않 anh (“to not be”) is the negative copula or an auxiliary in a negative clause.
- 있 iss (“to be”) is an auxiliary in affirmative clauses.
- 하 ha (“must, should”) is a necessitative modal auxiliary.
- 싶 sip (“will, to want”) is a desiderative modal auxiliary.
Instruction: Describe inherent and inflectional features for major word classes (at least NOUN and VERB). Describe other noteworthy features. Include links to language-specific feature definitions if any.
- Korean uses a nominative-accusative alignment. Direct objects are marked by the accusative morpheme 을 eul.
- The following relation subtypes are used in Korean:
- nsubj:pass for nominal subjects of passive verbs
- csubj:pass for clausal subjects of passive verbs
- nmod:poss for possessive (genitive) modifiers
- det:poss for possessive determiners
- acl:relcl for relative clauses
- obl:tmod for temporal adjuncts
- compound:lvc for light verb constructions
- flat:name for connection of parts of a flat multi-word named entity
- The following relation types are not used in Korean at all: expl, list, parataxis, reparandum
There are 3 Korean UD treebanks: