Jeju/Universal Dependencies

From LING073
Jump to: navigation, search

The corpus and the trained udpipe files(contents of /dev/ folder) are uploaded on Jeju corpus repostiroy. Jeju transducer repository additionally holds the udpipe files.

Evaluation

  • Number of sentences
    • annotated: 38
    • annotated2: 15
  • Number of forms
    • annotated: 200
    • annotated2: 61

withmorph model

Scores
UAS LAS
annotated 89.50% 77.00%
annotated2 72.13% 40.98%

nomorph model

Scores
UAS LAS
annotated 79.00% 61.00%
annotated2 77.05% 37.70%

Dependency relations

1. advcl

advcl represents a clause which modifiers a verb or an adverb. In Jeju, it is frequently used to modify a verb by specifying the manner, cause and many other conditions or specif details of the action(of the verb).

"<가시리>"
	"가시리" n @list #1->3
"<,>"
	"," cm @punct #2->7
"<행원리>"
	"행원리" n @list #3->5
"<,>"
	"," cm @punct #4->7
"<용수리에>"
	"용수리에" n loc @obl #5->7
"<설치뒝>"
	"설치뒝" v adv @advcl #6->7
"<잇어>"
	"잇어" v pres @root #7->0
"<.>"
	"." sent @punct #8->7
"<그건>"
	"그건" n top @nsubj #1->4
"<정낭이옌>"
	"정낭이옌" n adv @advmod #2->3
"<ᄒᆞ는>"
	"*ᄒᆞ는" v top @advcl #3->4
"<거우다>"
	"거우다" v hon pres @root #4->0
"<..>"
	".." sent @punct #5->4

2. cc

cc stands for coordinating conjunctions. In Jeju, many words serve the role of coordinating conjunction. They can appear at the beginning of the state(without joining sentences), or they can be used like English 'and' to join two or more sentences.

"<경ᄒᆞ연>"
	"*경ᄒᆞ연" adv @cc #1->5
"<삼별초가>"
	"삼별초가" n nom @nsubj #2->5
"<제주도ᄁᆞ지>"
	"제주도ᄁᆞ지" n loc @obl #3->5
"<와>"
	"와" v aux @advcl #4->5
"<낫수게>"
	"낫수게" v hon past @root #5->0
"<.>" 
	"." sent @punct #6->5
"<경ᄒᆞ난>"
	"*경ᄒᆞ난" adv @cc#1->6
"<돌담으로>"
	"*돌담으로" n dat @obl #2->3
"<에우싼>"
	"에우싼" adj @amod #3->4
"<불턱을>"
	"불턱을" n acc @obj #4->5
"<멘들앙>"
	"멘들앙" v adv @advcl #5->6
"<이용헷수게>"
	"이용헷수게" v hon pres @root #6->0
"<.>"
	"." sent @punct #7->6

3. advmod

advmod stands for adverbial modifier. In Jeju, they can either come before a verb or an adjective.

"<잘도>"
	"잘도" adv @advmod #1->2
"<공펭ᄒᆞ우다>"
	"*공펭ᄒᆞ우다" v hon pres @root #2->0
"<.>"
	"." sent @punct #3->2
"<ᄆᆞ멀은>"
	"ᄆᆞ멀은" n top @nsubj #1->4
"<피를>"
	"피" n acc @obj #2->4
"<ᄆᆞᆰ게>"
	"ᄆᆞᆰ게" v adv @advmod #3->4
"<ᄒᆞᆫ다>"
	"*ᄒᆞᆫ다" v pres @root #4->0
"<.>"
	"." sent @punct #5->4

4. amod

amod represents adjectival modifiers, which in Jeju appear before and modifiers a nominal. As in English, adverbial modifiers can precede adjectival modifiers in Jeju.

"<이>"
	"이" det @det #1->2
"<ᄎᆞᆯ레에선>"
	"ᄎᆞᆯ레에선" n loc top @obl #2->5
"<특이ᄒᆞᆫ>"
	"*특이ᄒᆞᆫ" adj @amod #3->4
"<내움살이>"
	"내움살이" n nom @nsubj #4->5
"<남수다>"
	"남수다" v hon pres @root #5->0
"<.>"
	"." sent @punct #6->5
"<막>"
	"막" adv @advmod #1->2
"<좋은>"
	"좋은" adj @amod #2->3
"<음식이우다양>"
	"음식이우다양" n v hon pres addi @root #3->0
"<.>"
	"." sent @punct #4->3

5. obj

obj refers to an object of a verb. In Jeju, they are almost always by an accusative particle(-를 or -을). However, the particle is often skipped in imperative mode. For instance,

  • 그놈(-을) 잡어!(Catch him!)

is allowed. Indirect objects(iobj) are indicated with other particles(-에게, 게).

"<ᄆᆞ멀은>"
	"ᄆᆞ멀은" n top @nsubj #1->4
"<피를>"
	"피" n acc @obj #2->4
"<ᄆᆞᆰ게>"
	"ᄆᆞᆰ게" v adv @advmod #3->4
"<ᄒᆞᆫ다>"
	"*ᄒᆞᆫ다" v pres @root #4->0
"<.>"
	"." sent @punct #5->4
"<다른>"
	"다른" adj @amod #1->2
"<지방에서는>"
	"지방에서는" n loc top @obl #2->6
"<냉국에>"
	"냉국에" n dat @obl #3->6
"<소금>"
	"소금" n @obl #3->4
"<간을>"
	"간을" n acc @obj #5->6
"<ᄒᆞ메>"
	"ᄒᆞ메" v pres @root #6->0
"<.>"
	"." sent @punct #7->6