Okinawan/Universal Dependencies

Eval

(Formatting borrowed from Darbus.)

	withmorph		nomorph
	LAS	UAS	LAS	UAS	Forms	Sentences
annotated	51.30%	63.48%	23.48%	36.52%	115	15
annotated2	42.37%	62.71%	44.07%	67.80%	59	11

Dependency relations

discourse

Since the corpora that I used for this assignment were from the "Pear story", there were a lot of discourse markers. I tried to connect them to the root (unless it was prohibitively far away).

"<いっけえらさぐとぅ>"
	"いっけえるん" v cnd @conj #1->5
"<な>"
	"な" dsc @discourse #2->5
"<るうん>"
	"るう" prn foc @nsubj #3->5
"<あり>"
	"あり" dsc @discourse #4->5
"<すいへや>"
	"すん" v cvb @root #5->0
"<.>"
	"." sent @punct #6->5

"<ちいちなち>"
	"ちいち" n dsc hs @obj #1->3
"<とぅらち>"
	"とぅらちゅん" v med @aux #2->3
"<きてぃんじゃん>"
	"きいん" v cvb prog @root #3->0
"<ばあよ>"
	"ばあよ" dsc @discourse #4->3
"<.>"
	"." sent @punct #5->3

advmod

Because Okinawan hides so much meaning in case markers, much of the semantics of the language are somewhat invisible in UD land. The advmod tag acted as a bit of a catchall for words that modified verbs.

"<あんちゃぐとぅ>"
	"あんちゃぐとぅ" dsc @discourse #1->3
"<うま>"
	"うま" prn dem abs @advmod #2->3
"<とぅまとぅん>"
	"とぅむん" v prog @root #3->0
"<ばあよ>"
	"ばあよ" dsc @discourse #4->3
"<.>"
	"." sent @punct #5->3

"<まあしが>" "まあ" intadv all foc @advmod #1->2 "<むたてぃんじゅる>" "むたりいん" v cvb imp @ccomp #2->3 "<わからん>" "わかるん" v neg @root #3->0 "<ばあよ>" "ばあよ" dsc @discourse #4->3 "<.>" "." sent @punct #5->3

=== obj ===

Most of the sentences that I used involved transitive verbs, so I used the obj tag for the (usually) nominals that were their objects.

<pre>
"<すぐ>"
	"すぐ" dsc @discourse #1->5
"<なあ>"
	"なあ" dsc @discourse #2->5
"<あり>"
	"あり" dsc @discourse #3->5
"<じてんしゃ>"
	"じてんしゃ" n abs @obj #4->5
"<ぬてぃきし>"
	"ぬいん" v cvb prog med @root #5->0
"<ぼおし>"
	"ぼおし" n abs @obj #6->7
"<かんてぃ>"
	"かんとぅん" v cvb @conj #7->5
"<.>"
	"." sent @punct #8->5

conj

I wasn't super confident about this one, but I generally attached this to non-auxiliary verbs that I did not deem to be the root.

"<いっけえらさぐとぅ>"
	"いっけえるん" v cnd @conj #1->5
"<な>"
	"な" dsc @discourse #2->5
"<るうん>"
	"るう" prn foc @nsubj #3->5
"<あり>"
	"あり" dsc @discourse #4->5
"<すいへや>"
	"すん" v cvb @root #5->0
"<.>"
	"." sent @punct #6->5

"<すぐ>"
	"すぐ" dsc @discourse #1->5
"<なあ>"
	"なあ" dsc @discourse #2->5
"<あり>"
	"あり" dsc @discourse #3->5
"<じてんしゃ>"
	"じてんしゃ" n abs @obj #4->5
"<ぬてぃきし>"
	"ぬいん" v cvb prog med @root #5->0
"<ぼおし>"
	"ぼおし" n abs @obj #6->7
"<かんてぃ>"
	"かんとぅん" v cvb @conj #7->5
"<.>"
	"." sent @punct #8->5

amod

Pretty much what's written on the tin. Most of the time, this attached to adjectives, but sometimes it would be used on a noun that was not suited to be `case`.

"<なあ>"
	"なあ" dsc @discourse #1->6
"<とぅすいが>"
	"とぅすい" n nom @amod #2->6
"<やら>"
	"やら" cop cvb @cop #3->2
"<また>"
	"また" dsc @discourse #4->6
"<わかある>"
	"わかさん" adj vz npst @amod #5->6
"<ちゅが>"
	"ちゅ" n nom @obj #6->8
"<やら>"
	"や" cvb @cop #7->5
"<わからにが>"
	"わかるん" v neg cvb nom @root #8->0
"<きいぬ>"
	"きい" n nom @obj #9->11
"<いいし>"
	"いい" adj all @advmod #10->11
"<ぬぶやあくん>"
	"ぬぶいん" v prog @conj #11->8
"<ばあよ>"
	"ばあよ" dsc @discourse #12->11
"<.>"
	"." sent @punct #13->11

"<いきがぬ>"
	"いきが" n gen @root #1->0
"<うをぅちゅる>"
	"うをぅちゅ" n emp @amod #2->1
"<やらはじろ>"
	"や" cop cvb amb dsc @cop #3->1
"<うれ>"
	"うれ" prn dem @det #4->1
"<.>"
	"." sent @punct #5->1

cop

The `cop` tag was well suited to Okinawan, as the language actually distinguishes between the copula and verb "to be" (similarly to Japanese). This often connected to its nominal.

"<いきがぬ>"
	"いきが" n gen @root #1->0
"<うをぅちゅる>"
	"うをぅちゅ" n emp @amod #2->1
"<やらはじろ>"
	"や" cop cvb amb dsc @cop #3->1
"<うれ>"
	"うれ" prn dem @det #4->1
"<.>"
	"." sent @punct #5->1

"<なあ>"
	"なあ" dsc @discourse #1->6
"<とぅすいが>"
	"とぅすい" n nom @amod #2->6
"<やら>"
	"やら" cop cvb @cop #3->2
"<また>"
	"また" dsc @discourse #4->6
"<わかある>"
	"わかさん" adj vz npst @amod #5->6
"<ちゅが>"
	"ちゅ" n nom @obj #6->8
"<やら>"
	"や" cvb @cop #7->5
"<わからにが>"
	"わかるん" v neg cvb nom @root #8->0
"<きいぬ>"
	"きい" n nom @obj #9->11
"<いいし>"
	"いい" adj all @advmod #10->11
"<ぬぶやあくん>"
	"ぬぶいん" v prog @conj #11->8
"<ばあよ>"
	"ばあよ" dsc @discourse #12->11
"<.>"
	"." sent @punct #13->11