Difference between revisions of "Khasi/Universal Dependencies"

From LING073
Jump to: navigation, search
(Dependency Relations)
 
(8 intermediate revisions by the same user not shown)
Line 24: Line 24:
 
| 52.59%
 
| 52.59%
 
|}
 
|}
 
 
 
==nomorph==
 
==nomorph==
 
{| class="wikitable"
 
{| class="wikitable"
Line 35: Line 33:
 
|-
 
|-
 
|annotated
 
|annotated
| 79.00%
+
| 87.70%
| 61.00%
+
| 80.33%
 
|-
 
|-
 
|annotated2
 
|annotated2
| 77.05%
+
| 45.93%
| 37.70%
+
| 40.74%
 
|}
 
|}
 
+
=Dependency Relations=
 
+
==@cc==
 
+
CC stands for coordinating conjunction.
annotated and with morph: Parsing from gold tokenization with gold tags - forms: 244, UAS: 85.66%, LAS: 78.69%
+
Khasi uses the word 'bad', meaning 'and' most frequently as a coordinating conjunction
 
+
In fact, there was no use of any of 'for, and, nor, but, or, yet, so' in the corpus I worked with.
annotated2 and with morph: Parsing from gold tokenization with gold tags - forms: 135, UAS: 57.04%, LAS: 52.59%
+
Coordinating conjunctions in Khasi may come at the beginning of the sentence or join clauses together.
 
+
Coordinating conjunctions point to the root
annotated and without morph: Parsing from gold tokenization with gold tags - forms: 244, UAS: 87.70%, LAS: 80.33%
+
Examples:
 
+
*"<.>"
annotated2 and without morph: Parsing from gold tokenization with gold tags - forms: 135, UAS: 45.93%, LAS: 40.74%
+
**"." sent @punct #35->6
 
+
*"<Bad>"
 +
**"bad" coocnj @cc #1->6
 +
And
 +
*"<,>"
 +
**"," cm @punct #12->18
 +
*"<bad>"
 +
**"bad" coocnj @cc #13->18
 
[[Category:sp17_UD]]
 
[[Category:sp17_UD]]
 +
==@det==
 +
Det stands for determiner.
 +
Khasi uses determiners in the context of articles and demonstratives
 +
These determiners modify their nouns.
 +
Examples:
 +
*"<ia>"
 +
**"ia" dirObj @case #7->10
 +
*"<kata>"
 +
**"kata" f sg dem dist osfar @det #8->10
 +
*"<ka>"
 +
**"ka" art f sg @det #9->10
 +
*"<jingshai>"
 +
**"jingshai" n f @obj #10->6
 +
And
 +
*"<<nowiki>U</nowiki>>"
 +
**"U" art m sg @det #19->20
 +
*"<Blei>"
 +
**"blei" n m @subj #20->23
 +
==@aux==
 +
Aux stands for auxiliary.
 +
Khasi uses aux most interestingly in the following ways:
 +
* past particles
 +
**"<la>"
 +
***"la" past @aux #5->6
 +
**"<pynlong>"
 +
***"pynlong" vbser @root #6->0
 +
* continued states or actions
 +
**"<da>"
 +
***"da" cont @aux #30->31
 +
**"<khih>"
 +
***"khih" vblex tv @conj #31->6
 +
* the imperative 'let', or in Khasi, 'to'
 +
**"<To>"
 +
***"to" imp @aux #8->9
 +
**"<long>"
 +
***"long" vbser @conj #9->6
 +
* A interesting case where the third person subject's personal pronoun is repeated before the verb
 +
**"<<nowiki>u</nowiki>>"
 +
***"u" prn pers p3 sg m @aux #9->8
 +
**"<la>"
 +
***"la" past @aux #10->11
 +
**"<khot>"
 +
***"khot" vblex tv @root #11->0
 +
==@case==
 +
case, as you might expect, can be used to mark case in languages with case systems, but it can also be used in languages without explicit case systems, but that have words functioning as if the case system was present. In Khasi, which doesn't explicitly have a case system, 'ia' can be considered an accusative marker, so case is used there:
 +
*"<ia>"
 +
**"ia" dirObj @case #8->10
 +
*"<ka>"
 +
**"ka" art f sg @det #9->10
 +
*"<bneng>"
 +
**"bneng" n f @obj #10->7
 +
Furthermore, words like 'halor' - prepositions - can take case. 'halor' means over.
 +
*"<halor>"
 +
**"halor" pr @case #32->34
 +
*"<ki>"
 +
**"ki" art GD pl @det #33->34
 +
*"<um>"
 +
**"um" n f @obl #34->31
 +
==@obl==
 +
Obl stands for oblique. Oblique can be used to mark temporal adverbs:
 +
*"<Te>"
 +
** "te" adv @obl #1->6
 +
*"<<nowiki>U</nowiki>>"
 +
** "U" art m sg @det #2->3
 +
*"<Blei>"
 +
** "blei" n m @nsubj #3->6
 +
As well as be the object of a case tag:
 +
*"<halor>"
 +
**"halor" pr @case #32->34
 +
*"<ki>"
 +
**"ki" art GD pl @det #33->34
 +
*"<um>"
 +
**"um" n f @obl #34->31

Latest revision as of 13:57, 25 April 2017

Evaluation

  • Number of sentences
    • annotated: 11
    • annotated2: 4
  • Number of forms
    • annotated: 244
    • annotated2: 135

withmorph

Scores
UAS LAS
annotated 85.66% 78.69%
annotated2 57.04% 52.59%

nomorph

Scores
UAS LAS
annotated 87.70% 80.33%
annotated2 45.93% 40.74%

Dependency Relations

@cc

CC stands for coordinating conjunction. Khasi uses the word 'bad', meaning 'and' most frequently as a coordinating conjunction In fact, there was no use of any of 'for, and, nor, but, or, yet, so' in the corpus I worked with. Coordinating conjunctions in Khasi may come at the beginning of the sentence or join clauses together. Coordinating conjunctions point to the root Examples:

  • "<.>"
    • "." sent @punct #35->6
  • "<Bad>"
    • "bad" coocnj @cc #1->6

And

  • "<,>"
    • "," cm @punct #12->18
  • "<bad>"
    • "bad" coocnj @cc #13->18

@det

Det stands for determiner. Khasi uses determiners in the context of articles and demonstratives These determiners modify their nouns. Examples:

  • "<ia>"
    • "ia" dirObj @case #7->10
  • "<kata>"
    • "kata" f sg dem dist osfar @det #8->10
  • "<ka>"
    • "ka" art f sg @det #9->10
  • "<jingshai>"
    • "jingshai" n f @obj #10->6

And

  • "<U>"
    • "U" art m sg @det #19->20
  • "<Blei>"
    • "blei" n m @subj #20->23

@aux

Aux stands for auxiliary. Khasi uses aux most interestingly in the following ways:

  • past particles
    • "<la>"
      • "la" past @aux #5->6
    • "<pynlong>"
      • "pynlong" vbser @root #6->0
  • continued states or actions
    • "<da>"
      • "da" cont @aux #30->31
    • "<khih>"
      • "khih" vblex tv @conj #31->6
  • the imperative 'let', or in Khasi, 'to'
    • "<To>"
      • "to" imp @aux #8->9
    • "<long>"
      • "long" vbser @conj #9->6
  • A interesting case where the third person subject's personal pronoun is repeated before the verb
    • "<u>"
      • "u" prn pers p3 sg m @aux #9->8
    • "<la>"
      • "la" past @aux #10->11
    • "<khot>"
      • "khot" vblex tv @root #11->0

@case

case, as you might expect, can be used to mark case in languages with case systems, but it can also be used in languages without explicit case systems, but that have words functioning as if the case system was present. In Khasi, which doesn't explicitly have a case system, 'ia' can be considered an accusative marker, so case is used there:

  • "<ia>"
    • "ia" dirObj @case #8->10
  • "<ka>"
    • "ka" art f sg @det #9->10
  • "<bneng>"
    • "bneng" n f @obj #10->7

Furthermore, words like 'halor' - prepositions - can take case. 'halor' means over.

  • "<halor>"
    • "halor" pr @case #32->34
  • "<ki>"
    • "ki" art GD pl @det #33->34
  • "<um>"
    • "um" n f @obl #34->31

@obl

Obl stands for oblique. Oblique can be used to mark temporal adverbs:

  • "<Te>"
    • "te" adv @obl #1->6
  • "<U>"
    • "U" art m sg @det #2->3
  • "<Blei>"
    • "blei" n m @nsubj #3->6

As well as be the object of a case tag:

  • "<halor>"
    • "halor" pr @case #32->34
  • "<ki>"
    • "ki" art GD pl @det #33->34
  • "<um>"
    • "um" n f @obl #34->31