Difference between revisions of "Jeju/Disambiguation"
From LING073
(→Final evaluation of ambiguity in the corpus) |
(→Ambiguous forms) |
||
| (4 intermediate revisions by the same user not shown) | |||
| Line 9: | Line 9: | ||
ᄒᆞ나 can tagged either as | ᄒᆞ나 can tagged either as | ||
| − | + | ᄒᆞ<v><dec><pres> | |
or | or | ||
| − | + | ᄒᆞ<v><csfcoo><pres> | |
In fact, the ambiguity can be generalized to | In fact, the ambiguity can be generalized to | ||
| − | + | (VerbRoot)<v><dec><pres> | |
or | or | ||
| − | + | (VerbRoot)<v><csfcoo><pres> | |
The following are sentences in which one of the taggings is the correct tagging. The first is from | The following are sentences in which one of the taggings is the correct tagging. The first is from | ||
the corpus, and the second is made up. But the first sentence contains the two possible taggings | the corpus, and the second is made up. But the first sentence contains the two possible taggings | ||
of the general (VerbRoot) case.(먹<v><dec><pres> and ᄒᆞ<v><csfcoo><pres>) | of the general (VerbRoot) case.(먹<v><dec><pres> and ᄒᆞ<v><csfcoo><pres>) | ||
| − | + | ^보끈/볶<v><det>$ ^보릴/보리<n><acc>$ ^ᄀᆞᆯ앙/골<v><inf>$ ^멘드는디/만들<v><top><csfsub>$ ^ᄋᆢ름의/ᄋᆢ름<n><dat>$ ^물에/물<v><dat>$ ^캉/캉<adv>$ ^들이쌈을/들이쌈<n><acc>$ ^ᄒᆞ나/ᄒᆞ<v><dec><pres>/'''ᄒᆞ<v><csfcoo><pres>'''$ ^밥에/밥<n><dat>$ ^비병도/비비<v><add>$ ^먹나/'''먹<v><dec><pres>'''/먹<v><csfcoo><pres>$. | |
| − | + | ||
| + | ^살인도/살인<n><add>$ ^ᄒᆞ나/'''ᄒᆞ<v><dec><pres>'''/ᄒᆞ<v><csfcoo><pres>$. | ||
=Disambiguation= | =Disambiguation= | ||
| Line 56: | Line 57: | ||
==Final evaluation of ambiguity in the corpus== | ==Final evaluation of ambiguity in the corpus== | ||
| − | 7019 | + | |
| − | 6321 | + | * Analyses: 7019 |
| − | + | * Tokens: 6321 | |
| + | *Ambiguity level: ~1.11042556557506723620 | ||
| + | |||
| + | As a result of the added rule, the ambiguity level has decreased by approximately 0.0474 percent. | ||
[[Category:Jeju]] | [[Category:Jeju]] | ||
[[Category:sp17_Disambiguation]] | [[Category:sp17_Disambiguation]] | ||
Latest revision as of 22:25, 2 March 2017
Contents
Identified Ambiguities
Initial evaluation of ambiguity in the corpus
- Analyses: 7022
- Tokens: 6321
- Ambiguity level: ~1.11090017402309761114
Ambiguous forms
I have identified an ambiguity, between two possible tagging of ᄒᆞ나.
ᄒᆞ나 can tagged either as
ᄒᆞ<v><dec><pres>
or
ᄒᆞ<v><csfcoo><pres>
In fact, the ambiguity can be generalized to
(VerbRoot)<v><dec><pres>
or
(VerbRoot)<v><csfcoo><pres>
The following are sentences in which one of the taggings is the correct tagging. The first is from the corpus, and the second is made up. But the first sentence contains the two possible taggings of the general (VerbRoot) case.(먹<v><dec><pres> and ᄒᆞ<v><csfcoo><pres>)
^보끈/볶<v><det>$ ^보릴/보리<n><acc>$ ^ᄀᆞᆯ앙/골<v><inf>$ ^멘드는디/만들<v><top><csfsub>$ ^ᄋᆢ름의/ᄋᆢ름<n><dat>$ ^물에/물<v><dat>$ ^캉/캉<adv>$ ^들이쌈을/들이쌈<n><acc>$ ^ᄒᆞ나/ᄒᆞ<v><dec><pres>/ᄒᆞ<v><csfcoo><pres>$ ^밥에/밥<n><dat>$ ^비병도/비비<v><add>$ ^먹나/먹<v><dec><pres>/먹<v><csfcoo><pres>$.
^살인도/살인<n><add>$ ^ᄒᆞ나/ᄒᆞ<v><dec><pres>/ᄒᆞ<v><csfcoo><pres>$.
Disambiguation
Added constraints
- Proposed rule in prose
Observe that one of the possible tagging is a conjunctive suffix. Also, note that Jeju sentences usually end with a verb. Since the other tagging is a verb that ends a sentence, it is possible to say a form is (VerbRoot)<v><dec><pres> if there a punctuation mark after it.(meaning the sentence has ended.)
- Proposed rule in rlx file Constraint Grammar form
The following code only contains relevant parts.
SETS # defining POS into sets LIST CONJUNCTION = (v csfcoo pres) ; LIST NOUN = (n sg) (n pl) ; LIST VERB = (v dec pres) ; LIST NONEMPTY = "^(?!\s*$).+" ; LIST STOP = "." ; LIST NUMBER = (num ord) ; LIST ALL = CONJUNCTION VERB NOMEPTY ; SECTION # If there is nothing to the right, it cannot be a conjunction. REMOVE CONJUNCTION IF (0 CONJUNCTION) (0 VERB) (1 STOP) ; # If there is something to the right, it cannot be a verb ending. REMOVE VERB IF (0 CONJUNCTION) (0 VERB) ;
Final evaluation of ambiguity in the corpus
- Analyses: 7019
- Tokens: 6321
- Ambiguity level: ~1.11042556557506723620
As a result of the added rule, the ambiguity level has decreased by approximately 0.0474 percent.
