Okinawan/Disambiguation
From LING073
Initial Ambiguity
Before this disambiguation rule was implemented, there were roughly 1.09808917197452229299 analyses per word.
CG rule
Consider the following two sentences:
- じょうきんし いくん. (I walk towards the dustcloth.)
- じょうきんし ぬぐいん. (I wipe with the dustcloth.)
In the first sentence, the case marker し marks the allative case, while it marks the instrumentative case in the second. They analyze as follows:
- ^じょうきんし/じょうきん<n>/じょうきん<n><all>$ ^いくん/いくん<v><npst>$^./.<sent>$^./.<sent>$
- ^じょうきんし/じょうきん<n><ins>/じょうきん<n><all>$ ^ぬぐいん/ぬぐいん<v><npst>$^./.<sent>$^./.<sent>$
The homophonous し means that じょうきんし can be analyzed as either allative or instrumentative. Notably, we would like to select <all> in the first case, and <ins> in the second case. Therefore, we can identify two situations where the word is definitely allative:
- If the word itself is a pronoun or locational nominal, remove <ins>.
- If the verb following it is directive, remove <ins>.
After implementing CG versions of these rules, disambiguity dropped a shave to ~1.09554140127388535032. Clearly, (unless the ambiguity is very common), disambiguation is a painstaking process.
