It is a very critical problem to determine the scope of parallel structure in Korean noun phrase coordination. The problem is selecting the correct syntactic structure formed by an individual coordiante conjunction from all possible candidates.
Although it is difficult to select the right coordinate conjunctive structure, conjunctions show the tendency to conjoin the same kind of syntactic patterns, called "parallelism". Kurohashi used the similarity for analyzing conjunctive structure, and this method is found effective. While Korean language has several coordinate conjunction structures which are very similar to thoes of Japanese, applying directly this method doesn't produce good result. So, we examine the cause of failure of analysis.
This thesis presents the coordinate structure scope analyzer using the information extracted from tree tagged corpus. The current rules are based on the Part-of-Speech strings and use the longest-matching method. They can be extended to fix the errors occurred in the real sentences. In order to resolve the ambiguity of Korean noun phrase coordination, we need more information such as lexical or semantic information.
At current state, this analyzer shows nearly equivalent accuracy as that of the analyzer based on the parallelism on the detection of the scope. The presented analyzer is expected to get a good result with the information about resolution of the ambiguity.
