In technical domain, documents include many technical terminologies, which degrade the accuracy of morphological analyzer and lower the system performance of technical document processing. For these defects, the system requires a method of treating the technical terminologies. Since the previous researches for handling the unknown words, however, focus on the common words, their methods are inadequate for processing the technical terminologies. Even if we simply add the technical terminologies into the dictionary for morphological analysis, it is difficult to cover newly created technical terminologies. The technical terminologies can give rise to two problems during the process of morphological analysis. The one is a failure of morphological analysis, and the other is a wrong analysis by common words.
In this thesis, we propose a morphological analyzer dealing with both problems for the technical documents. For the first problem, our system recognizes the technical terminologies with the foreign word information and the word formation patterns of the technical terminologies. And, for the second one, the system make corrections to wrong morphological analyses with the domain relative words and domain irrelative words extracted from technical domain corpus. Experimental results show that our system performs better in the technical domain than morphological analyzer with the method processing general unknown words and that with dictionary added the technical terminologies.
