Natural language processing systems repeatedly have to solve word-sense ambiguity problem. The word-sense ambiguity problem is the selection of the intended word-meaning of a word from the set of its possible meanings.
To solve the problem, it is necessary to use several sources of knowledges such as lexical knowledge, morphlogical knowledge, syntactic knowledge, semantic knowledge, various kinds of contextual knowledge etc. This knowledge sources must concurrently not sequentially participate in parsing input sentence and must interact with each others to help disambiguating the word meanings.
Considering this problem, in this thesis, a word meaning selection system is designed. By building a language comprehension model which maps input sentences ultimately into internal meaning representations, a knowledge-source based hierarchical multiprocess word meaning selection system is designed with respect to the model. It is partially implemented using the PEARL Al package. The formalization of word meaning selection in a context using schemata is also presented.
자연언어 처리시스템은 항상 단어의미의 모호성 문제를 해결해야만 한다. 단어의미의 애매모호성 문제는 한 단어가 가지는 모든 가능한 의미로부터, 단 하나의 의도하는 의미를 선택하는 문제이다. 이 문제를 해결하기 위해서는, lexical, morphological, syntactic, semantic knowledge 그리고 여러종류의 contextual knowledge등의 knowledge source들이 필요하다. 이러한 Knowledge Source들은 입력문장을 parsing하는데 있어서, 순차적인 방식이 아니라 동시에 관여하여서 단어의 애매성 문제를 해결하는데 있어서 상호작용해야 한다.
이러한 관점하에서, 지식구조를 이용한 계층적(hierarchical)인 멀티프로세스 단어의미 선택 시스템을 설계하고 PEARL Al package를 이용하여 부분적으로 implement하였다.