전문용어의 의미관계 정보를 이용한 도메인 온톨로지의 구축 = Domain ontology construction based on semantic relation information of terminology

다국어 초록 (Multilingual Abstract)

An ontology is an explicit specification of a conceptualization. That is, an ontology is a description (like a formal specification of a program) of the concepts and relationships that can exist for an agent or a community of agents. This thesis suggests a method of constructing domain ontology semi-automatically using terminology processing and applies the method to document retrieval. In order to construct ontology, we propose an algorithm that classifies the patterns of nouns and suffices which compose terminology, in domain texts, extracts terminology, and build a hierarchical structure. The experiment used documents related to pharmacy domain. As singleton terms combined with specific nouns or suffices were identified, 2,864 sub-concepts were added and the algorithm showed accuracy of 92.57% was achieved on the average. In case of multi-word terms, 574 concepts were added and the average accuracy was 66.64%. Constructed ontology, which forms natural groups of senses centering on specific nouns or suffices composing the terminology with semantic information, can be utilized in approaching the knowledge of special areas such as document retrieval. According to the result of document retrieval based on the constructed ontology, the system improved precision by 4.97% compared to keyword-based document retrieval with traditional TFㆍIDF method. Also recall was improved by 0.78%. As stated above, Ontology constructed by analyzing texts in a specific domain adds concepts and relationships automatically. As a result, it maintains richer information, can answer various queries, and enhance the accuracy of retrieval. This suggests that concepts and rules defined in ontology may be used as the base of inference to improve the performance of retrieval. Subsequent research will be focused on how to apply the proposed method of ontology construction to general domain.

목차 (Table of Contents)

• 목차 = i
• 표 차례 = ii
• 그림 차례 = iii
• 1. 서론 = 1
• 1.1 연구 배경 = 1
• 1.2 연구의 동기 및 내용 = 3
• 2. 관련연구 = 6
• 2.1 온톨로지 = 6
• 2.1.1 온톨로지의 표현언어 = 9
• 2.1.2 온톨로지의 구축 사례 = 14
• 2.2 전문용어의 추출 = 17
• 2.3 정보 검색 시스템 = 20
• 2.3.1 검색 모델 = 20
• 2.3.2 재현율과 정확률 = 22
• 3. 도메인 온톨로지의 구축방안 = 24
• 3.1 온톨로지의 구축단계 = 24
• 3.1.1 온톨로지의 구조와 표현 = 27
• 3.1.2 기본 온톨로지 = 31
• 3.1.3 개념의 추출 = 32
• 3.1.4 관계의 추가 = 34
• 3.2 전문용어의 추출 = 38
• 3.2.1 단일 어절의 형태 = 40
• 3.2.2 다중 어절의 형태 = 43
• 3.3 온톨로지의 확장 = 46
• 4. 온톨로지의 활용 = 50
• 4.1 문서 검색에의 이용 = 50
• 4.2 문서의 순위화 = 53
• 5. 실험 및 평가 = 60
• 5.1 전문 용어 인식 = 60
• 5.2 검색 성능 평가 = 66
• 6. 결론 = 72
• 참고문헌 = 74
• Abstract = 79
• 부록1 불용어 리스트 = 81
• 부록2 단어 “중이염”의 텍스트별 가중치 = 85
• 부록3 10개 질의에 대한 재현율의 비교 = 86
• 부록4 10개 질의에 대한 정확률의 비교 = 87
• 부록5 Medicine Ontology (일부분) = 88