퍼지정보검색시스템의 검색효율에 관한 실험적 연구 = (An) Experimental Study on Evaluating Effectiveness of Fuzzy Information Retrieval System

다국어 초록 (Multilingual Abstract)

Boolean retrieval systems have been most widely used among commercially available information retrieval systems. This is because uers are able to express their information need more conviniently and briefly by using a Boolean query formulation.
But, Boolean retrieval systems have some disadvanteges. First, document ranking is amenable because various similarity coefficients cannot be computed between queries and documents. Second, it cannot retrieve partially related documents because of the nature of Boolean operators.
Among the many retrieval techniques that have been developed to compensate for these weaknesses of Boolean retrieval, the fuzzy set model can solve the above defects. Fuzzy set theory generalizes traditional set theory by permitting partial membership in a set and makes it possible to recognize different degree to which a document can match a request.
This study compared the retrieval effectiveness of Boolean and fuzzy set retrieval strategies through a retrieval experiment. Recall and precision ratio was used to evaluate effectiveness of fuzzy set retrieval strategies. The fuzzy information retrieval system was designed for this experiment. This designed system has three modules-automatic indexing, keyword connection matrix construction, and fuzzy set retrieval modules. Experimental data consisted of 207 index terms, selected from 135 articles in library and information science journals containing Korean abstracts. Eight search queries were used for this retrieval experiment.
The results of this experiment are as follows.
(1) The precision ratio of fuzzy set retrieval is 63%, higher than the 52% of Boolean retrieval. On the other hand, the recall ratio of fuzzy set retrieval is 75%, a little bit lower than the 72% of Boolean retrieval.
(2) In Boolean retrieval systems, any relevant document which has 'and not' operator search terms as index terms cannot be retrieved. But, in fuzzy retrieval systems, these kinds of relevant documents can be retrieved.
(3) In order to improve the ratio of recall and precision of designed fuzzy retrieval system, the learning algorithm, which modifies the initial relation value of keyword connection matrix by using subject specialist's feedback, have to be devised and included in the system.

목차 (Table of Contents)

• 목차 = i
• 1. 서론 = 1
• 1.1 연구의 목적 = 2
• 1.2 연구의 방법 및 범위 = 3
• 1.3 용어의 정의 = 4
• 1.4 선행연구 = 7
• 1.4.1 외국의 선행연구 = 7
• 1.4.2 국내의 선행연구 = 10
• 2. 이론적 배경 = 11
• 2.1 자동색인모형 = 11
• 2.2 퍼지키워드 관계값 구축모형 = 14
• 2.2.1 퍼지키워드 관계행렬 = 14
• 2.2.2 퍼지시소러스 생성 알고리듬 = 15
• 2.3 퍼지집합검색모형 = 18
• 2.3.1 MIN-MAX 연산자를 이용한 퍼지집합모형 = 19
• 2.3.2 T-연산자를 이용한 퍼지집합모형 = 21
• 2.3.3 평균-연산자를 이용한 퍼지집합모형 = 23
• 3. 퍼지정보검색시스템 구축 = 27
• 3.1 실험환경 = 27
• 3.2 시스템 구축 = 28
• 3.2.1 자동색인 기능 = 28
• 3.2.2 퍼지키워드 관계행렬 구성기능 = 31
• 3.2.3 퍼지정보검색 기능 = 33
• 3.3 시스템구현 = 39
• 3.4 검색실험 = 41
• 3.4.1 질문의 구성 = 41
• 3.4.2 검색실험 = 42
• 3.4.3 검색결과 비교분석 = 42
• 4. 결론 = 45
• 부록 = 48
• 1. 실험데이타 = 48
• 2. 색인어 화일 = 94
• 3. 도치색인화일 = 96
• 4. 조사표 = 102
• 5. 조사형명사사전 = 102
• 6. 불용어 단어사전 = 102
• 7. 불용어 어간사전 = 102
• 참고문헌 = 104
• ABSTRACT = 109