초록 ▼

5. 결론
○ 본 연구에서 개발한 지식맵은 신기술의 사전 타당성 평가에 매우 유용한 정보와 객관적인 근거를 제공할 수 있었으며, 현재의 전문가 회의(Peer review)방식과 지식맵 방식을 통합하는 형태의 새로운 사전 조정방식은 매우 유용할 것으로 판단됨
○ 사전 심의에 있어서 신기술의 특성상 적합한 전문가를 선정하기 곤란하였으나 주요한 동향이 객관적이고 이해하기 용이한 도식화 형태로 제공되기 때문에 현재 전문성 위주로 구성되는 순수 전문가 회의(Peer review)에 비해 보다 광범위한 이해 관계자들 즉 사용자, 정

5. 결론
○ 본 연구에서 개발한 지식맵은 신기술의 사전 타당성 평가에 매우 유용한 정보와 객관적인 근거를 제공할 수 있었으며, 현재의 전문가 회의(Peer review)방식과 지식맵 방식을 통합하는 형태의 새로운 사전 조정방식은 매우 유용할 것으로 판단됨
○ 사전 심의에 있어서 신기술의 특성상 적합한 전문가를 선정하기 곤란하였으나 주요한 동향이 객관적이고 이해하기 용이한 도식화 형태로 제공되기 때문에 현재 전문성 위주로 구성되는 순수 전문가 회의(Peer review)에 비해 보다 광범위한 이해 관계자들 즉 사용자, 정책관련자 등이 참여하는 사용자 평가(Merit review) 방식으로 전환시킬 수 있음
○ 지식맵 분석과정이 DB에 의해 진행되므로 일단 특정 신기술에 대한 분석이 완료된 후 추가적인 데이터의 입력만 이루어지면 지속적인 변화를 추적할 수 있으므로 신규 과제뿐만 아니라 단계 평가에 활용하면 매우 효과적일 것임
○ 본 연구에서는 지식맵을 사전 타당성 평가에 활용하고자 개발하였으나 지식맵은 연구관리 전 과정을 지원할 수 있는 매우 유용한 도구로서 활용할 수 있을 것이므로 향후 다음과 같은 분야에 활용 노력을 지속적으로 할 필요가 있음
- 연구기획서 작성, TRM 작성, 국제협력지도 작성, 기술분류체계 작성,경쟁국 기술동향 분석, 기술예측, Bio-informatics, 기술혁신조사(LBIS,Literature-based innovation survey), 과학기술 조기 경보시스템(S&T early warning system)구축 등에 핵심적인 도구로서 활용 할 수 있음
- 특히 현재 연구기획이 실시되고 있는 차세대 성장동력산업관련 핵심기술에 대해 지식맵 작성과 이의 DB화를 시도하면 적은 비용으로 정부,연구자는 물론 광범위한 이해 관계자들의 참여를 활성화시키고 추진타당성에 대한 합의 도출에 도움이 될 수 있을 것임

Abstract ▼

1. Introduction
Nowadays science and technology analyses appeared as standing on the basis of the scope and quantity of literature in the field of rapidly progressed science and technology. And using both of them, it is named "bibliometric" or "scientometrics". These analysis results answer emerg

1. Introduction
Nowadays science and technology analyses appeared as standing on the basis of the scope and quantity of literature in the field of rapidly progressed science and technology. And using both of them, it is named "bibliometric" or "scientometrics". These analysis results answer emergence in science and technology and provide researchers, responsible persons,technological information specialists, and planners with useful value.However, in much literature of science and technology, it is not easy for researchers to find study fields and link each study field together with giving difficulties to policy planners for schematizing changing aspects in science and technological fields when establishing study plans.
For this schematizing, several methods were tried. A traditional method used in science research and science policy is Peer Review to ask some specialists for opinions (Law & Whittaker, 1992). However, this method has weak points as follows. First, it takes much cost if research scale of experts is not big. Secondly, if research scale is small, representativeness becomes a problem. Thirdly, it is too complicated to contrast a scope of points of view determining if a science field has finished to be developed or still be in progress(He, 1999). Therefore, bibliometric methodology is another way to conduct this kind of work in terms of a standard capacity,and one of representative methods is Co-word analysis.
Traditional bibliometric methods such as co-citation analysis authors and journals were made on the basis of analysis on quotation databases contained science papers. This kind of analysis brings an interesting result while it cannot provide an immediate figure about actual contents of research themes using papers. Co-word analysis measuring co-occurrence and analyzing key words in papers dealing with a provided theme has the potential to handle this kind of analytical problem in brevity.Co-word analysis method simply shows data in a uniquely visualized way keeping basic information contained databases. This has fundament of a natural language as the provider of scientific notion, idea and knowledge(van Raan & Tijssen, 1993).
For long, many scholars have used Co-word analysis method when grasping changing aspects and doing scope analysis on various fields.There are researches of software engineering field(Coulter et al., 1998),Polimer chemistry(Callon et al., 1991), research of nervous ne (Noyons & van Rann, 1998a; van Raan & Tijssen, 1993), optical engineering(Noyons & van Raan, 1994) Bioelectronics(Hinze, 1994), pharmacy(Rikken et al., 1995),biology(Rip & Courtial, 1984; Looze & Lemarie, 1997), concentrated material physics(Bhattacharya & Basu, 1998).
Co-word analysis is based on the hypothesis that key words of a paper scribe properly its contents or connections of various problems, and consists of a paper. The case that two key words are extracted in one paper means the paper is connected with two mentioned themes(Cambrosio et al., 1998). The case the same word pair is written frequently means the research theme is bound of significance in papers.
Co-word analysis shows a pattern or a tendency of a specific field by measuring related strength between representative terms of related issues published in the specific field. The main form of Co-word analysis is to describe mental structure of the specific study field as a map of notional space of the field, or to visualize in a visible map by tracing changes in conceptual space.
Accordingly, the purpose of this research is to suggest schematized change aspects and new research directions hidden in bibliographical phenomenon of a paper in the science and technology field to science and technology researchers and policy planners. For this, the field this paper researches is the next generation OLED and BIOCHIP field. The reason to choose this field is due to the importance of the next generation OLED and BIOCHIP, and wide scope connected by various fields. Besides, to measure application and technological fields by using bibliographical method and technology is as important as in a basic research field.
2. Technology Mapping Method through Co-word analysis
Co-word analysis method should include ways to express interactions among subordinate structures and to select analysis field, choose analysis data and format, measure similarity and formality of co-occurrence frequency of words and cluster and map.
2.1. Data selection
It is desirable to select data in a document unit, rather than a journal unit in that a word is the most significant factor in Co-word analysis. There is two ways to extract these words from journal, a conference paper, a report or contents of books. First, we can extract words from keyword indices, themes, abstracts or data that have any analysis code. Many journals, a paper abstract service, and databases already provide those key words. Cambrosio et al.(1993) extracted a key word after adding several indices and theme words to supplement the quality caused by the lack structure of data which would be used in their hands for study. Descriptive words provided from the result were standardized expressions differently formed when using different spellings and the same term. Coulter et al.(1998) used descriptive language gotten from professional index writer. They believed this way would be very useful when conducting the research of the fixed system requiring ordinary names. Experiments of professional index writers give us conviction that they use taxologically standardized tools. Looze & Lemarie(1997) went through Co-word analysis based on key words suggested by experts. Several researchers were given key words from online databases added by index writer and authors(Courtial, 1994; Law & Whittaker, 1992; Courtial et al., 1994). Noyon & van Raan(1998b) diagrammatized the whole unqualified structure of the neural network with utilizing synchronous appearance of a classified code.
The most important condition for collecting databases from these regulated terms is the possibility of "indexer effect". In other words, when collecting key words by professional index writer working with other databases, the problem is that their predictions or points of view can be reflected when selecting words (King, 1978). However, this effect can be offset by the rightly performed interview (Law & Whittaker, 1992;Cambrosio et al., 1993; Tijssen, 1993; Courtial, 1994).
The second data collection method is to extract directly words from the whole text using software such as Nptools(Voutilainen, 1993). Words of proper frequency or phrases are selected as a subject of Co-word analysis method to present a core theme in a specific field. These methods are chosen not to meet difficulties or problems of time consumed for repairing and keeping a classifying related field that would be developed, useless efforts of index composers, and consideration for time-consuming of constructing classifying system and index searching system memorized in a computer.
2.2. Similarity matrix creation
In Co-word analysis method, once a research theme is chosen, next step is to compose co-occurrence matrix with having a basis of co-occurrence of words. The cell value of two words is determined by the number of times for two words to take up at the same time in one context. Highly leveled co-occurrence frequency occurs between two words displays there is a close relationship of them(Ding et al., 2001). Co-occurrence matrix is possibly used as matrix of the first similarity degree.
However, as usual similarity factors are used for composing matrix of similarity degree, the purpose of using similarity factor is for the ability to standardize an appearing frequency difference between frequent occurring words and less occurring words by means of regulating scopes of co-occurrence frequency. In according to characteristics of consisting components, these similarity factors are divided into two kinds: for weight vector and for binary vector(Callon et al., 1986; Peter & van Rann, 1993; Coulter et al., 1998). Also divided into two kinds according to levels of word occurrence frequency: preference for high frequency and preference for low frequency (Peter and van Rann, 1993). Currently, Ding et al.(2001) measured the degree of word association using Pearson relation coefficient.Pearson coefficient r is valued from -1 to 1 and his result is gained from using co-occurrence frequency of two words and other words in Co-word matrix to measure similarity between two words. However, it is not clear what similarity coefficient is good for being selecting to present current technology status and ongoing developing status.
2.3. Clustering
Once composing a matrix of similarity degree, we do clustering to classify terms and documents. So far, the clustering method usually used in Co-word analysis has two kinds: Principal Component Analysis (PCA; for example, Porter 1999), and Hierarchical Clustering(for example, van Rann,199; Ding et al., 2001).
Main component analysis is the method to reduce multi-dimensional data up to data in two or three dimension aspect together with minimizing loss of data. We can visually find where an observed object is located with using main component analysis. Also, main component analysis composes the total index containing lots of indices, divides the observing stuff into several groups, and is used in the purpose of reading data for the multiple regression analysis or discriminating analysis in other points of view.
Hierarchical clustering is already existed clusters that are combined by the distance value step by step to make a bigger clustering, and is possibly presented by dendrogram. In this hierarchical clustering, there are four kinds: single linkage that is a way of connecting components of clusters only using one linage, complete linkage that uses the pair in the lowest degree of similarity to determine a degree among each cluster, group average that determines an similarity degree of clusters as the average similarity degree of all text included in a specific cluster, Ward's method that produces clusters with minimizing the Euclid distance between each cluster centrality.
Based on these clusters, we compose cluster co-occurrence matrix, and also the second similarity matrix. The composing way is the same as mentioned above. One difference is the basis of composition is the word contained a cluster.
2.4 Mapping
There are some ways to map data. The most generally used way is Multidimensional scaling. It places objects in multidimensional space in accordance with the degree of similarity. Things in low degree are placed far from each other, things in high degree are placed closely.
Other methods are surely used for special programs. LEXIMAPPE program was invented as science policy instrument for Co-word mapping,and already has been used when analyzing issues in various research fields(Looze & Lemarie, 1997; Law & Whittaker, 1992; Cambrosio et al.,1993; Courtial, 1994); Content Analysis and Information Retrieval(CAIR) was developed by Software Engineering Institute at Caregie Mellon University,and used researchers working for Co-word analysis(Courlter et al., 1998);Astrian Research Centers invented Bibliometric Technology Monitoring(BibtechMon) and it is one of software developed for Co-word analysis.
Neural network algorism of Kohonen is another accessing method to diagrammatize data. Polanco et al.(1998) applied artificial neural network technology including Associative network with unsupervised learning(KOHNRN) to diagrammatizing and to analyzing the fields of science and technology information. WEBSOM Research group is one of the examples. They constructed Self-Organizing Map interface in the basis of web and with using it,diagrammatized words from lots of papers(Kohonen, 1995). WEBSOM performed the whole content analysis of a paper set not controlled and totally naturally occurred by using Self-Organizing Maps.
This analysis result, Map with arranged gap of documents presents directly the similarity relation of contents that are themes of papers. On Document map, it shows distances of relations. Within a text, the degree of concentration in different places can be indicated by different colors when presenting a literal map(Honkela et al., 1996).
Like this, if finishing drawing multidimensional scaling map, we need to the degree of clusters on the basis of words contained in clusters. Here we have two ways. First, Callon et al.(1991) used transformation index t to measure dissimilarity, and Coulter(1998) found the degree of similarity among networks that have different periods with using SI(Similarity index) derived from a numerical alteration index. Next, Peter & van Rann(1993) suggested a way of calculating degree of similarity among clusters with using cosine coefficient in the use of weight.

목차 Contents

• 표지 ... 1
• 서언 ... 2
• 목차 ... 4
• 표차례 ... 8
• 그림차례 ... 10
• 요약 ... 13
• Ⅰ. 서론 ... 40
• 제1절 연구의 배경 및 목적 ... 40
• 1. 연구의 배경 ... 40
• 2. 연구의 목적 ... 42
• 제2절 연구의 내용 ... 43
• Ⅱ. 신기술 국가연구개발사업의 사전 타당성 평가 현황 및 개선 방안 ... 46
• 제1절 사전 타당성 평가의 범위 및 중요성 ... 46
• 1. 국가연구개발사업의 추진단계와 사전 타당성 평가 ... 46
• 2. 신기술 개발과정에 있어서 사전 타당성 평가의 역할 ... 47
• 3. 새로운 사전 타당성 평가 방법의 필요성 ... 52
• 제2절 국내외 사전 타당성 평가의 현황 ... 54
• 1. 우리나라의 사전 타당성 평가 현황 ... 54
• 2. 선진국의 사전 타당성 평가 및 평가 기준 ... 65
• 3. 일본 ... 73
• 4. 유럽연합(EU) ... 75
• 제3절 우리나라 사전 타당성 평가 기준의 검토 ... 76
• 1. 사전 타당성 평가의 문제점 ... 76
• 2. 개선 방향 ... 78
• Ⅲ. 연구기획 사전타당성 분석을 위한 새로운 분석방법:지식맵(Knowledge Map)과 지표(S&T Indicator)개발 ... 80
• 제1절 데이터 수집 ... 82
• 1. 주제영역 결정 및 문헌 수집 ... 82
• 2. 색인어 수집 ... 87
• 제2절 유사도 행렬 작성 ... 92
• 1. 동시출현행렬 ... 92
• 2. 유사계수 ... 93
• 제3절 클러스터링 ... 99
• 1. 요인분석 ... 99
• 2. 계층적 클러스터링 ... 102
• 3. 네트워크 분석 ... 105
• 제4절 매핑 ... 109
• 1. 다차원척도법 ... 110
• 2. 소프트웨어를 이용한 방법 ... 117
• Ⅳ. OLED 분야의 사전타당성 분석을 위한 지식맵 및 지표개발 결과 ... 118
• 제1절 유기발광다이오드(Organic Light Emitting Diodes: OLED) ... 118
• 1. 유기발광다이오드(OLED)의 개요 및 중요성 ... 118
• 2. 국내외 기술개발동향 ... 123
• 3. OLED 시장 현황 ... 125
• 4. OLED 업체 현황 ... 131
• 5. OLED의 향후 전망 ... 135
• 제2절 세계 기술환경 및 우리나라 수준으로 본 도전 가능성 ... 136
• 1. 세계 연구개발자원의 분포와 우리나라의 위치 ... 136
• 2. 세계와 우리나라 연구개발자원의 질적수준 비교 ... 152
• 3. 선진국과 우리나라의 연구개발 위치 비교 ... 164
• 4. Domain Map을 통한 국가 간 R&D 분야 구조 및 상호연관성 격차의 비교 ... 167
• 5. 도전가능성 및 전략기술 ... 175
• 제3절 정부지원의 타당성 ... 178
• 1. 산․학․연 주체별 연구활동의 특성 비교 ... 179
• 2. 우리나라와 미국의 정부지원 개입시기 및 지원전략 비교 ... 182
• 3. 정부지원 타당성 및 향후 과제 ... 195
• 제4절 국제협력 전략의 적합성 ... 197
• 1. 국가 간 국제협력 현황 ... 200
• 2. 국가협력체계의 연도별 변화 ... 202
• 3. 기관 간 협력체계의 연도별 변화 ... 204
• 4. 연구자 간 국제협력 현황 ... 207
• 5. 연구자 간 협력체계의 연도별 변화 ... 209
• 6. 국제협력 전략 및 향후 과제 ... 212
• V. Biochip분야의 사전타당성 분석을 위한 지식맵 및 지표개발결과 ... 213
• 제1절 바이오칩 기술의 개요 및 중요성 ... 213
• 1. Biochip이란? ... 213
• 2. 바이오칩의 분류 및 특징 ... 216
• 3. 바이오칩의 국내․외 기술개발동향 ... 229
• 4. 바이오칩의 시장 현황 ... 230
• 제2절 세계 기술환경 및 우리나라 수준으로 본 도전 가능성 ... 238
• 1. 세계 연구개발 자원의 분포와 우리나라의 위치 및 질적수준 비교 ... 240
• 2. 선진국과 우리나라의 연구개발 위치 비교 ... 244
• 3. Domain Map을 통한 국가간 R&D 분야 구조 및 상호연관성 격차의 비교 ... 248
• 4. Biochip의 도전 가능성 ... 255
• 제3절 정부지원의 타당성 ... 257
• 1. 산․학․연 주체별 연구활동의 특성 비교 ... 258
• 2. 우리나라와 미국의 정부지원 개입시기 및 지원전략 비교 ... 260
• 3. 정부지원의 타당성 및 향후 방안 ... 270
• 제4절 국제협력 전략의 적합성 ... 271
• 1. 국가간 협력체계의 변화와 현황 ... 272
• 2. 기관간 협력체계의 변화와 현황 ... 275
• 3. 연구자간 협력체계의 변화와 현황 ... 277
• 4. 국제협력 전략 및 향후 방안 ... 278
• Ⅵ. 결론 및 향후 연구과제 ... 280
• 제1절 결 론 ... 280
• 제2절 향후 연구방향 ... 283
• 참고문헌 ... 285
• SUMMARY ... 311
• CONTENTS ... 319
• 끝페이지 ... 321