(A) study on future core technology for defense weapon system automatic concept extraction using LDA and inverse term popularity rank

다국어 초록 (Multilingual Abstract)

This study aims to automatically extract future core technology keywords for defense weapon system using LDA topic modeling and Inverse Term Popularity Rank, and quantitative analysis and performance evaluation of interconnection relationships. Due to the nature of closedness and security in the defense field, the sharing, analysis, and field recycling of technical information are limited, and there are not many previous studies on experimental quantitative trend analysis such as text mining. Many previous studies are mainly using qualitative methods such as Delphi techniques to concentrate experts' opinions, which is feared to be focused on specific weapons systems and core technologies. Due to security, the complexity of the data collection procedure, the increase in the time and economic cost of mobilizing experts, and the distortion caused by the excessive influence of a small number of experts are also limitations. As a way to supplement them, various key technologies that are currently under R&D or have plans were collected through various methods. First, based on the Defense Acquisition Program Act and the Defense Technology Innovation Promotion Act, the Defense Technology Innovation Basic Plan, the Defense Technology Investigation Report, the Defense Technology Innovation Implementation Plan, and the Defense Technology Plan. Second, the Defense Science and Technology Information Magazine (last 5 years and 8 months, a total of 60 issues) published by the Defense Science and Technology Quality Institute was used as data. Morphological analysis was performed by pre-processing the data collected through torque aging. By implementing Automatic Concept Extraction using LDA topic modeling, meaningful Key Phrases for the core technology of the defense weapon system were extracted. And performance evaluation was conducted with LDA modeling and B-Cubed Precision & Recall Metrics as a result of direct investigation and organization of future defense core technologies of the expert group (Doctor 2 and Master 3) for more than 20 years in the defense field. For improving F-Score, the Inverse Term Popularity Rank was applied to give weight by Google search amount.
Based on the results of this study, it is possible to raise the need to derive quantitative trends for future core technology for defense weapon system by using Automatic Concept Extraction. Moreover, it is possible to overcome the time and economic limitations of security and mobilization of experts, which are special characteristics of the defense field, while redefining the classification system of future core technology for defense weapon system. Along with the theoretical implications of the long-term defense weapon system development, contributing in national defense competitiveness enhancement within international community is intended.

국문 초록 (Abstract)

본 연구는 LDA 토픽모델링과 Inverse Term Popularity Rank을 활용하여 국방 무기체계 핵심기술 키워드를 자동화 추출하고 상호간 연결 관계의 정량적 분석 및 성능평가를 하고자 한다. 국방분야의 폐쇄성, 보안성이라는 특성상 기술정보의 공유, 분석, 현업 재활용이 제한되어 텍스트마이닝 등 실험적인 정량적 동향 분석에 대한 선행연구들을 많이 찾을 수 없는 상황이다. 다수의 선행연구들이 델파이 기법 등 정성적인 방법으로 전문가들의 의견을 집약하는 방식이 주로 활용되고 있어 특정 무기체계 및 핵심기술에 치중될 우려가 있다. 또한, 보안성으로 인하여 데이터 수집 절차의 복잡성과 전문가 인력 동원의 시간 및 경제적 비용 증가, 소수 전문가의 과도한 영향력 발휘에 의한 왜곡 등도 한계점으로 들 수 있다. 이들을 보완하기 위한 방법으로 현재 연구개발 중이거나 계획이 있는 여러 핵심기술들을 다양한 방법을 통해 수집했다. 첫째, 방위사업법과 국방과학기술혁신 촉진법에 근거하여 방위사업청과 출연기관 등에서 발간한 무기체계 핵심기술에 대해 검토한 문서로서 국방과학기술혁신 기본계획, 국방과학기술조사서(총론 및 각 무기체계 분야별), 국방과학기술혁신 시행계획, 국방기술기획서 및 각각의 부록 자료이다. 둘째, 국방과학기술품질원에서 발간하는 국방과학기술정보지(최근 5년 8개월, 총 60개 호)를 데이터로 활용하였다. 토크나이징(Tokenizing)을 통해 수집한 데이터를 전처리하여 형태소 분석을 실시하였다. 그리고 LDA 토픽모델링을 이용하여 Automatic Concept Extraction을 구현하여 국방 무기체계 핵심기술에 대한 유의미한 KeyPhrase들을 추출했다. 그리고 국방분야 20년 이상 전문가 그룹(박사2, 석사 3)의 미래 국방 핵심기술에 대해 직접 조사·정리한 결과와 LDA 모델링과 B-Cubed Precision & Recall Metrics로 성능평가를 수행하였다. 추가로 F-Score 향상을 위해 Inverse Term Popularity Rank를 적용하여 구글 검색량에 의한 가중치(weight)를 부여하였다.
본 연구의 결과를 바탕으로 국방 무기체계 핵심기술 Automatic Concept Extraction을 활용하여 핵심기술에 대한 정량적 동향을 도출해낼 필요성 제기할 수 있다. 또한 국방분야의 특수성인 보안성과 전문가 인력 동원의 시간 및 경제성의 제한사항을 극복함과 동시에 다가올 미래 새로운 무기체계 핵심기술의 분류체계를 재정립할 수 있다. 아울러 국방 무기체계 핵심기술의 정책적 시사점과 연구의 한계점을 제시한다. 장기 무기체계 발전방향의 이론적 함의와 함께 국제사회의 국방 경쟁력을 제고하는데 기여하고자 한다.

목차 (Table of Contents)

• 1. Introduction 1
• 1.1. Background 1
• 1.2. Study Purpose 7
• 1.3. Methodology and Organization of the Study 13
• 2. Theoretical Background 16
• 2.1. Current Korea Security Status and Changes in Future Warfare 16
• 2.1.1. Current Korea Security Status 16
• 2.1.2. Changes in Future Warfare Scenario 18
• 2.2. Defense Weapon System Core Technology 21
• 2.2.1. Definition of Defense Weapon System 21
• 2.2.2. Characteristics and Importances of Defense Weapon System 21
• 2.2.3. Definition of Defense Core Technology 24
• 2.3. Automatic Concept Extraction 25
• 2.3.1. Big Data & Text Mining 26
• 2.3.2. Tokenizing 27
• 2.3.3. Automatic Concept Extraction Precedent Study 28
• 2.4. LDA Topic Modeling 33
• 2.4.1. Concept of LDA Topic Modeling 33
• 2.4.2. Precedent Study of LDA Technique 37
• 3. Research Method 40
• 3.1. Research Subject and Collecting Data 40
• 3.2. Development Environment System for LDA 45
• 3.3. LDA Modeling Framework Concept 48
• 3.3.1. Step 1 : Collection and File Conversion 48
• 3.3.2. Step 2 : Pre-processing 49
• 3.3.3. Step 3 : LDA Topic Modeling 50
• 3.3.4. Step 4 : Performance Evaluation 53
• 3.3.4.1. Concept of B-Cubed Metrics and Precedent Study 53
• 3.3.4.2. Performance Evaluation : Comparison with Baselines 54
• 4. Analysis Results 58
• 4.1. LDA Implementation Results 58
• 4.2. BaseLines Model for Comparison of Performance Evaluation Results 60
• 4.2.1. TextRank 60
• 4.2.2. TopicRank 61
• 4.2.3. RAKE(Rapid Automatic Keyword Extraction) 62
• 4.3. Performance Evaluation Results 63
• 4.4. Results of Automation Concept Extraction Using LDA and
• Presentation of Future Strategic Technologies 67
• 4.5. Research Contributions 74
• 5. Conclusion 78
• 5.1. Research Summary 78
• 5.2. Significance and Implications of the Study 81
• 5.3. Limitations of the Study 86
• 5.2.1. Technical Limitations in Implementing LDA 86
• 5.2.2. Limitations of Research Procedures 86
• References 89
• Abstract 106

참고문헌 (Reference)

1 Neville, P., "Growing trees for stratified modeling", 30, 528-533, 1998

2 Cho, T. M., Lee, J. H., "Latent Keyword Extraction Using LDA Model", 25(2), 180-185, 2015

3 Lee, C. H., "Future Army Super-Connected New Weapon System", 521, 74-86, 2020

4 Han, H. N., Sin, N. R., Kim, Y. J., "Text Mining Analysis of Blockchain Logistics Industry", 12(5), 2567-2578, 2021

5 Kim, M., "Topic Modeling on Fine Dust Issues Using LDA Analysis", 29(2), 23-29, 2020

6 Choi, H., Lee, H., Kim, T., "A study on the research trends in fintech using topic modeling", 17(11), 670-681, 2016

7 Ahn, J. Y., Na, S. T., Kim, J. H., Jung, M. H., "Smart grid research trend analysis using dynamic topic analysis", 6, 613-620, 2017

8 Ko, W., "Changes in the War Paradigm and Implications for the Korean Army", 90(0), 9-46, 2011

9 Kim, M. C., Yoon, S. W., "Topic Modeling Analysis of Fine Dust Issues Using LDA Techniques", 29(2), 23-29, 2020

10 Kim, H., Kim, W., "A Study on Automatic Analysis System of National Defense Articles", 21(1), 86-93, 2018

11 Kim, S. M., Park, Y. K., Lee, H. W., "Identification of Failure Types of Weapons Using Machine Learning", 19(8), 64-70, 2018

12 Yoo, S. M., "ICBM Industry Development Plan and Technology Introduction Strategy", 12(2), 9-16, 2014

13 Erol, S., Sihn, W., Schumacher, A., "Strategic guidance towards Industry 4.0–a three-stage process model", International conference on competitive manufacturing, 9(1), 495-501, 2016

14 Gambatese, J. A., Hallowell, M. R., "Qualitative research: Application of the Delphi method to CEM research", 136(1), 99-107, 2010

15 Lee, W. J., "The Fourth Industrial Revolution and Changes in Management and Marketing", 22(1), 177-193, 2018

16 Hwang, S. J., You, S. T., Park, K. S., Lee, Y. S., Kim K. S., "An Analysis of Industrial Security Trends in Korea Using LDA Topic Modeling", 10(2), 79-103, 2020

17 Han, E. J., Park, J. S., Kim, N. R., "Analysis of trends in science and technology using keyword network analysis", 23(2), 63-73, 2018

18 Hong, S. G., Park, J. S., Kim, J. W., "A Study on the Trend and Prediction of Science and Technology Using Topic Modeling", 22(4), 19-28, 2017

19 Brusilovsky, P., Thaker, K., Labutov, I., He, D., Chau, H., "Automatic concept extraction for domain and student modeling in adaptive textbooks", 31(4), 820-846, 2021

20 Jo, W. S., Yoo, H. K.,, Nam, E. M., Na, J. H.,, "Prediction Model of Slope Collapse through Quantification of Qualitative Variables", 14(5), 281-288, 2009

21 Yoo, S. H., "Weapon System Exploring the Future Electronic Warfare Development Plan of our Army", 507, 94-99, 2021

22 Jung E. H., "A Study on the Spatial Perception of Lotte World Tower through Text Mining Analysis", 16(5), 203-211, 2021

23 Amigo, E., Verdejo, F., Gonzalo, J., Artiles, J., "A comparison of extrinsic clustering evaluation metrics based on formal const-raints", 12(4), 461-486, 2009

24 Kim, J. Y, Yang, J. Y.,, "A study on public perception of wales Millenium Centre architecture using textmining", 12(5), 193-201, 2017

25 Kim, C. S., Song, M. J., "An Empirical Analysis of International Evaluation Conditions between Korea and China", (8), 109-135, 2010

26 Jo, H. S., "Artificial Intelligence, Autonomous Weapons System and Future Warfare Transformation", 28(1), 115-140, 2018

27 Kang, H. J., Kim, C., Kang, K., "Analysis of the trends in biochemical research using Latent Dirichlet Allocation (LDA)", 7(6), 379, 2019

28 Kim, D. H., "An Empirical Study on Future New Technologies in the Field of Defense Unmanned Robotics", 19(4), 611-616, 2018

29 Lee, S. B, Oh, C. D., Lee, T. J., "A comparative study of government-media communication strategies on nuclear power issues", 17(3), 172-229, 2017

30 Cho, S. W., Lee, G. M., Lee, C., Kim, J. J., Jinxian Q., Im, H., "Prediction of Aviation Security Technologies through Cluster Analysis of Patent Big Data", 15(5), 43-63, 2017

31 Kang, J. H., "A study of atypical data analysis based on text mining Focused on writing pattern analysis", 42(8), 373-391, 2020

32 Suh, C. J., Yoon, J. E., "Research trend analysis on smart healthcare by using topic modeling and ego network analysis", 19(5), 981-993, 2018

33 Kim, S. W., Song, J. O., Ko, S. S., "Future Defense Power Development Using New Technology Plasma of the Fourth Industrial Revolution", 507), 100-107, 2021

34 Jang, J. H., Lee S. J., "Combined CNN algorithm for reducing false detection rate when identifying aircraft in satellite images", 76(3), 61-90, 2020

35 Cho, T. H., Lee, S. S., Choi, I. S., "Future Development of Aerospace Weapons System Based on Aeronautics and Space Force Technology Modeling", 24(5), 368-373, 2020

36 Kang, H. K., Park, Y. J., Park, J. H., "A Study on the Future Defense Weapon System and Key Technology Derivation Method Based on Big Data Analysis", 46(2), 225-242, 2018

37 Ahn, J. W., Nho, S. W., Kim, T. H., "An Empirical Study on the Prediction of Future Defense New Technologies in the Field of Artificial Intelligence", 21(9), 458-465, 2020

38 Blei, D. M., "Probabilistic Topic Models Surveying a suite of algorithms that offer a solution to managing large document archives", 55(4), 77-85, 2012

39 Jo, H. I., Lee, B. G., Kim, J. W., "A study on research trends of block-chain using LDA topic modeling: focusing on United States, China, and South Korea", 20(7), 1453-1460, 2019

40 Ha, G. H., Vu, T. M., Vu, G. T., Tran, B. X., Sahin, O., Nghiem, S., Ho, R. C., "Modeling research topics for artificial intelligence applications in medicine: latent Dirichlet allocation application study", 21(11), e15511, 2019

41 Jeong, M. S., Lee, J. Y., "Analysis of research activities and trends related to artificial intelligence (AI) technology based on the Latent Dirichlet Allocation (LDA) model", 23(3), 87-95, 2018