A Latent Topic Modeling approach for Subject Summarization of Research on the Military Art and Science in South Korea

Within the military art and science research articles, many of studies have focused on the empirical examination or theoretically review interests related to military phenomena (e.g., military power, security relations). Furthermore, because the technological and social change in recent years have spurred convergence and generalization research in Korean military art and science, multidisciplinary research related to science and technology and various disciplines is emphasized as well as social science that studies military issues. In this sense, to maintain sustainable development of current research, it is necessary to explore specific topics on academic studies. To accomplish this purpose, we conducted a sequence of three analytic stages First, we selected a web database in KCI OAI-PMH (Korea Citation Index Open Archives Initiative Protocol for Metadata Harvesting) for literature search. And then we performed subsequently snowballing sampling via author affiliation and related-keywords (i.e., military, defense, weapon etc.) of reference list based on initial DB search to find comprehensive articles from relevant studies. It comprised data sets including English Abstract from a total of 4,193 studies (314 journals) during 2002-2019. Second, using these data sets, we extracted token, lemma, and morphological features of potentially useful NOUN by employing Universal Dependencies (UD) pipeline for joint sentence segmentation, word segmentation. Third, based on the topic modeling using topicmodels, OpTop, and topicdoc package with Latent Dirichlet Allocation (LDA) algorithms on this corpus, we presented new subject classification including ten topics (i.e., Defense Reform, security alliance, defense industry, defense R&D, combat simulation, dynamic analysis of weapon systems, reliability evaluation of weapon system, target detection, characteristics analysis of materials or performance). As a result, the current study explored latent topic cluster (subject classification) was divided into 1) weapon system acquisition and management (military force building and intangible force maintenance) and 2) defense R&D (military force operation, military force development, tangible force maintenance) based on the topic network analysis. These findings are preliminary, but it enhances our understanding of the existing sub-subject areas by extending subject classification for military art and science in the Knowledge Classification Scheme of National Research Foundations of Korea.