Sung-Hwan Kim

Dr. Sung-Hwan Kim


Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 January 2013 in International Journal of Machine Learning and Computing
Verified
Journal Article
Closed access
GPGyeong-Mi Park
Gyeong-Mi Park
SKSung-Hwan Kim
Sung-Hwan Kim
HHHye-Ryeon Hwang
Hye-Ryeon Hwang
HCHwan Gue CHO
Hwan Gue CHO
Recently we witnessed that the social network analysis focusing on social entities is applied in the social science and web-science, behavioral sciences, as well as in economics, marketing. In this paper we present one method to construct the social network from literary fictions by a simple lexical analysis, not using the complex natural language processing tools. And we will show that those social graphs, saying literary social graph, shows the power law distribution of some features, which is the typical characteristics of complex systems. We showed that the social network extracted from literary data reflects the similar network structure which was semantically designed by authors of fictions. And we newly proposed the concept of the kernel of literary social network by which we can classify the abstract level of protagonists appeared in fictions. Our study shows that the metric distance among characters written in linear text is very similar to the intrinsic and semantic relationship described by fiction writers, which implies the proposed social network from fictions could be another representation of literary fiction. So we can apply other scientific and quantitative approach by analyzing the concrete social graph model extracted from textual data

14 May 2020 in Applied Sciences
Verified
Journal Article
Open access
PDF file
SKSung-Hwan Kim
Sung-Hwan Kim
HCHwan Gue CHO
Hwan Gue CHO
Analyzing user behavior in online spaces is an important task. This paper is dedicated to analyzing the online community in terms of topics. We present a user–topic model based on the latent Dirichlet allocation (LDA), as an application of topic modeling in a domain other than textual data. This model substitutes the concept of word occurrence in the original LDA method with user participation. The proposed method deals with many problems regarding topic modeling and user analysis, which include: inclusion of dynamic topics, visualization of user interaction networks, and event detection. We collected datasets from four online communities with different characteristics, and conducted experiments to demonstrate the effectiveness of our method by revealing interesting findings covering numerous aspects.

Verified
Journal Article
Open access
PDF file
SKSung-Hwan Kim
Sung-Hwan Kim
KLKi-Joune Li
Ki-Joune Li
HCHwan Gue CHO
Hwan Gue CHO
Utilizing indoor spaces has become important with the progress of localization and positioning technologies. Covering and partitioning problems play an important role in managing, indexing, and analyzing spatial data. In this paper, we propose a multi-stage framework for indoor space partitioning, each stage of which can be flexibly adjusted according to target applications. One of the main features of our framework is the parameterized constraint, which characterizes the properties and restrictions of unit geometries used for the covering and partitioning tasks formulated as the binary linear programs. It enables us to apply the proposed method to various problems by simply changing the constraint parameter. We present basic constraints that are widely used in many covering and partitioning problems regarding the indoor space applications along with several techniques that simplify the computation process. We apply it to particular applications, device placement and route planning problems, in order to give examples of the use of our framework in the perspective on how to design a constraint and how to use the resulting partitions. We also demonstrate the effectiveness with experimental results compared to baseline methods.

1 January 2014 in IEICE Transactions on Information and Systems
Verified
Journal Article
Open access
PDF file
JSJong-Kyu Seo
Jong-Kyu Seo
SKSung-Hwan Kim
Sung-Hwan Kim
HCHwan Gue CHO
Hwan Gue CHO
A social network is a useful model for identifying hidden structures and meaningful knowledge among social atoms, which have complicated interactions. In recent years, most studies have focused on the real data of the social space such as emails, tweets, and human communities. In this paper, we construct a social network from literary fiction by mapping characters to vertices and their relationship strengths to edges. The main contribution of this paper is that our model can be exploited to reveal the deep structures of fiction novels by using graph theoretic concepts, without the involvement of any manual work. Experimental evaluation showed that our model successfully classified fictional characters in terms of their importance to the plot of a novel.

1 January 2013 in International Journal of Machine Learning and Computing
Verified
Journal Article
Closed access
SKSung-Hwan Kim
Sung-Hwan Kim
COChang-Seok Ock
Chang-Seok Ock
JSJong Kyu Seo
Jong Kyu Seo
HCHwan Gue CHO
Hwan Gue CHO
Sequence alignment is one of the methods widely used to determine string similarity. It computes similarity by aligning the component characters a string, and summing the similarity scores of pairs of matched characters. Using an appropriate character similarity measure is important when performing the alignment-based similarity calculation, since the string similarity is highly depending on the character similarity. In this paper, we focus on the character similarity learning process for string classification, particularly for when one set of strings that belong to the same class is given. Our method uses the matching frequency to calculate the character similarity. The performance of the method is also demonstrated by experimental evaluation