A5101538818- Corporate Finance and Governance
- Text and Document Classification Technologies
- Firm Innovation and Growth
- Advanced Text Analysis Techniques
- Web Data Mining and Analysis
- Technology and Data Analysis
- Parallel Computing and Optimization Techniques
- Recommender Systems and Techniques
- Topic Modeling
- Data Management and Algorithms
- Impact of AI and Big Data on Business and Society
- Interconnection Networks and Systems
- Data Mining Algorithms and Applications
- Semantic Web and Ontologies
- Anomaly Detection Techniques and Applications
- Private Equity and Venture Capital
- Supercapacitor Materials and Fabrication
- Innovation Policy and R&D
- Natural Language Processing Techniques
- Financial Reporting and Valuation Research
- Imbalanced Data Classification Techniques
- Machine Learning and Algorithms
- Machine Learning and Data Classification
- Advanced Data Storage Technologies
- Time Series Analysis and Forecasting
University of Seoul
2014-2024
Hanyang University
2010-2023
Anyang University
2023
Hoseo University
2011-2018
Korea Advanced Institute of Science and Technology
2009-2015
Kootenay Association for Science & Technology
2014
Yonsei University
2013
Seoul National University
2002-2010
Government of the Republic of Korea
2009
Cheongju University
2008
This paper presents a keyword extraction technique that can be used for tracking topics over time. In our work, keywords are set of significant words in an article gives high-level description its contents to readers. Identifying from large amount on-line news data is very useful it produce short summary articles. As text documents rapidly increase size with the growth WWW, has become basis several mining applications such as search engine, categorization, summarization, and topic detection....
On-chip networks are critical to the scaling of future multicore processors. Recent processors have adopted ring topologies because its simplicity and high bandwidth. In this paper, we first describe a bufferless router microarchitecture for an on-chip network topology. We propose extend with extra buffer entry create lightweight microarchitecture. The proposed approaches ideal latency by reducing complexity through minimizing amount buffers simplifying switch allocation. how does not need...
Network-on-chip (NoC) bandwidth has a significant impact on overall performance in throughput-oriented processors such as GPG-PUs. Although it been commonly assumed that high NoC can be provided through abundant on-chip wires, we show increasing router frequency results more cost-effective NoC. However, arbitration critical path limit the frequency. Thus, propose direct all-to-all network overlaid mesh (DA2mesh) architecture exploits traffic characteristics of GPGPU and removes from...
This paper investigates the determinants of capital structure large corporations headquartered in United States and Korea. We consider five explanatory variables: profit, company size, non‐debt tax shields, growth, business‐risk, along with several industry indicator variables as independent examine, for each country, relationship to market value based leverage ratio. With our rigid criteria inclusion study, we study top thirteen companies (by size) seven industries. The majority findings...
Bufferless on-chip networks are an alternative type of network organization that can improve the cost-efficiency by removing router input buffers. However, bufferless performance degrades at high load because increased contention and large number deflected packets. The energy benefit is also reduced deflection. In this work, we propose a novel flow control for in high-throughput manycore accelerator architectures to reduce impact deflection routing. By using clumsy (CFC), instead per-hop...
Recently, novelty detection with reconstruction along projection pathway (RaPP) has made progress toward leveraging hidden activation values. RaPP compares the input and its autoencoder in spaces to detect samples. Nevertheless, traditional autoencoders have not yet begun fully exploit this method. In paper, we propose a new model, Extended Autoencoder Model, that adds an adversarial component take full advantage of RaPP. The matches latent variables reconstructed original samples high...
One of the most difficult problems in developing question-answering (QA) system is that it so hard to generate natural language questions and find an answer a query question. In order avoid number difficulties QA systems, we propose new style architecture actively uses sentences within document as source question/answer. Basically, our proposed gives user set candidate question for information needs, are automatically generated from significant expected contain meaningful facts or events....
Collaborative filtering, which is a popular approach for developing recommendation systems, exploits the exact match of items that users have accessed. If access different items, they are considered as unlike-minded even though may actually be semantically like-minded. To solve this problem, we propose semantic collaborative filtering model represents semantics users’ preferences and with their corresponding concepts. In work, extend Bayesian belief network (BBN)-based because it provides...
A cost-efficient network-on-chip is needed in a scalable many-core systems. Recent multicore processors have leveraged ring topology and hierarchical can increase scalability but presents different challenges, including higher hop count global bottleneck. In this work, we describe that refer to as transportation-network-inspired (tNoC) leverages principles from transportation network particular, propose novel hybrid flow control for scale the efficiently. The channels are allocated on flit...
This paper presents a novel way of building the user profile concept network for personalized search. The is defined as network, in which each approximately represented with formal analysis (FCA) theory. We assume that concept, called `session interest concept', subsume user's query intention during session and it can reflect preference. Whenever issues his/her query, generated. Then, new concepts are merged into current (i.e., profile) recent preferences accumulated. According to FCA, pair...
Document clustering is inherently an unsupervised learning process that organizes document (or text) data into distinct groups without depending on pre-specified knowledge. However, real-world applications, such as building a topical hierarchy for large collection, need to perform under various kinds of constraints. This paper presents new type supervised organize information in way reflects knowledge provided by user. As means which external human can be incorporated the process, quadratic...
Global competition has increased the importance of patents as a means to protect and strengthen technology competitiveness. The purposes our study were identify what industries in South Korea are strong or weak terms patent applications some strategies enable become strong. For this, we gathered statistics on seven variables follows: number businesses, employees, research development investment, full-time equivalent researchers, institutions, domestic market size, applications. Especially,...
The study tests the hypothesis that there are country and industry effects on capital structure of large corporations headquartered in United States Korea. So far, very few studies have focused systematically these issues between developed less countries. For book value–based leverage ratios, Korean firms had higher debt ratios than their U.S. counterparts, while were no or significant construction chemical industries at market ratios. Moreover, statistically differences using value measures...
본 논문은 기계학습 기법에 기반한 웹문서 자동분류 시스템의 성능을 높이기 위한 새로운 형태의 특징가공 기법을 제안한다. 제안 기법은 하이퍼텍스트 웹문서에 대한 자동분류를 효과적으로 수행하기 위해 하이퍼링크 관계를 활용하여 특징 집합을 확장시킨다. 웹문서는 통해 서로 연결된 구조를 가지며, 그 관계는 많은 경우 연관도가 높은 문서들 간에 존재한다. 이러한 링크 정보가 분류모델의 주요 인자가 되는 집합의 질을 높이는 중요한 역할을 수행할 수 있다. 기법의 기본 아이디어는 워드넷 온톨로지를 기반으로 분류 대상 문서와 인접 문서들에 포함된 단어(특징)들 간의 의미적 유사도를 평가함으로써 다수의 특징들로 구성된 추상화된 개념적 특징을 생성하는 것이다. 여기서 유사도 함수는 안에서 특징들 상/하위어 관계 정보를 정량적으로 계산하게 된다. 구축시 개념 특징은 일반 특징과 동일하게 간주하여 보다 정확한 모델을 구축하는데 기여한다. Web-KB 문서집합을 이용한 실험을 기법이 기존 기법...
본 연구는 국내 자본시장에서 지금까지 상대적으로 낮은 관심을 받아왔던 KOSDAQ 상장기업들의 재무적 특성 분석(자본구조 포함에 관한 실증적인 연구이다. 연구에서는 기업들(우량기업부와 벤처기업부 소속)에 대한 포괄적인 가설검정을 실행하였다. 첫 번째 가설은 동 상장회사들의 자본구조(부채) 결정에 영향을 미칠 수 있는 설명변수의 통계학적인 분석이다. 다중 회귀식을 이용한 결과, 장부가 기준의 종속변수에 영향력과의 관계에서, 자산크기, 성장성, 시가-장부가 비율, 변동성, 시가총액, 그리고 내의 소속부에 더미변수 들이 유의성을 보였고, 시장가 설명변수들의 영향력은 체계적 위험의 대용치인 베타계수, 소속부 더미변수에서 나타났다. 두 가설검정은 로지스틱 모델을 분류 기준상, 해당 기업들의 재무적인 특성분석 검정이었으며, 마지막 분석은 산업별 평균 부채비율로서의 '회귀성'관련 검정이었다. 회귀식 결과와 관련하여 재무특성 중, 수익성, 시가대비 그리고, 외국인 지분률이 각각 높을수록...
This paper presents a novel text space model that represents textual documents for document clustering, which contains the 'concept' independently of 'document' and 'term' spaces. The described here as matrices (i.e., 2nd-order tensors), corpus is represented 3rd-order tensor. For this, it necessary to produce concept vector each term occurs in given document, related word sense disambiguation. As an external knowledge source weighting, we employ Wikipedia encyclopedia.