• Annual Conference on Human and Language Technology
• /
• 2018.10a
• /
• pp.185-188
• /
• 2018

띄어쓰기와 문장 경계 인식은 그 성능에 따라 자연어 분석 단계에서 오류를 크게 전파하기 때문에 굉장히 중요한 문제로 인식되고 있지만 각각 서로 다른 자질을 사용하는 문제 때문에 각각 다른 모델을 사용해 순차적으로 해결하였다. 그러나 띄어쓰기와 문장 경계 인식은 완전히 다른 문제라고는 볼 수 없으며 두 모델의 순차적 수행은 앞선 모델의 오류가 다음 모델에 전파될 뿐만 아니라 시간 복잡도가 높아진다는 문제점이 있다. 본 논문에서는 띄어쓰기와 문장 경계 인식을 하나의 문제로 보고 한 번에 처리하는 다중 클래스 분류 시스템을 통해 시간 복잡도 문제를 해결하고 다중 손실 선형 결합을 사용하여 띄어쓰기와 문장 경계 인식이 서로 다른 자질을 사용하는 문제를 해결했다. 최종 모델은 띄어쓰기와 문장 경계 인식 기본 모델보다 각각 3.98%p, 0.34%p 증가한 성능을 보였다. 시간 복잡도 면에서도 단일 모델의 순차적 수행 시간보다 38.7% 감소한 수행 시간을 보였다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2019.05a
• /
• pp.83-86
• /
• 2019

This study proposes a methodology for predicting co-authorship of contributors to a highly anticipated paper through link analysis and learning, taking into account the result of collaborative research. Previous studies predict the co-authorship with high accuracy, but this shows limitations in that the quality of the predicted relationship is not considered. Therefore, to solve the above problem, we propose three steps to predict the co-authorship that will help with the expected performance: (1) Construct a heterogeneous graph to measure results of collaborative research. (2) Analyze and learn links based on results of collaborative research. (3) Predict links that are anticipated to have high expectation. It is expected to be useful for increasing confidence in the predicted co-authorship.

• Journal of the Korean Society for Library and Information Science
• /
• v.56 no.4
• /
• pp.473-490
• /
• 2022

This study was conducted to examine Sci-Hub, which provides the original text of academic papers to be provided for a fee by bypassing copyright, and to establish a basic basis for understanding the usage behavior of researchers with access restriction in Korea by analyzing the domestic Sci-Hub usage based on the dataset released by Sci-Hub in 2017. Therefore, after grasping the current status of the world related to Sci-Hub, the categories were set and analyzed by region where the dataset was downloaded, subject matter of academic papers, publisher, OA status, and published year. As a result of the study, the most downloaded areas were Seoul and the metropolitan area, and papers and journals in the field of natural science were downloaded the most, and about 20% of papers were in Open Access state. The papers published between 2010 and 2017 were the most downloaded, and IEEE's papers were the most downloaded, showing that recently published academic papers in the natural sciences were the most downloaded by the time they were downloaded.

• Journal of the Korean Society for information Management
• /
• v.16 no.2
• /
• pp.183-199
• /
• 1999

The three key parts of citation index are the Citation Index, the Source Index and the Permuterm Subject Index. To identify the core elements for citation index database system, it analyzed the items that have been cited in references and source items of Journal of the Korean Society for Information Management from vol. 1 no. l(1984) to vol. 15 no. 3(1998). Ten core elements were identified and the description format was specified respectively. The core elements are author, organization that the author is affiliated, title of article, title or journal, volume/numberlpage numbers, year of publication, keyword, language, subject category, statistics of references in article. Also, characteristics of errors in reference citations were analyzed and categorized in the viewpoint of the citation index development of academic journals in Korea.

• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.109-111
• /
• 2004

발성율은 일정한 시간동안 발성되는 음성신호 내에 몇 개의 음절이 포함되어 있는 지를 나타낸다. 발성율은 화자마다 다르고 각 음소들의 특징에 따라 변화할 수 있다. 발성율의 사전 측정이 이루어 진다면 음성부호화 측면에서도 중용한 정보로 사용될 수 있다. 기존의 음성부호화기는 발성율에 관계없이 고정적인 분석 구간을 정하여 전송률을 결정하고 있다. 따라서, 발성율을 미리 측정한다면, 발성율이 느린 부분과 빠른 부분에 각기 다른 부호화 방법을 적용하여 음질을 향상할 수도 있고 전송률을 가변적으로 적용할 수 도 있게 된다. 정확한 발성율을 측정하기 위해서는 음절의 변화를 추정하여야 한다. 음절의 변화를 추정하기 위한 방법으로 음성신호의 에너지 포락선 측정법과 LSP를 이용한 측정법이 각각 제안된 바 있으나, 본 논문에서는 위 두 가지 방법을 혼합한 방법을 사용하였다. 에너지 변동은 음성신호의 시간영역 처리방법으로 LSP 파라미터는 음성신호의 선형예측 분석에 의해 구해질 수 있다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2017.11a
• /
• pp.120-122
• /
• 2017

정보통신기술의 발달에 따라 컴퓨터, 스마트폰과 같은 스마트 기기가 대중화됨에 따라 개인 블로그나 SNS 등을 기반으로 하여 개인이 다양한 콘텐츠를 생산하고 공유하는 커뮤니케이션 플랫폼인 1인 미디어가 활성화 되었다. 효율적으로 1인 방송시스템을 구축하기 위해서는 기존 방송 시스템의 체계는 유지하면서 개인의 협소한 공간을 효율적으로 활용할 수 있는 방안을 요구하게 된다. 따라서 본 논문에서는 각각의 방송장비들을 노트북, 어플과 같은 스마트 기기와 연동시켜 효율적이고 편리한 1인 미디어 방송 시스템을 구축하기 위한 방법을 제안하였다.

• Proceedings of the IEEK Conference
• /
• 2003.07e
• /
• pp.2447-2450
• /
• 2003

화자인식은 음성의 특성을 이용해서 화자의 신원을 확인하는 기술이다. 이러한 기술은 등록된 화자집단 중 화자를 식별하는 화자식별(speaker Identification)과 지금 발성한 화자만을 비교하여 확인하는 화자확인(speaker verification)이 있다. 이러한 화자인식은 음성에 내재되어 있는 화자정보를 추출하여 개인을 확인하는 기술로 전화망을 통한 서비스가 확산되어 가고 있는 현대사회에 가장 효과적인 기술 중 하나이다. 또한 PDA를 이용한 증건거래 시스템 등 현대사회에서는 실시간으로 화자인식이 이루어져야 한다. 본 논문에서는 이와 같이 실시간 화자인식을 위한 처리시간 단축에 관하여 연구하였다. 처리시간 단축을 위하여 우선 피치주기 단위로 음성 파형을 분해한 다음 분해된 피치 단위에 윈도우 함수를 곱해서 단구간 신호의 열로 만들고 분해된 단위를 조절하는 PSOLA 합성방식을 이용하여 인식 시스템의 전처리단을 재구성하였다. 이와 같은 방식으로 제안한 인식시스템의 처리시간, 인식률을 기존의 화자인식 시스템과 비교하였다.

• Journal of the Korean Society for information Management
• /
• v.38 no.1
• /
• pp.53-69
• /
• 2021

This study aims to understand the research landscape of South Korea using the data of 2020 Google Scholar Metrics. To achieve the goal, we constructed and analyzed four types of networks including the university collaboration network, the keyword co-occurrence network, the journal citation network, and the discipline citation network. Through the analysis of the university collaboration network, we found major universities such as Seoul National University, Keimyung University, and Sungkyunkwan University that have led collaborative research. Job related keywords such as job change intention and job satisfaction have been frequently studied with other keywords. Through the analysis of the journal citation network, we found multiple journals such as The Journal of the Korea Contents Association, Korean Journal of Sociology, and Korean Journal of Culture and Social Issues that have been widely cited by the other journals and influenced them. Finally, Education, Business administration, and Social welfare were identified as the top influential disciplines that have influenced other disciplines through the knowledge diffusion. The study is the first of its kind to use the data of Google Scholar Metrics and conduct a stepwise network analysis (e.g., keyword, journal, and discipline) to broadly understand the research landscape of South Korea. Our results can be used by government agencies and universities to develop effective strategies of promoting university collaboration and interdisciplinary research.

• Journal of Korean Library and Information Science Society
• /
• v.53 no.4
• /
• pp.211-232
• /
• 2022

Research performance evaluation has been based heavily on quantitative criteria. Various limitations however were revealed and the demand for excellence in research quality was growing and the importance of qualitative evaluation factors was increasing as well. The study investigates and introduces cases for improving research evaluation that have appeared around the world since 2010. To extract qualitative factors that can be introduced in research achievement evaluation in Korea this study investigated and organized the performance evaluation cases of foreign universities and proposed the introduction of the evaluation method in Korea. Qualitative evaluation factors were extracted by analyzing achievement evaluation documents evaluation procedures and evaluation standards of top 10 universities around the world. As a results this study proposes the use of various performance evaluation documents mulch-faceted evaluation and multi-level performance evaluation process and various qualitative performance evaluation standards. In traditional bibliometrics, while the degree of citation was measured when generating research evaluation indicators, future indicators may reflect recommendation scores as indicators, or include industrial connectivity and utilization in education / industry. It is hoped that the performance evaluation method will be balanced through this paper and ultimately the professor's research achievement evaluation will be a process that can help individuals grow into excellent researchers and develop the research environment.

• Journal of the Korean Society for information Management
• /
• v.41 no.2
• /
• pp.1-18
• /
• 2024

This study aims to compare ego-centered journal citation identities among four domestic journals in library and information science. Ego-centered citation identity refers to the set of authors that an author frequently cites. The target journals for this study are Journal of the Korean Society for Library and Information Science (KSLIS), Journal of the Korean Biblia Society for Library and Information Science (KBIBLIA), Journal of Korean Library and Information Science Society (KLISS), and Journal of the Korean Society for Information Management (KOSIM). As a result of citation/citee ratio (CCR), self-citing rates (SCR), and journal co-cited analysis, the journal citation identities of four journals contained the other three journals besides the ego journal and JASIST. Furthermore, KOSIM had the most diverse range of journal citation identity and the four journals mattered the intra-journal information. KLISS showed the most unique cited journal network structure among the four journals.