• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.281-286
• /
• 2021

뉴스 헤드라인에 제3자의 발언을 직접 인용해 전언하는 이른바 '따옴표 저널리즘'이 언론 보도의 객관주의 원칙을 해치는지는 언론학 및 뉴스 구독자에게 중요한 문제이다. 이 연구는 온라인 포털사이트를 통해 실시간 유통되는 한국어 기사의 정확성을 판별하기 위한 기계학습(Machine Learning) 모델을 제안한다. 이 연구에서 제안하는 모델은 Edit Distance와 FastText 기법을 활용해 기사 제목과 본문 내 인용구의 유사성을 측정하고, XGBoost 모델을 활용해 최종 분류한다. 아울러 이 모델을 통해 229만 건의 뉴스 헤드라인에 대해 직접 인용구가 포함된 기사가 취재원의 발언을 주관적인 윤색없이 독자들에게 전하고 있는지를 판별했다. 이뿐만 아니라 딥러닝 기반의 KoELECTRA 모델을 활용해 기사의 제목 내 인용구에 대한 감성 분석을 진행했다. 분석 결과, 윤색이 가미되지 않은 직접 인용형 기사의 비율이 지난 20년 동안 10% 이상 증가했으며, 기사 제목의 인용구에 나타나는 감정은 부정 감성이 긍정 감성의 2.8배 정도로 우세했다. 이러한 시도는 앞으로 계산사회과학 방법론과 빅데이터에 기반한 언론 보도의 평가 및 개선에 도움을 주리라 기대한다.

• Journal of the Korean Society for Library and Information Science
• /
• v.54 no.2
• /
• pp.197-223
• /
• 2020

In this study, citation-weighted author profiling (CWAP) was developed as a new method that combines the advantages of both author profiling (AP) method and author co-citation analysis (ACA) method. In AP method, words reflect the author's main topics of study. On the other hand, what words reflect in CWAP is topics that the author mainly influences. This enables detailed topic identification, which is the advantage of AP method, and at the same time determines the subjects in which the author has influence, as with ACA method. The proposed CWAP method was applied experimentally to analyze the intellectual structure of school library research in Korea. The results of the trial application revealed in detail what topics each author has a high influence on, and the change of influence over time was also clearly revealed. The CWAP method proposed in this study is expected to be used as a technique to grasp detailed topics from the viewpoint of research influence on which topics the author has been cited for, not as a research productivity perspective of how many papers the author has published.

• Journal of the Korean Society for information Management
• /
• v.32 no.4
• /
• pp.77-106
• /
• 2015

By document co-citation analysis with Korean Citation Index (KCI) data, this study accurately identified the research fronts and hot topics in Korean library and information science (LIS) from 2004 to 2013. 159 core papers in LIS domain and their citations are scraped manually from Korean Citation Index web site. In the cluster analysis and network analysis, 159 core papers were grouped into 27 clusters with multiple papers and 8 singlton clusters. Among the 27 clusters which have multple papers, 'LIS education' cluster was the largest with 16 core papers, and 'citation analysis & intellectual structure analysis' cluster had the strongest citation impact according to the ehs-index. Closer observation of the citations to the core papers in each research front showed that 67.5% of the citations were made by LIS research papers and 32.5% of the citations were made by non-LIS research papers. Considering the share of citations and the citation impact growth index, 'local documentation', 'citation analysis & intellectual structure analysis', and 'research trends analysis' were identified as the most emerging research front in Korean library and information science. The analytical methods used in this study have great potential in discovering the characteristics of research fronts in Korean interdisciplinary research domains.

• Journal of the Korean Society for Library and Information Science
• /
• v.36 no.4
• /
• pp.21-47
• /
• 2002

The present study, on the bass of a premise that reference lists are one of the social products, reflecting various social environments of their own society, was made as part of an attempt to explore the co-occurrence phenomenon of both Korean and Non-Korean Literatures occurred within the Korean references. 321 authors (articles) of 43 Korean journals on Social Sciences were surveyed on research channels and citation motivations and their 11358 references were analyzed. The findings are as follows : 1) The extent of the co-occurrence was that Non-Korean literatures were more 1.9 times (65.3%) cited than Korean ones and English (61.5%)-American (50.4%) predominancy was heavily found. 2) Research channel, worked as an indicator of the identity of researcher as well as the source of research ideas was most Non-Korean channel orientedness (55.8%). 3) Citation motivations were significantly depended on whether Korean or Non-Korean literatures and non-Korean literatures were cheifly cited to be conceptual motivations than other motivations. 4) Research channel among variables was worked as a main effect predicting major citation motivations on Non-Korean literatures. Finally, this study is very suggestive : 1) It might be a new approach and interpretation by adopting citation motivations to explore a process of knowledge producting of researchers 2) Partly, it proved empirically the relationship of knowledge producted by Korean researchers to Non-Korean knowledge through the analysis of citation motivations.

• Proceedings of the Korean Society for Language and Information Conference
• /
• 2001.06a
• /
• pp.269-282
• /
• 2001

이 논문은 한국어 간접인용조사 ‘고’가 종결표지로 사용될 때 발생하는 화용적 기능을 설명한다. [묘출발화/사고(RST)의 관점에서 1인칭 화자와 3인칭 인물 사이의 관계 개념에 근거하여, 인용 정보에 대한 화자의 반응이 언어적으로 수용되는 과정에 대한 표상의 문제를 취급한다. 인용 명제에 대해 화자가 취하는 주관적 관점을 평가하기 위해 필요한 요인들을 변별하여 주고, 1인칭 화자의 내적 판단 영역이 인용의 정보 상태와 대응하기 위해 필요한 제약들을 파악한다.

• Proceedings of the Korean Society for Information Management Conference
• /
• 2005.08a
• /
• pp.305-310
• /
• 2005

이 연구에서는 주제검색을 통해 검색된 문헌들의 인용정보를 이용한 질의확장 기법을 제안하였으며 이 제안된 기법의 성능을 일반적 질의확장 기법인 지역적 질의확장 및 전역적 질의확장과 비교 평가하였다. 연구 결과 인용기반 질의확장 기법이 전역적 및 지역적 질의확장 기법에 비해 우수한 성능을 보임을 확인하였으며, 특히 피인용 표제어를 이용한 질의확장 검색의 효용성을 실험을 통해 밝혀냈다.

• Proceedings of the Korean Society for Information Management Conference
• /
• 2014.08a
• /
• pp.3-6
• /
• 2014

국내 인용 데이터베이스에서도 해외와 마찬가지로 저널 영향력 지수가 핵심 지표로 제공되고 있으며 저널 평가나 연구성과 평가의 기준 지표로 사용되는 경향이 확산되고 있다. 그러나 국내 저널에서는 논문이 발표된 당해년도에 인용되는 비율이 해외에 비해서 높기 때문에 이를 반영하지 않는 영향력 지수는 국내 저널에 대해서는 더 불완전한 지수라고 할 수 있다. 이 연구에서는 여러 해에 걸쳐 발행된 논문이 특정 연도에 인용된 횟수로 평가하는 저널 영향력 지수와 달리, 특정 연도에 발행된 논문이 여러 해에 걸쳐 인용된 횟수로 평가하는 통시적 영향력 지수의 개념과 산출방법을 알아보고 정보관리학회지를 대상으로 시험 측정한 결과를 살펴본다.

• Proceedings of the Korean Society for Information Management Conference
• /
• 2015.08a
• /
• pp.77-82
• /
• 2015

본 연구는 의학 분야를 중심으로 A 대학교 의과대학 도서관의 Science Direct 전자학술지 이용통계와 Scopus 인용지수를 비교분석하였다. 이에 594종의 의학 분야 전자학술지를 추출하여, 그 이용통계와 Scopus 인용지수에 대해 빈도분석, 상관관계분석을 실시하였다. 그 결과 의학 분야 전자학술지의 이용이 소수의 학술지에 집중되어 있고, 그 이용통계와 인용지수 사이에 양의 상관관계가 성립함을 확인할 수 있었다.

• Journal of the Korean Society for Library and Information Science
• /
• v.47 no.2
• /
• pp.247-264
• /
• 2013

Wikipedia represents a revolutionary experiment in the sense that it allows anonymous Internet users to contribute, change, and edit Encyclopedia contents used by tens of millions of people. While a very popular information source on the Internet, because of its questionable information credibility and accuracy, citing Wikipedia articles is being regarded as a risky behavior for scholars. The present study identified 282 scholarly articles from Thomson Reuters' Web of Science citation database that cite Wikipedia at least once. Out of the millions of articles indexed in Web of Science, the proportion of articles citing Wikipedia is extremely small. On the other hand, the numbers are showing a marked increase since 2011. Wikipedia citing articles are distributed in subject areas, such as library and information science, business, psychology, education, and communication more often than in other areas. The distribution of a total of 577 citations from 267 articles for which we were able to obtain full texts shows that Wikipedia is being cited mainly in studies of Wikipedia (139 citations, 24.1%) or as a ready reference source (331 citations, 57.4%). At the same time, about 15% of total citations turned out to be cases of potentially risky behaviors in which Wikipedia is being cited as a crucial basis or data source for study.

• Journal of the Korean Society for information Management
• /
• v.30 no.3
• /
• pp.273-296
• /
• 2013

The purpose of this study is to examine the data coverage and citation threshold for analyzing SPLC(Search Path Link Count) as a main path of a historiograph of a certain topic in order to provide 'core' papers of global research trends to a researcher affiliated with a local R&D institution. 5 datasets were constructed by retrieving and collecting 2,318 articles on RGB LED on Web of Science published from 1990-2013 and 20,109 articles which cited these original 2,318. The SPLC analysis was performed on each dataset by increasing the threshold of citation counts, and the changes and resilience of the 28 extraced networks were compared. The results of user feedback on 198 unique core papers from 28 SPLC networks received from LED researchers affiliated with a Korean government-sponsored research institution were also analyzed. As a result, it is found that the nodes in each SPLC network in each dataset were differentiated by the citation counts, while the changes in the structure of SPLC networks were slight after the networks' citation counts were set at 40. Additionally, the user feedback showed that personalized research interest generally matched to the global research trends identified by the SPLC analysis.