Will Cryptocurrency-Based Loyalty Programs Thrive as the Coin Price Rises?

Euro Bae(배유로),Hyeokkoo Eric Kwon(권혁규),Jaywon Lee(이재원),Daegon Cho(조대곤) 한국지능정보시스템학회 2022 한국지능정보시스템학회 학술대회논문집 Vol.2022 No.6

전자공시시스템(DART)을 활용한 국내 텍스트 분석(Textual Analysis) 환경에 관한 연구

김형준 ( Hyungjun Kim ),박종원 ( Jongwon Park ),이재원 ( Jaywon Lee ) 한국회계학회 2015 회계저널 Vol.24 No.4

본 연구는 금융감독원의 전자공시시스템(DART)을 통해 공시된 대량의 보고서들을 활용한 국내 텍스트 분석 연구 환경을 평가한다. 먼저 저자들은 대량의 공시서류를 내려받기 어려운 국내 전자공시 환경을 고려하여, 웹 크롤링 등을 활용한 우회적인 방법을 통해 1999년부터 2013년까지 공시된 총 111,497 건의 분기, 반기 및 사업보고서들을 내려받는다. 이후, Li(2008)의 연구방법을 응용하여 이들 보고서에서 "임원의 현황"표를 추출하는 비정형 데이터 마이닝 프로그램을 개발하고, 이를 바탕으로 전체 표본에 대한 텍스트분석 가능여부를 점검한다. 연구 결과, 전자공시시스템에 공시된 모든 분기, 반기, 및 사업보고서들 중 약 40%가 기계판독 불가능한 것으로 나타났다. 또한 기계판독이 가능한 보고 서들의 경우에도, 문서가 PDF 형식으로 제공되고 있어 효과적인 텍스트 분석을 위해서는 TXT 혹은 HTML 형식으로의 추가적인 변환작업이 필요한 것으로 나타났다. 최근 미국을 중심으로 각종 기업 공시자료들을 활용한 다양한 텍스트 분석이 진행되는 등 그 중요성이 점차 대두되고 있는 상황을 고려할 때, 본 연구의 결과는 국내 텍스트 분석 연구를 보다 활성화시키기 위해 DART 시스템의 개선이 선행되어야 한다는 점을 시사한다. In the United States, there is an increasing interest in the application of computer-based textual analysis and a large body of prior accounting literature has analyzed the textual contents of corporate filings such as annual reports. However, due to the lack of institutional supports and relevant experiences, neither practitioners nor researchers in Korea have used textual analysis. Given the circumstances, this study examines whether the Korea Financial Supervisory Service(FSS)``s on-line Data Analysis Retrieval and Transfer(DART) system offers adequate research environment for the textual analysis on corporate filings. By replicating a textual analysis procedure used in prior study, we point out some flaws and limitations of current DART system and suggest potential solutions to encourage the textual analysis in Korea. First, we start by collecting the full sample of corporate annual reports provided in DART system. Unlike the EDGAR(Electronic Data Gathering, Analysis, and Retrieval) system in U.S. which allows information users to easily download massive amount of corporate reports through the anonymous FTP server, the DART system only provides web-based downloading system and does not facilitate users to download massive amount of reports instantly. In this study, we overcome this issue by utilizing a web-crawler program and collect 111,497 corporate annual(quarterly) reports from 1999-2013 period. However the development of web-crawler program requires information users to have certain level of programming skills, implying that this approach is not appropriate to be widely used. Moreover, it took approximately 372 hours to complete our collection using web-crawler which is three times longer that the time needed to collect the similar amount of corporate reports using EDGAR anonymous FTP server. Overall, our findings from the collection step indicate that current DART system is not suitable for downloading massive amount of corporate reports which is necessary to run the textual analysis. Second, from the collected samples, we extract "top executives status" table from individual reports with a text analyzing program that we develop following Li(2008). This procedure intends to check the machine-readability of texts and tables included in corporate reports and to further assess the environment of textual analysis using Korean data. With a similar method used by Li(2008) when extracting MD&A sections and footnotes from his U.S. sample annual reports, we try to obtain "top executives status" table from our Korean sample reports. We deem specific sentences that meet certain regular-expression conditions we define to be starting sentences of the "top executives status" tables. This text analyzing program is written by Python programming language and tested to ensure 99% or higher accuracy of extracting the table. The results indicate that the Korean DART system offers inadequate research environment for textual analysis on corporate filings. Particularly, corporate reports provided in PDF format act as constraints on analyzing the textual contests. In detail, 43,407 reports out of 111,497, which approximately is 40% of the total samples are found to be not machine readable. For these reports, computer failed to extract "top executive status" table meaning that none of the texts from the report could be used for textual analysis. Moreover, given the technical difficulties in reading PDF format, corporate filings that are machine readable require another conversion process before the textual analysis adding to the difficulty of analyzing massive amounts of data. Alternatively, we develop web automation program and web-crawler program in an attempt to collect the sample reports in a machine readable format. However, using these programs are time consuming works which require approximately 1,800 hours to complete our collection process. Considering that textual analysis on corporate filings are on the rise, our overall results reveal that the refinements of DART system must precede the active research on textual analysis of corporate filings in Korea. We propose two potential solutions to resolve the problems related to current DART system and encourage the textual analysis in Korea. First, as in the EDGAR system, DART system must construct an anonymous FTP server where the information users can easily access and download massive amount of reports. This may allow practitioners and researchers to actively engage in textual analysis and thereby extend the application of disclosed corporate reports. Furthermore, corporate managers would pay more attention to the non-numeric information included in their reports increasing the quality of disclosure. Second, DART system must make technical complements to its report format by providing additional format of TXT or HTML. PDF format report which is the only format of report that DART system provides has its advantage in terms of clarity and security. However, PDF reports require additional conversion process which significantly increases the time spent on running textual analysis. In addition, TXT and HTML formats are perfectly machine readable and can resolve the problems related to machine readability issue. This study contributes to the literature by examining the textual analysis research environment using DART system. In typical, we point out the flaws and limitations of DART system by comparing its relevant attributes with EDGAR system. Furthermore, we propose two potential solutions to resolve the problems and in turn encourage the textual analysis research in Korea.

소아 입원환자에서의 A형 및 B형 인플루엔자 임상 비교

정승원 ( Seungwon Jung ),이준희 ( Joon Hee Lee ),강진한 ( Jin Han Kang ),이학성 ( Hak Sung Lee ),최재원 ( Jae Won Choi ),마상혁 ( Sang Hyuk Ma ),이재원 ( Jaywon Lee ) 대한소아감염학회 2017 Pediatric Infection and Vaccine Vol.24 No.1

목적: 소아 입원환자에서 A/H1N1, A/H3N2형 및 B형 인플루엔자 감염을 비교하고 항바이러스제의 효용성을 분석하고자 하였다. 방법: 2014년 1월부터 4월까지 창원파티마병원에 인플루엔자 감염으로 입원한 소아 환자들을 후향적 으로 분석하였다. 결과: 총 302명 중 인플루엔자 A/H1N1형 15명(5.0%), A/H3N2형 100명(33.1%), B형 187명(61.9%) 이었다. A는 24개월 미만, B는 24개월-6세 사이 감염자에서 높은 분포를 보였고(P =0.005). B형 인 플루엔자 감염군에서 발열 기간이 유의하게 길었다(P =0.001). 총 161명(53.3%)가 백신 접종자였으며, 감염 환자군 모두에서 oseltamivir를 복용한 환자들의 발열 기간이 유의하게 더 짧은 것으로 나타났다. 결론: A형과 B형 인플루엔자 환자는 연령 분포 및 임상 경과에 유의한 차이를 보였으며, oseltamivir 는 효과의 차이는 있었으나 두 군 모두에서 효용성이 있다. Purpose: The objective of this study was to compare the clinical characteristics of influenza A and B infections and analyze the effect of oseltamivir in hospitalized children. Methods: We investigated children under the age of 15, who were diagnosed with influenza A/H1N1, A/H3N2, or B from January to April 2014. The subjects were admitted to the Changwon Fatima Hospital and diagnosed using a rapid antigen test from nasopharyngeal swabs. The medical records of the patients were retrospectively reviewed. Results: A total of 302 pediatric patients with influenza were enrolled. Influenza B infection was the most common type (n=187, 61.9%), followed by A/H3N2 (n=100, 33.1%) and A/H1N1 (n=15, 5.0%). Compared to patients diagnosed with influenza A, patients diagnosed with influenza B were older (P =0.005), and the duration of fever was significantly longer (P =0.001). A total of 161 patients (53.3%) had been vaccinated against influenza during the season, before admission. Among the patients infected with A/H3N2 and B, the duration of fever was shorter in oseltamivir recipients compared to oseltamivir non-recipients (P =0.026 and P =0.004, respectively). Conclusions: There were significant differences between influenza A and B groups in terms of age, demographics, and clinical course. Although the effectiveness of oseltamivir on influenza differs according to the type of influenza, our data provides evidence that oseltamivir is beneficial for both A and B infections.