뇌동맥류 환자의 방사선학적 중재적 시술과 신경외과적 수술 방법에 따른 재원기간 비교

이정범 충남대학교 보건·바이오산업기술대학원 2008 국내석사

뇌동맥류의 치료방법인 방사선학적 중재적 시술(coil embolization)과 신경외과적 수술(clipping operation)을 받은 환자의 재원기간을 비교하였다. 뇌동맥류(cerebral aneurysm) 환자의 치료방법에 따른 재원기간을 비교한 결과 Hunt-Hess grade, Fisher grade가 낮은 집단일수록, clipping operation 보다는 coil embolization을 이용하여 치료한 집단일수록, 파열되지 않은 동맥류일수록 재원기간이 짧았다. The purpose of this study is to compare and analyze length of stay (LOS) according to the endovascular treatment using detachable coils (coilembolization) and neuro surgical operation (clipping operation) in patients of cerebral aneurysm. The data was obtained from the patients who underwent CT (computed tomography) or TFCA (transfemoral cerebral angiography) for suspicion of cerebral aneurysm. Subjects of this study included patients who underwent coil embolization (n=69) and clipping operation(n=15) between January 1, 2006 to December 31, 2006 for cerebral aneurysms. Length of stay was compared between 69 cases of coil embolization and 15 cases of clipping operation. The results were as follows; 1. In cerebral aneurysm patients male to female ratio was about 4:6. 2. There were about 1.7 times more frequent in ruptured aneurysms than in unruptured ones. 3. There were 13 times more frequent in anterior circulation aneurysms than in posterior circulation ones. 4. There were 4.5 times more frequentin coil embolizations than in clipping operations, about 7 times in male patients. 5. Coil embolization group showed significantly shorter length of stay than clipping operation group (p<0.05). 6. There were significant differences among patients of different Hunt-Hess grades, and length of stay was shorter in patients of lower grades (p<0.05). 7.There were significant differences among patients of different Fisher grades, and length of stay was shorter in patients of lower grades (p<0.05). 8.Hunt-Hess grades,Fisher grades, rupture, and treatment modalities (coiling, clipping) affected length of stay. In conclusion, length of stay was shorter in patients of lower Hunt-Hess and Fisher grades.In addition, those were also shorter in coil embolization group than in clipping operation group. Therefore, to shorten the length of stay, coil embolization should be initially considered for treatment of patients with cerebral aneurysms.

국소진행성 직장암에서 선행항암화학방사선요법 후 종양퇴행에 영향을 미치는 요인

이정범 인하대학교 대학원 2017 국내석사

Purpose: Neoadjuvant chemoradiotherapy is the standard of care for locally advanced rectal cancer. The aim of this retrospective study was to identify factors affecting tumor regression after neoadjuvant chemoradiotherapy of locally advanced rectal cancer. Methods: From January 2011 to December 2016, retrospective analysis was performed on 93 patients with locally advanced rectal cancer who were received neoadjuvant chemoradiotherapy followed by radical surgery. The total dose of radiation was 54Gy, 2Gy was irradiated once a day for 5 days per week. Of 93 patients, 72 patients received concurrent infusional 5-fluorouracil and leucovorin, while 21 patients received oral capecitabine during radiotherapy. Tumor regression grade(TRG) system was applied by simplifying Dworak TRG. Results: Of the 93 patients, 54 (58.1%) cases were excellent tumor regression response(TRR)(Dworak TRG 3-4). Univariate analysis demonstrated that age, over 65 years old, was significant predictive factor for excellent TRR (P=0.003). The results from the multivariate analysis also indicated that age (Odd ratio, 0.21; P=0.002) was significant predictive factor. Conclusion: Over 65 years old age patients was predictor for excellent TRR. That would be helpful for treatment guidelines for old age patients. However, there are many restriction in retrospective study, so we recommend the randomized prospective study. Key Words: Locally advanced rectal cancer, Neoadjuvant chemoradiotherapy, Tumor regression

自己學習을 通한 地理科의 學習相談 事例 硏究 : 中學校 社會Ⅰ 敎科書를 中心으로

이정범 高麗大學校 1990 국내석사

토마스 아퀴나스의 정의론에 대한 연구

이정범 서강대학교 대학원 2014 국내석사

Thomas Aquinas drew Aristotle's ethics, which was new to Europeans at that time, in constituting his own. In this process, he incorporated some concepts and ideas prevailed at that time which were related to the problem of justice with Aristotle's, while dismissed others. Most of the previous studies on justice of Thomas Aquinas doesn't take the intellectual context of that time into consider. Even if a study of MacIntyre(1989) does, it merely sees Thomas as the one who only incorporated prevailing concepts of that time with Aristotle's. In this study, I focused on not only the historical and intellectual context of that time and what were incorporated into his understanding on justice, but also uniquenesses of Thomas Aquinas' own and what were dismissed by him. To achieve this goal, in Chapter 2, I examined Augustine's and Aristotle's views on human being as well as their views on justice, which compose philosophical background of their justice, to comprehend some contexts of concepts related to understanding on justice of that time. After finishing this work, in chapter 3, I inquired into the concept of Thomas Aquinas' own justice. In this process, I also examined what concepts related to justice of that time had Thomas Aquinas incorporated into his own version of justice and dismissed from his predecessor's. Then I tried to show some uniquenesses of Thomas Aquinas' justice. As a result, firstly, I found Thomas Aquinas rejected Augustine's understanding on justice, defining it as belonged to 'fas'(divine law), not 'jus'(human law; just deed. Thomas Aquinas set justice as belonged to human affairs.). Secondly, Thomas Aquinas transformed or deleted some Greek concepts like 'dikaion(justice in Greek)' or 'nomisma(money in Greek)', which Aristotle used in explaining his justice, to make his account of justice sensible to Latin readers of that time. In chapter 4, with analyses from chapter 2 and 3, I compared Thomas Aquinas' understanding on justice with contemporary catholic view on justice from Theology of Justice(1973). By doing this, I tried to show some features of catholic church's justice in general, transcending hundreds of years. Finally, I argued some contemporary meanings of Thomas Aquinas' view on justice as a closing remark. 토마스 아퀴나스는 13세기 당시 유럽에는 새로웠던 아리스토텔레스의 윤리학을 자신의 정의론을 구성하는 데에 끌어들였으며, 이 과정에서 기존의 정의에 대한 관점을 일정 부분 기각하거나 통합시킴으로서 자신만의 정의에 대한 관점을 구축해 나갔다. 기존의 토마스 아퀴나스의 정의론에 대한 연구는 토마스 아퀴나스의 정의개념의 역사성을 아예 고려하지 않거나, 고려하더라도 토마스 아퀴나스가 기존의 제 개념들을 어떻게 통합시켰는지에 주목하고 있을 뿐, 토마스의 독창성 및 그가 기각한 부분에 대해서는 그다지 주목하고 있지 않다. 이러한 인식 하에서, 본 논문에서는 토마스 아퀴나스의 정의론을 역사적 맥락에서 고찰하며, 또한 토마스 아퀴나스의 독창성과 그가 기각한 부분은 무엇인지를 집중적으로 조망하고자 한다. 이러한 작업을 수행함에 있어, 본 논문의 2장에서는 토마스 아퀴나스 사상의 지성사적 토대를 검토하기 위하여 크게 아우구스티누스와 아리스토텔레스의 정의론과, 그 근간이 되는 인간론을 우선적으로 검토하였다. 이렇게 토마스 아퀴나스 이전의 정의개념에 대한 검토를 마친 후, 3장에서는 토마스 아퀴나스 자신의 정의론을 구체적으로 살펴보았다. 이 과정에서 토마스 아퀴나스의 정의론을 구성하는 제 개념들 중 아리스토텔레스, 아우구스티누스로부터의 유산은 무엇이고, 차이점은 무엇인지 검토하였다. 그 결과, 토마스 아퀴나스는 단순히 정의에 대한 기존의 제 개념들을 통합하였다기보다는, 정의개념을 fas(divine law)가 아닌 jus(human law; just deed. 토마스 아퀴나스는 정의를 인간들의 문제에 속하는 것으로 보았다.)에 한정함으로써 아우구스티누스의 정의론을 기각하였다는 점을 알 수 있었다. 또한 아리스토텔레스의 정의에 대한 관점과 연관되어 있는 그리스어에서의 고유 어휘들, 예컨대 dikaion과 nomisma 등을 정의 개념을 설명하는 데에 있어서 삭제하고, 라틴 독자들에게 걸맞게 변형하였음을 또한 알 수 있었다. 이렇게 토마스 아퀴나스의 정의개념을 당대의 지성사적 맥락에서 파악한 후에, 4장에서는 이러한 분석을 토대로 Theology of Justice(1973)에서의 정의에 대한 관점과 토마스 아퀴나스의 정의론을 비교함으로서 시대적 차이를 뛰어넘는 가톨릭 정의론 일반에 대하여 개괄하고자 하였다. 이후 토마스 아퀴나스의 정의론이 지니고 있는 현대적 의미에 대하여 짧게 살펴보며 나름의 결론을 내렸다.

「する」動詞에 관한 考察 : 형태적 유형과 의미적 유형의 관계를 중심으로

이정범 숭실대학교 대학원 2008 국내석사

본 논문은 일본어 「する」동사의 의미적 다양성에 관한 것으로, 형태적 유형과의 관계를 중심으로 그 본질을 규명한 것이다. 「する」는 자동사·타동사가 같은 형태로 존재하는 양용동사로서 統語的 측면에서 다양성을 보일 뿐만 아니라 기능적 측면에서도 和語·漢語·外來語 등과 결합해 「VN+する」의 형태로 新語를 만들어내는 造語力이 강한 어휘이다. 이러한 造語力은 실제 언어생활에서 그대로 반영되어 사용빈도에서도 높은 수치를 보인다. 일본어의 경우 일반적으로 자동사와 타동사의 구별은 對格「を」를 취하는가 그렇지 않은가 하는 統語的인 기준과, 開く와 開ける 등과 같이 형태적 기준에 의해 이루어진다. 「する」는 자동사·타동사가 같은 형태로 존재하는 양용동사이므로 문장 안에서 자동사와 타동사를 구별하는데 있어 형태적 기준을 적용하는 것은 불가능하다. 또한 취하는 格도 「を」·「に」·「と」·「が」 등 형태가 다양하고, 같은 格에 의한 표현일지라도 주체의 행위·事象(event) 등 「する」에 미치는 요소가 다수 존재하여 의미가 混在해 있다. 따라서 格에 의한 자동사와 타동사의 구별도 쉽지 않다. 그렇기 때문에 문장 안에서 「する」의 쓰임을 제대로 파악하기 위해서는 형태적 유형의 분류에 따른 분석뿐만 아니라 의미적 유형의 분류를 통한 분석이 요구된다. 이에 본 논문은 「する」의 형태적 유형과 의미적 유형의 분류를 통해 그 관계성을 고찰함으로써 「する」의 의미적 다양성을 규명하고자 하였다. 고찰 결과 형태적으로 對格 「を」를 취해 가장 타동사에 가깝다고 할 수 있는 「~をする」유형에서 자동사의 대표적인 속성으로 취급되는 「상태」를 나타내는 의미적 유형이 존재했다. 한편 「~にする」유형의 경우 정도의 차이는 있으나 모든 유형에서 주체의 행위성이 나타났고 「상태」의 의미는 갖지 않았다. 즉 의미적으로는 「~にする」가 오히려 「~をする」보다 타동사에 가깝다고 할 수 있다. 본 논문에서는 「する」동사의 형태와 의미에 따른 본질적 특징 분석에 초점을 두어 고찰하였다. 종래에는 「VN+する」, 즉 「する」의 기능적 측면에서의 연구가 「VN」영역에 편중되어 진행되어 왔다. 앞으로는 본 연구의 결과를 토대로 「する」에 의한 「VOICE」나 「ASPECT」의 영역까지 연구범위를 확장해갈 수 있을 것으로 기대된다. This study made a searching examination with a property of "SURU" verb in Japanese. "SURU" has a meaning of an intransitive verb and a transitive verb with the same form. In other words it may be said that this verb is various syntactically. Moreover, in the functional side "SURU" is put together with "a peculiar Japanese, Chinese characters, words of foreign origin" and makes new vocabulary. A form of this time is "VN+SURU". Such a coined word is reflected by everyday life and the number of the use frequency appears highly. In the case of Japanese, There are generally two distinction methods of an intransitive verb and a transitive verb. At first, it is a syntactic standard. This is judged from presence of accusative "WO(を)". The second is a method by a morphological standard But as for "SURU", an intransitive verb and a transitive verb emerge with the same form in a sentence. Therefore, Distinction by a morphological standard is impossible. Besides, "SURU" combines with various forms such as "WO(を)·NI(に)·TO(と)·GA(が)". Meanings are various with a form of the status that is the same in it. Therefore, distinction of an intransitive verb by status and a transitive verb is not easy, too. So, Analysis to keep a classification of a type of a meaning and that of a type of a morphological type is essential to grasp that it is used "SURU" in a sentence. So, In the study making a searching examination in essence of "SURU", I put a focus to discuss relations of a morphological type and a semantic type. As a result, There was a type of a meaning to show "a state" of rapid completion of an intransitive verb with a type of "-WOSURU(-をする)" which was near to the first transitive verb at time to look morphologically. On the other hand, In the case of "-NISURU(-にする)" type, There was a difference of degree, but act characteristics of the subject appeared with all types. The meaning of "a state" did not last. In other words, "-NISURU(-にする)" can tell that rather character of a transitive verb is stronger than "-WOSURU(-をする)" semantically. This thesis focused on analysis of an essential characteristic through forms and meanings of "SURU". I expect such a result by laying the foundation to enlarge a study range from now on to a domain of "SURU" by "VOICE" and "ASPECT".

추간판 탈출증 환자에서 퇴행성변화와 추간판 높이의 변화

이정범 충남대학교 대학원 2004 국내석사

The changes of disc height in patients who had been diagnosed with herniated nucleus pulposus(HNP) and undergone discectomy were studied retrogradely. The goal of this study is to find the relationship between the disc height and degenerative disc changes in patients who were diagnosed with HNP and undergone discectomy. The patients in this study are those who were diagnosed with HNP and undergone discectomy in our hospital. The fifty seven patients whom we have observed were those who had discectomy in our hospital from March, 2001 through March, 2003. Patients who had posterior fusion during surgery were excluded. There were 35 males and 22 females, their average age was 35.9 yrs old(17-55) and mean follow-up period was 2.7 yrs. The degree of HNP was examined with MRI, and the degree of degenerative disc was measured through the T2 MRI before operation. The disc heights were measured before and after surgery, and the anterior, posterior and middle disc heights were measured with simple lumbar lateral X-ray. There were 6 cases of degeneration grade Ⅱ, 20 cases of degeneration grade Ⅲ, 25 cases of degeneration grade Ⅳ and 6 cases of degeneration grade Ⅴ. There were also 27 cases of protruded disc, 25 of extruded disc and 5 examples of sequestrated disc. When patients with HNP went through discectomy, the changes of disc height in people with degenerative disc change before surgery were notable, but there was no difference in disc height change regarding the degree of HNP. In HNP patients after discectomy, the changes of disc height decrease significantly in severe degenerative disc for short term period and we need to study more for long term period.

심층 신경망(DNN, Deep Neural Network)을 활용한 PM2.5 단기 예보모델 개발 연구

이정범 안양대학교 일반대학원 2021 국내석사

환경부는 대기 중의 오염물질 중 하나인 PM2.5로 인한 피해를 사전 예방하기 위해서 2015년 1 월부터 19개 권역을 대상으로 PM25 예보를 수치모델을 통해 하고 있다. 하지만, 수치모델을 통 한 예보는 배출량 및 기상자료 등의 입력 자료의 불확실성과 수치모델 자체가 복잡한 대기현상 을완전히 반영하지 못하는 근본적인 한계를 가지고 있다. 따라서 본 연구에서는 서울지역을 대 상으로 인공지능 모델을 개발하였고, 인공지능 모델의 성능을 향상시키기 위해 입력 변수의 구 성을 다르게 하여 모델의 성능을 분석해 보았다. 또한, 선정된 인공지능 모델과 수치모델의 성 능을 비교 분석하여 수치모델의 한계를 보완하고 향상된 예보 성능을 가져오는지 파악하였다. PM2.5를 예보하는 DNN(Deep Neural Network) 모델을 개발하기 위해서, 본 연구에서는 15시 예 보시간을 기준으로 3일 예보를 하는 6시간 간격 Time Table을 구축하였다. 구축한 Time Table 에 따라서 서울지역의데이터를 수집하였다. 학습 데이터 변수의 이상치(Outlier)를 1사분위(Ql) 과 3사분위수(Q3)를 통해 정의되는 2.5IQR(Inter Quartile range)규칙을 통해서 판단하였다. 이 상치는 각각 Q1-2.5IQR 및 C3+2.5IQR 값으로 정의하였다. 정규화 작업은 Min-Max Scaler 방법 을 적용하여 0~1범위 정규화 하였다. 학습 데이터는 2015~2018년이고, 평가 데이터는2019년도 이다. 본 연구에서는 5개의 층(Layer)을 가지는 구조의 DNN 모델을 구축하였고, 오류 역전파(Back-propagation) 학습법을 통해 DNN 모델을 구축하였다. 총 세 가지의 DNN 모델을 하이퍼 파라미터는 모두 동일하게설정하고, 입력되는 변수를 다르게 하여 구축하였다. 세 가지 의 DNN 모델의 R(Corre1ation coefficient), I0A(Index of Agreement), RMSE(Root Mean Square Error) 값은 유의한 것으로 나타났다. 또한 DNN 모델들이 측정값에 과소 모의하는 경향 이 나타나는 것으로 판단이 된다. 따라서 통계적으로 유의한 DNN 모델 중 D+l과 D+2의 P0D (Probabi1ity of Detection)가 가장 높은 0FC1_AC_J 모델을 대표 모델로 선정하였다. 선정된 OFC1_AC_J 모델의 과적합 보정 전, 후 성능을 비교해 보았다. 과적합 보정 후에 ACC (Accuracy)와 FAR(Fa1se Alarm Rate)의 성능은 감소하는 경향을 보였다. 또한, 새로운 과적합 값이 나타나면서 POD의 성능은 유사하거나 증가하는 것으로 나타났다. 따라서 본 연구에서 사 용한 과적합 보정 방법은 적용하지 않는 것이 유리하다고 판단이 된다. 과적합 보정을 하지 않 은 OFC1_AC_J 모델과 CASE4 모델의 성능평가 결과, ACC와 FAR는 OFC1_AC_J 모델이 유리한 것으로 나타났고, POD는 CASE4 모델이 유리한 것으로 나타났다. CASE4 모델의 경우 NMB (Norma1ized Mean Bias)값이 일별로 약 23~32%로 측정값에 과대모의하는 경향이 있어 POD 의 성능이 뛰어난 것으로 판단이 된다. 반면, OFC1_AC_J 모델은 NMB 값이 일별로 -5~1%로 과대 모의하는 CASE4 모델의 특성을 보완하는 것으로 나타났다. 따라서 본 연구에서는 현업 에서 사용 중인 수치모델의 한계를 보완하는 모델로서 인공지능 모델인 DNN 모델의 사용 가능 성이 있다고 판단하였고, ACC와 FAR의 측면에서는 향상된 예보가 가능하다고 판단하였다. In order to prevent the public health from PM (Particulate Matter) exposure, the Korea Ministry of Environment has been providing PM2.5 forecast for 19 regions using numehcal models since January 2015. However, air quality forecasts by numehcal models have intrinsic limitations such as uncertainties in input data of emissions and weather forecast data as well as the numehcal model itself. The artificial intelligence model based on DNN(Deep Neural Network) algorithm has been developed to overcome shortcoming of the numehcal model in this study. In order to develop the DNN model for PMa.s forecasting, forecasting time table is established by 6 hours in 3 forecasts days. According to the time table, data from Seoul area was collected. The outlier of the DNN learning input vahable was determined by the 2.5IQR(Inter Quartile Range) rules defined by values of QKthe first quarti1e)-2.5IQR and Q3(the third quarti1e)+2.5IQR. Normalization was performed range from 0 to 1 by applying the Min-Max scaler method and then normalized data was divided into the learning data from 2015 to 2018 and the validation data for 2019. In this study, the DNN model consists of 5 layers stmcture and back propagation leaming methods which is a representative method of supervised learning to obtain optimal weigh and cost. Three DNN models were developed based on various combination of input variables under same hyper-parameters. The performance of three DNN models showed similar statistics values of R(Corre1ation coefficient), I0A(Index of Agreement), and RMSE (Root Mean Square Error) compahng with air quality measurements. Among the DNN models, we choose a representative model, 0FCL_AC_J, which had the highest P0D (probabi1ity of detection) of D+l and D+2. The performance of 0FC1_AC_J model with overfitting correction was decreased ACC(Accuracy) and FAR(Fa1se Alarm Rate) and was increased or similar to POD compared with model without overfitting. It implies that the overfitting correction is not vahable method to improve the forecasting accuracy in this DNN model. We compared the performance of the 0FC1_AC_J model without overfitting correction and the CASE4 model. The 0FC1_AC_J model showed improvements in ACC and FAR and a decrease in POD compared with CASE4 model. CASE4 model results tends to overestimate compahng with measurements as 23-32% in NMB(Norma1ized Mean Bias), resulting in POD high performance, ACC and FAR low performance. On the other hand, the 0FC1_AC_J model improved -5~1% in NMB per day compared with CASE4 model, resulting in ACC, FAR high performance and POD low performance The results in thls study demonstrate that artificial intelligence model based on the DNN algohthm is a feasible method that could overcome limits of the numerical model by improving ACC and FAR in PM2.5 forecasting.

포르말린 반응 실험에서 척수강내 병행 투여한 Morphin 과 Magnesium의 항 침해효과

이정범 강원대학교 대학원 2005 국내석사

대기업의 인터넷 마케팅 사례연구 : 오프라인과 온라인의 통합을 중심으로

이정범 연세대학교 경영대학원 2002 국내석사

The major obstacle facing the on-line businesses is relative lack of off-line basis or capability. Most of the on-line businesses were launched with poor off-line business basis or experiences. Since the Internet marketing is still less reliable, the conventional off-line businesses still prosper despite their high marketing costs. Quite ironically, in this age of Internet, the traditional off-line businesses are being given more chances. For example, the off-line businesses can reduce their costs further by reducing the steps of channel or purchasing, while interacting with customers only to respond to their needs more quickly than on-line businesses. With such basic conceptions in mind, this study was aimed at analyzing the success cases of large businesses who have integrated their off-line marketing into the on-line one in the new Internet environment. The integrated on-and off-line marketing may use the resources and methods of both systems. The integrated marketing may be characterized by three-fold features ; First, new on-line marketing activities may be combined with the traditional off-line marketing in an organistic way. Second, the new on-line marketing activities may support the conventional off-line marketing activities. Third, new off-line marketing activities may be unfolded independently from the on-line marketing activities. In the first case, two kinds of marketing activities are performed separated but for a synergic effect, while in the second and third cases, emphasis would be put on either marketing and the other marketing activities would be supported. In a broader terms, the integrated marketing means that the off-line marketing will be gradually replaced with the on-line marketing. As people spend more hours in the Internet and thereby, our society undergoes a great change with diversifying economic activities, either on-line or off-line alone cannot appeal perfectly to consumers. Therefore, it is deemed essential to overcome the limits of the off-line marketing, while approaching the consumers staying on-line. This is nothing other than the integrated marketing concept. For this case study, those large companies who have succeeded in integrating on-line and off-line marketing activities were sampled to review the synergic effects of the integrated marketing. The results of this study can be summarized as follows ; SK Inc., invented a new type of marketing called "OK Cash-back" based on the database constructed of existing SK Telecom and Enclean customers and has succeeded in establishing the new marketing system. The "OK Cash-back" site is associated with some two thousand off-line businesses to create a synergic effect. SK "OK Cash-back" marketing program must be a success case of integrated on-and off-line marketing. On the other hand, Korea Telecom Inc., supplemented the reduced wire business sales with the on-line marketing. The existing wire business recorded lower sales, but their Internet-dedicated line rental business and ADSL and other Internet-related businesses recorded a high blue ink. Samsung Life Insurance Inc., has designed an on-line marketing program which supports the off-line marketing forces who sell 90% of the whole insurance commodities. By separating the off-line commodities from the on-line counterparts, the company has succeeded in creating a synergic effect, while solving the conflicts between two marketing sectors. As a result of analyzing the success cases of integrating on-line and off-line marketing, it was confirmed that the successful companies differentiated their integrated marketing strategies, while adapting it to their business conditions and establishing a marketing system for the symbiosis between two marketing sectors. The off-line marketing which only supported the on-line one or the on-line for off-line could not succeed. In other words, when on-line supports off-line which in turn support on-line, the integrated marketing strategies could succeed.

海左 丁範祖의 長篇古詩 硏究

류정민 고려대학교 2001 국내석사