http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
정석원 ( Seok-won Jeong ),최맹식 ( Maengsik Choi ),김학수 ( Harksoo Kim ) 한국정보처리학회 2016 정보처리학회논문지. 소프트웨어 및 데이터 공학 Vol.5 No.11
상호참조 해결은 문서 내에 존재하는 멘션들을 식별하고, 참조하는 멘션끼리 군집화하는 것으로 정보 추출, 사건 추적, 질의응답과 같은 자연어처리 응용에 필수적인 과정이다. 최근에는 기계학습에 기반한 다양한 상호참조 해결 모델들이 제안되었으며, 잘 알려진 것처럼 이런 기계학습 기반 모델들은 상호참조 멘션 태그들이 수동으로 부착된 대량의 학습 데이터를 필요로 한다. 그러나 한국어에서는 기계학습 모델들을 학습할 가용한 공개 데이터가 존재하지 않는다. 그러므로 본 논문에서는 다른 기계학습 모델보다 적은 학습 데이터를 필요로 하는 효율적인 상호참조 해결 모델을 제안한다. 제안 모델은 시브-가이드 자질 기반의 랜덤 포레스트를 사용하여 상호참조하는 멘션들을 구분한다. 야구 뉴스 기사를 이용한 실험에서 제안 모델은 다른 기계학습 모델보다 높은 0.6678의 CoNLL F1-점수를 보였다. Coreference resolution is to identify mentions in documents and is to group co-referred mentions in the documents. It is an essential step for natural language processing applications such as information extraction, event tracking, and question-answering. Recently, various coreference resolution models based on ML (machine learning) have been proposed, As well-known, these ML-based models need large training data that are manually annotated with coreferred mention tags. Unfortunately, we cannot find usable open data for learning ML-based models in Korean. Therefore, we propose an efficient coreference resolution model that needs less training data than other ML-based models. The proposed model identifies co-referred mentions using random forests based on sieve-guided features. In the experiments with baseball news articles, the proposed model showed a better CoNLL F1-score of 0.6678 than other ML-based models.
정석원(Seok-won Jeong),최맹식(Maengsik Choi),김학수(Harksoo Kim) 한국정보과학회 2015 정보과학회논문지 Vol.42 No.8
지식베이스는 자연어 처리 기반의 다양한 응용 시스템 성능에 영향을 미치는 중요한 요소이다. 영어권에서는 WordNet, YAGO, Cyc, BabelNet과 같은 지식베이스들이 널리 사용되고 있다. 본 논문에서는 위키백과와 YAGO로부터 YAGO 형식의 한국어 지식베이스(이하 K-YAGO)를 자동 구축하는 방법을 제안한다. 제안 시스템은 YAGO와 위키백과 인포박스간의 간단한 매칭을 통해 초기 K-YAGO를 구축한 뒤, 기계학습을 이용하여 초기 K-YAGO를 확장한다. 실험 결과, 제안 시스템은 초기 K-YAGO 구축 실험에서 0.9642의 신뢰도를 보였고, K-YAGO 확장 실험에서 0.9468의 정확도와 0.7596의 매크로 F1 척도를 보였다. The performance of many natural language processing applications depends on the knowledge base as a major resource. WordNet, YAGO, Cyc, and BabelNet have been extensively used as knowledge bases in English. In this paper, we propose a method to construct a YAGO-style knowledge base automatically for Korean (hereafter, K-YAGO) from Wikipedia and YAGO. The proposed system constructs an initial K-YAGO simply by matching YAGO to info-boxes in Wikipedia. Then, the initial K-YAGO is expanded through the use of a machine learning technique. Experiments with the initial K-YAGO shows that the proposed system has a precision of 0.9642. In the experiments with the expanded part of K-YAGO, an accuracy of 0.9468 was achieved with an average macro F1-measure of 0.7596.
저온 O<sub>sub2</sub> 어닐링 공정을 통한 HfSi<sub>x</sub>O<sub>y</sub>의 전기적 특성 개선
이정찬,김광숙,정석원,노용한,Lee, Jung-Chan,Kim, Kwang-Sook,Jeong, Seok-Won,Roh, Yong-Han 한국전기전자재료학회 2011 전기전자재료학회논문지 Vol.24 No.5
We investigated the effects of low temperature ($500^{\circ}C$) $O_2$ annealing on the characteristics of hafnium silicate ($HfSi_xO_y$) films deposited on a Si substrate by atomic layer deposition (ALD). We found that the post deposition annealing under oxidizing ambient causes the oxidation of residual Hf metal components, resulting in the improvement of electrical characteristics such as flat band voltage shift (${\Delta}V_{fb}$) by hysteresis without oxide capacitance reduction. We suggest that post deposition annealing under oxidizing ambient is necessary to improve the electrical characteristics of $HfSi_xO_y$ films deposited by ALD.
기판 온도 변화에 따른 AlN 박막 성장에 잔류 산소가 미치는 영향
김병균,이을택,김응권,정석원,노용한,Kim, Byoung-Kyun,Lee, Eul-Tack,Kim, Eung-Kwon,Jeong, Seok-Won,Roh, Yong-Han 한국전기전자재료학회 2008 전기전자재료학회논문지 Vol.21 No.5
Aluminum nitride (AlN) thin films have been deposited on Au electrodes by using reactive RF magnetron sputtering method in a gas mixture of Ar and $N_2$ at different substrate temperature. It was found that substrate temperature was varied in the range up to $400^{\circ}C$, highly c-axis oriented film can be obtained at $300^{\circ}C$ with full width at half maximum (FWHM) $3.1^{\circ}$. Increase in surface roughness from 3.8 nm to 5.9 nm found to be associated with increase in grain size, with substrate temperature; however, the AlN film fabricated at $400^{\circ}C$ exhibited a granular type of structure with non-uniform grains. The Al 2p and N 1s peak in the X-ray photoelectron spectroscopy (XPS) spectrum confirmed the formation of Al-N bonds. The XPS spectrum also indicated the presence of oxynitrides and oxides, resulting from the presence of residual oxygen in the vacuum chamber. It is concluded that the AlN film deposited at substrate temperature of $300^{\circ}C$ exhibited the most desirable properties for the application of high-frequency surface acoustic devices.
조기 대장암의 내시경 절제술 후 림프절 전이의 위험 인자
임경훈 ( Kyung Hoon Lim ),정석원 ( Seok Won Jung ),정인두 ( In Du Jeong ),방성조 ( Sung Jo Bang ),신정우 ( Jung Woo Shin ),박능화 ( Neung Hwa Park ),최혜정 ( Hye Jeong Choi ),최대화 ( Dae Hwa Choi ),김도하 ( Do Ha Kim ) 대한장연구학회 2008 Intestinal Research Vol.6 No.2
Background/Aims: Although endoscopic resection is widely used for the treatment of early colorectal cancer, the risk factors for lymph node metastasis are not clear. This study was designed to determine the risk factors for lymph node metastasis in patients with colorectal cancer who are treated by endoscopic resection. Methods: The medical records of patients with histologically-proven early colorectal cancers who were treated by endoscopic resection between January 2002 and September 2008 were retrospectively reviewed. Information regarding the demographic data of patients and the clinicopathologic characteristics were recorded and analyzed. Results: Twenty-nine patients who underwent subsequent surgical treatment after endoscopic resection for early colorectal cancer were enrolled in this study. Six patients (20.7%) had lymph node metastases on surgical pathologic examination. The predictive factors for lymph node metastasis were tumor morphology (non-polypoid flat tumors [p=0.019]), absence of background adenomas (p=0.033), and deep submucosal invasion ≥2,000 um (p=0.012). Unexpectedly, the presence of vascular invasion was not associated with lymph node metastasis. Conclusions: The presence of vascular invasion might not be an absolute indication for additional surgical treatment of early colorectal cancer; however, deep submucosal invasion, accompanied by a gross tumor with a non-polypoid flat morphology, and the absence of background adenomas are potential risk factors for lymph node metastasis. (Intest Res 2008;6:103-109)