• Annual Conference on Human and Language Technology
• /
• 2005.10a
• /
• pp.7-12
• /
• 2005

본 논문은 은닉 마코프 모델(hidden Markov Model: HMM)을 2 단계로 적용하여 논문 모집공고(Call-for-Paper: CFP)에서 필요한 정보를 추출하는 방법을 제안한다. HMM은 순차적인 흐름의 정보를 담고 있는 데이터를 잘 설명할 수 있으며 CFP가 담고 있는 정보에는 순서가 있기 때문에, CFP를 HMM으로 설명할 수 있다. 하지만, 문서를 전체적으로(global) 파악하는 HMM만으로는 정보의 정확한 경계를 파악할 수 없다. 따라서 첫 번째 단계로 CFP문서에서 구(phrase) 단위를 구성하는 단어의 열에 대한 HMMs을 통해 국부적으로(local) 정보의 경계와 대강의 종류를 파악한다. 그리고 두 번째 단계에서 전체적인 문서의 내용 흐름에 근거하여 구축된 HMM을 이용하여 그 정보가 세부적으로 어떤 종류의 정보인지 정한다. PASCAL challenge에서 제공받은 Cff 말뭉치에 대한 첫 번째 단계의 실험 결과, 0.60의 재현률과 0.61의 정확률을 보였으며, 정확률과 재현률을 바탕으로 F-measure를 측정한 결과 0.60이었다.

• Annual Conference on Human and Language Technology
• /
• 2011.10a
• /
• pp.9-15
• /
• 2011

여러 언어로 작성되는 위키피디아의 경우 언어 간에 등록되어 있는 정보의 양과 내용이 달라 언어 간 정보를 상호 추출하고 서로 통합하는 연구에 대한 관심이 증가하고 있다. 특히, 위키피디아의 요약본으로써 의미가 있는 인포박스는 위키피디아 아티클에 존재하는 구조화된 정보 중 가장 근간이 되는 정보로, 본 논문에서는 위키피디아에 존재하는 인포박스를 1)소스 언어 자원으로부터 획득하여 타겟 언어로 번역하고, 2)번역된 결과물과 웹그래프를 이용하여 타겟 언어 데이터에서 획득하는 정보와 결합하는 과정을 통해 자동으로 인포박스를 생성하는 기법에 대하여 설명한다. 웹그래프는 위키피디아에 존재하는 링크 구조를 통해 서로 다른 두 용어간의 관련도를 측정하여 인포박스에 추가될 내용을 파악하는데 사용한다. 본 논문의 기법은 언어 간 인포박스를 생성하는 측면에서, 영어 인포박스 데이터를 입력으로 하여 한국어 인포박스 데이터를 생성하는 방식으로 진행하였다. 평가를 위하여 기존 한국어에 실제 존재하는 인포박스 데이터와 비교 실험하는 방식을 사용하여 평균적으로 40%의 정확률과 83%의 재현율을 나타내었다. 하지만, 기존 한국어에 존재하는 인포박스 데이터의 내용이 인포박스에 포함될 완전한 데이터를 모두 포함했다고 볼 수 없으므로 본 논문에서 제안하는 수행한 실험의 정확률이 상대적으로 낮게 나온 것으로 분석되었다. 실제 사람이 수작업으로 새롭게 생성된 인포박스 데이터의 적합성을 판별한 경우 평균 76%의 정확률과 91%의 재현율을 나타내었다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.512-513
• /
• 2023

CNN 기반 Deep Learning 분야에서 객체 탐지 정확도를 높이기 위해 모델의 많은 Parameter 가 사용된다. 많은 Parameter 를 사용하게 되면 최소 하드웨어 성능 요구치가 상승하고 처리속도도 감소한다는 문제가 있어, 최소한의 정확도 하락으로 Parameter 를 줄이기 위한 여러 Pruning 기법이 사용된다. 본 연구에서는 Neural Architecture Search(NAS) 기반 Channel Pruning 인 Artificial Bee Colony(ABC) 알고리즘을 사용하였고, 기존 NAS 기반 Channel Pruning 논문들이 Classification Task 에서만 실험한 것과 달리 Object Detection Task 에서도 NAS 기반 Channel Pruning 을 적용하여 기존 Uniform Pruning 과 비교할 때 파라미터 수 대비 정확도가 개선됨을 확인하였다.

• Journal of the Korean Society for Library and Information Science
• /
• v.44 no.1
• /
• pp.117-135
• /
• 2010

We designed a structured folksonomy system in which queries can be expanded through tag control; equivalent, synonym or related tags are bound together, in order to improve the retrieval efficiency (recall and precision) of image data. Then, we evaluated the proposed system by comparing it to a tag-based system without tag control in terms of recall, precision, and user satisfaction. Furthermore, we also investigated which query expansion method is the most efficient in terms of retrieval performance. The experimental results showed that the recall, precision, and user satisfaction rates of the proposed system are statistically higher than the rates of the tag-based system, respectively. On the other hand, there are significant differences among the precision rates of query expansion methods but there are no significant differences among their recall rates. The proposed system can be utilized as a guide on how to effectively index and retrieve the digital content of digital library systems in the Library 2.0 era.

• Journal of KIISE:Software and Applications
• /
• v.37 no.6
• /
• pp.470-479
• /
• 2010

In relation to the extraction of opinion words, it may be difficult to directly apply most of the methods suggested in existing English studies to the Korean language. Additionally, the manual method suggested by studies in Korea poses a problem with the extraction of opinion words in that it takes a long time. In addition, English thesaurus-based extraction of Korean opinion words leaves a challenge to reconsider the deterioration of precision attributed to the one to one mismatching between Korean and English words. Studies based on Korean phrase analyzers may potentially fail due to the fact that they select opinion words with a low level of frequency. Therefore, this study will suggest the k-Structure (k=5 or 8) method, which may possibly improve the precision while mutually complementing existing studies in Korea, in automatically extracting opinion words from a simple sentence in a given Korean product review. A simple sentence is defined to be composed of at least 3 words, i.e., a sentence including an opinion word in ${\pm}2$ distance from the attribute name (e.g., the 'battery' of a camera) of a evaluated product (e.g., a 'camera'). In the performance experiment, the precision of those opinion words for 8 previously given attribute names were automatically extracted and estimated for 1,868 product reviews collected from major domestic shopping malls, by using k-Structure. The results showed that k=5 led to a recall of 79.0% and a precision of 87.0%; while k=8 led to a recall of 92.35% and a precision of 89.3%. Also, a test was conducted using PMI-IR (Pointwise Mutual Information - Information Retrieval) out of those methods suggested in English studies, which resulted in a recall of 55% and a precision of 57%.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.25 no.4
• /
• pp.63-80
• /
• 2022

As a technology to analyze and predict an issue has been developed by constructing real space into virtual space, it is becoming more important to acquire precise spatial information in complex cities. In this study, images were acquired using an unmanned aerial vehicle for urban area with complex landscapes, and land cover classification was performed object-based image analysis and semantic segmentation techniques, which were image classification technique suitable for high-resolution imagery. In addition, based on the imagery collected at the same time, the replicability of land cover classification of each artificial intelligence (AI) model was examined for areas that AI model did not learn. When the AI models are trained on the training site, the land cover classification accuracy is analyzed to be 89.3% for OBIA-RF, 85.0% for OBIA-DNN, and 95.3% for U-Net. When the AI models are applied to the replicability assessment site to evaluate replicability, the accuracy of OBIA-RF decreased by 7%, OBIA-DNN by 2.1% and U-Net by 2.3%. It is found that U-Net, which considers both morphological and spectroscopic characteristics, performs well in land cover classification accuracy and replicability evaluation. As precise spatial information becomes important, the results of this study are expected to contribute to urban environment research as a basic data generation method.

• The Journal of Korean Society for Radiation Therapy
• /
• v.19 no.2
• /
• pp.123-129
• /
• 2007

Purpose: CT scan shows that significant tumor movement occurs in lesions located in the proximity of the heart, diaphragm, and lung hilus. There are differences concerning three kinds of type to get images following the Scan type called Axial, Helical, Cine (4D-CT) mode, when the scanning by CT. To know how each protocol describe accurately, this paper is going to give you reappearance using the moving phantom. Materials and Methods: To reconstruct the movement of superior-inferior and anterior-posterior, the manufactured moving phantom and the motor following breathing were used. To distinguish movement from captured images by CT scanning, a localizer adhered to the marker on the motor. The moving phantom fixed the movement of superior-inferior upon 1.3 cm /1 min. The motor following breathing fixed the movement of anterior-posterior upon 0.2 cm /1 min. After fixing each movement, CT scanning was taken by following the CT protocols. The movement of A localizer and volume-reappearance analyzed by RTP machine. Results: Total volume of a marker was 88.2 $cm^3$ considering movement of superior-inferior. Total volume was 184.3 $cm^3$. Total volume according to each CT scan protocol were 135 $cm^3$ by axial mode, 164.9 $cm^3$ by helical mode, 181.7 $cm^3$ by cine (4D-CT) mode. The most closely describable protocol about moving reappearance was cine mode, the marker attached localizer as well. Conclusion: CT scan should reappear concerning a exact organ-description and target, when the moving organ is being scanned by three kinds of CT protocols. The cine (4D-CT) mode has the advantage of the most highly reconstructible ability of the three protocols in reappearance of the marker using a moving phantom. The marker on the phantom has always regular motion but breathing patients don't move like a phantom. Breathing education and devices setting patients were needed so that images reconstruct breathing as exactly as possible. Users should also consider that an amount of radiation to patients is being bombed.

• Journal of the Korean Society of Hazard Mitigation
• /
• v.9 no.6
• /
• pp.99-109
• /
• 2009

In this study, often the spatial distribution of precipitation was analyzed using the quantile with regional frequency analysis and spatial analysis to find out the detail distribution of extreme precipitation for preventing the disaster in the region of Gangwon. The hourly precipitation data of 66 stations in Gangwon were used. As the results of regional frequency analysis, it shows that the generalized logistic (GLO) distribution is the best for the region of Gangwon. As the results of spatial analysis, the quaniles have high vaules nearby Seolakdong, Daegwallyeong and Cheongil as the duration of precipitation increase, and the change of spatial distribution occurs severely according to the duration of precipitation. The spatial characteristics of precipitation appears clearly as the return period of quantile increases. As the results of the spatial distribution of precipitation in Gangwon heavy quantiles usually are appeared in Yongdong, and the spatial distributions of quantile in Yongseo are various according to the duration and the return period of quantile. Therefore, to estimate more accurate quantiles in Gangwon, various geographical and weather conditions are considered additionally for the regional precipitation frequency analysis.

• Journal of Korea Multimedia Society
• /
• v.17 no.1
• /
• pp.43-51
• /
• 2014

Social Network Service has the sufficient potential can be widely and effectively used for various fields of society because of convenient accessibility and definite user opinion. Above all Twitter has characteristics of simple and open network formation between users and remarkable real-time diffusion. However, real analysis is accompanied by many difficulties because of semantic analysis in 140-characters, the limitation of Korea natural language processing and the technical problem of Twitter is own restriction. This thesis paid its attention to human's political attitudes showing permanence and assumed that if applying it to the analytic design, it would contribute to the increase of precision and showed it through the experiment. As a result of experiment with Tweet corpus gathered during the election of national assemblymen on 11st April 2012, it could be known to be considerably similar compared to actual election result. The precision of 75.4% and recall of 34.8% was shown in case of individual Tweet analysis. On the other hand, the performance improvement of approximately 8% and 5% was shown in by-timeline political attitude analysis of user.

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.8
• /
• pp.141-148
• /
• 2010

In mobile terminal recommend service system has general information restrictive recommend that individuality considering to user's information find and recommend. Also it has difficult of accurate information recommend bad points user's not offer individuality information preference recommend service. Therefore this paper is propose user's information individuality preference considering by user's individuality preference recommendation system using improved k-means algorithm. Propose method is correlation coefficients using user's information individuality preference when user's individuality preference recommendation using improved k-means algorithm. Restrictive information recommend to fix a problem, information of restrictive general recommend that user's information individuality preference offer to accurate information recommend. Performance experiment is existing service system as compared to evaluating the effectiveness of precision and recall, performance experiment result is appear to precision 85%, recall 68%.