• Proceedings of the Korea Information Processing Society Conference
• /
• 2009.04a
• /
• pp.226-229
• /
• 2009

최근 들어 인터넷 게시판이나 개인 블로그 등은 온라인상에서 사람들의 정보 공유나 의견 교환의 중요한 매체가 되고 있다. 많은 수의 블로그들은 현재 사회적으로 이슈가 되는 여러 문제들을 반영하고 있다. 또한 최근 댓글을 통해 적극적으로 자신의 의사 표현하거나 다른 사람들의 의견을 살피는 인터넷 사용자의 증가로 인터넷 뉴스나 블로그 기사에 많은 수의 댓글이 달리고 있다. 그러나 대부분의 블로그나 인터넷 포털 사이트의 경우 기사나 댓글들을 순차적인 목록 형태로 제공하므로 자신이 원하는 내용의 댓글을 검색하거나 전체 댓글에 대한 전반적인 파악은 힘든 일이다. 따라서 본 논문에서는 기사에 달린 많은 수의 댓글들을 분류하고, 이를 시각화 하는 시스템인 TRIB(Telescope for Responding comments for Internet Blog)을 제안한다. TRIB은 미리 정의된 사용자 정의 사전을 이용하여 댓글을 내용에 따라 분류하여 시각화 하므로 사용자들은 자신의 관심과 흥미에 따라 개인화 된 뷰를 볼 수 있다. 1,000개 이상의 댓글을 가진 뉴스 기사들을 대상으로 한 실험을 통해 TRIB 시스템의 댓글 분류와 시각화 성능을 보인다.

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.532-535
• /
• 2021

다양한 SNS 플랫폼이 등장하고, 이용자 수가 급증함에 따라 온라인에서 얻을 수 있는 정보의 활용 가치가 높아지고 있다. 문장은 자연어 처리 시스템의 기본적인 단위이므로 주어진 문서로부터 문장의 경계를 인식하는 작업이 필수적이다. 공개된 문장 경계 인식기는 SNS 문서에서 좋은 성능을 보이지 않는다. 본 논문에서는 문어체로 구성된 일반 문서뿐 아니라 SNS 문서에서 사용할 수 있는 문장 경계 인식기를 제안한다. 본 논문에서는 SNS 문서에 적용하기 위해 다음과 같은 두 가지를 개선한다. 1) 학습 말뭉치를 일반문서와 SNS 문서 두 영역으로 확장하고, 2) 이모티콘을 사용하는 SNS 문서의 특징을 반영하는 어절의 유형을 자질로 추가하여 성능을 개선한다. 실험을 통해서 추가된 자질의 기여도를 분석하고, 또한 기존의 한국어 문장 경계 인식기와 제안한 모델의 성능을 비교·분석하였다. 개선된 모델은 일반 문서에서 99.1%의 재현율을 보이며, SNS 문서에서 88.4%의 재현율을 보였다. 두 영역 모두에서 문장 경계 인식이 잘 이루어지는 것을 확인할 수 있었다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2008.11a
• /
• pp.1500-1503
• /
• 2008

H.264와 같은 고압축 비디오처리 기법의 등장으로 기존의 MPEG2와 같은 비디오 압축에서 H.264로의 비디오 트랜스코딩이 증가되고 있지만, 고압축 비디오 콘텐츠의 온라인과 오프라인에서 불법배포는 현재 문제가 되고 있다. 본 논문에서는 다이렉트쇼 환경 기반에서 고압축과 저작권 보호를 위한 비디오 트랜스 코딩과 워터마킹을 구현한다. 제안한 방법은 다이렉트쇼의 필터를 이용하여 MPG,WMV를 H.264로 비디오 트랜스코딩을 하고 이와 함께 비디오의 공간영역 특성을 이용하여 저작권 보호를 위한 강인한 워터마킹을 구현한다. 실험 결과 MPG,WMV를 H.264로 트랜스코딩에서 H.264의 QP(Quantization parameter)를 15로 하고 화면간 반복을 10프레임으로 하였을 경우 저작권 보호를 위하여 삽입된 워터마크는 평균 99% 검출됨을 확인하였고, 또한 트랜스코딩중 워터마크삽입에 따른 시간지연은 전체 트랜스코딩시간의 5.7%가 됨을 확인할 수 있었다. 제안한 방법은 저작권 삽입 기능가지는 트랜스코딩 소프트웨어를 필요로 하는 Digital TV방송, IPTV, DVD 사업에 사용 될 수 있을 것이다.

• The Journal of the Acoustical Society of Korea
• /
• v.24 no.7
• /
• pp.423-430
• /
• 2005

In this paper. a microphone array beamforming algorithm that reduces the signal distortion caused by reverberation and near-field effect in car environment is proposed. When reverberation or near-field effect is present, an optimum beamformer should be constructed with a steering vector consisting of transfer functions between source and microphones, but it is generally difficult to estimate transfer functions on-line without knowledge of the source signal. Instead, a sub-optimal beamforming algorithm that reduces signal distortion is proposed. It is constructed with steering vectors consisting of relative transfer functions between reference sensor and other sensors. In order to evaluate the performance of the proposed algorithm. we had recorded noisy speech database in a car, and performed speech recognition experiments with HMM Toolkit (HTK) released by Cambridge University. The recognition rate of the proposed algorithm was 15 percents higher than that of the conventional far-field beamformers in best case.

• Journal of KIISE:Computer Systems and Theory
• /
• v.27 no.3
• /
• pp.324-336
• /
• 2000

The paper proposes an efficient buffer management scheme to enhance performance of the disk I/O system. Without any user level information, the proposed scheme automatically detects the block reference patterns of applications by associating block attributes with forward distance of a block. Based on the detected patterns, the scheme applies an appropriate replacement policy to each application. We also present a new block allocation scheme to improve the performance of buffer cache when kernel needs to allocate a cache block due to a cache miss. The allocation scheme analyzes the cache hit ratio of each application based on block reference patterns and allocates a cache block to maximize cache hit ratios of system. These all procedures are performed on-line, as well as automatically at system level. We evaluate the scheme by trace-driven simulation. Experimental results show that our scheme leads to significant improvements in hit ratios of cache blocks compared to the traditional schemes and requires low overhead.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.41 no.6
• /
• pp.69-80
• /
• 2004

This paper proposes a method to learn action patterns of opponent characters for intelligent characters. For learning action patterns, intelligent characters learn the past actions as well as the current actions of opponent characters. Therefore, intelligent characters react more properly than ones without the knowledge on action patterns. In addition, this paper proposes a method to learn moving actions whose fitness is hard to evaluate. To evaluate the performance of the proposed algorithm, we experiment with four repeated action patterns in a game similar to real games. The results show that intelligent characters learn the optimal actions for action patterns and react properly against to random action opponent characters. The proposed method can be applied to various games in which characters confront each other, e.g. massively multiple of line games.

• Journal of Digital Contents Society
• /
• v.10 no.4
• /
• pp.537-542
• /
• 2009

The constraints of printed text have been lifted through developments in computer technology which has been identified as a revolutionary force. Hypertexts can be simply defined as electronic text that is found online, in a non-linear manner. In contrast to traditional printed texts, electronic writing depends upon an emergent technology, which is still subject to transformation. Unfortunately more research is needed on the experiences readers have when reading documents in hypertext formats for the purpose of knowledge retention. This study is to research the contrast between the traditional printed texts and hypertexts. Other areas where the literature has been relatively silent will be explored such as the experiences subjects have in reading hypertexts, and printed texts. It was clearly founded that the format of text does significantly influence the recall comprehension level of readers in the Printed Text and Hypertext Groups.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.2C
• /
• pp.306-314
• /
• 2004

Some receiver(and most transmit) beamforming algorithms with an array antenna at a cellular CDMA base stations require accurate internal and external calibrations. The external calibration, which usually needs to be done only once, determines the array manifolds, i.e. the complex response of each antenna as a function of DOA(Directions of Arrival). The internal calibrations are necessary because characteristics of RF/IF circuity of each receiver chain vary differently in response to temperature or humidity changes. We propose an iterative subspace-based calibration algorithm for an asynchronous CDMA-based antenna away in the presence of unknown gain and phase error is presented. We verify the subspace-based calibration algorithms by performing the experiment using measured data. Also, we propose an efficient algorithm using the simulated annealing technique. This algorithm overcomes the problem of the initial guessing in the subspace-based approach.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.7 no.9
• /
• pp.317-325
• /
• 2017

In this paper, we try to classify Korean hedge sentences, which are regarded as not important since they express uncertainties or personal assumptions. Through previous researches to English language, we found dependency information of words has been one of important features in hedge classification, but not used in Korean researches. Additionally, we found that word embedding vectors include the word usage information. We assume that the word usage information could somehow represent the dependency information. Therefore, we utilized word embedding and neural networks in hedge sentence classification. We used more than one and half million sentences as word embedding dataset and also manually constructed 12,517-sentence hedge classification dataset obtained from online news. We used SVM and CRF as our baseline systems and the proposed system outperformed SVM by 7.2%p and also CRF by 1.2%p. This indicates that word usage information has positive impacts on Korean hedge classification.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.7
• /
• pp.19-27
• /
• 2023

In the process of collecting social reviews, a number of noise reviews irrelevant to a given search keyword can be included in the search results. To filter out such reviews, machine learning can be used. However, if the number of reviews is insufficient for a target place to be analyzed, filtering accuracy can be degraded due to the lack of training data. To resolve this issue, we propose a supervised learning method to improve accuracy of the noise review filtering for the places with insufficient reviews. In the proposed method, training is not performed by an individual place, but by a group including several places with similar characteristics. The classifier obtained through the training can be used for the noise review filtering of an arbitrary place belonging to the group, so the problem of insufficient training data can be resolved. To verify the proposed method, a noise review filtering model was implemented using LSTM and BERT, and filtering accuracy was checked through experiments using real data collected online. The experimental results show that the accuracy of the proposed method was 92.4% on the average, and it provided 87.5% accuracy when targeting places with less than 100 reviews.