• Journal of KIISE
• /
• v.44 no.5
• /
• pp.496-502
• /
• 2017

Pointer Networks is a deep-learning model for the attention-mechanism outputting of a list of elements that corresponds to the input sequence and is based on a recurrent neural network (RNN). The coreference resolution for pronouns is the natural language processing (NLP) task that defines a single entity to find the antecedents that correspond to the pronouns in a document. In this paper, a pronoun coreference-resolution method that finds the relation between the antecedents and the pronouns using the Pointer Networks is proposed; furthermore, the input methods of the Pointer Networks-that is, the chaining order between the words in an entity-are proposed. From among the methods that are proposed in this paper, the chaining order Coref2 showed the best performance with an F1 of MUC 81.40 %. The method showed performances that are 31.00 % and 19.28 % better than the rule-based (50.40 %) and statistics-based (62.12 %) coreference resolution systems, respectively, for the Korean pronouns.

• KIISE Transactions on Computing Practices
• /
• v.24 no.2
• /
• pp.93-98
• /
• 2018

Recommendation systems are used to provide items of interests for users to maximize a company's profit. Matrix factorization is frequently used by recommendation systems, based on an incomplete user-item rating matrix. However, as the number of items and users increase, it becomes difficult to make accurate recommendations due to the sparsity of data. To overcome this drawback, the use of text data related to items was recently suggested for matrix factorization algorithms. Furthermore, a word-level convolutional neural network was shown to be effective in the process of extracting the word-level features from the text data among these kinds of matrix factorization algorithms. However, it involves a large number of parameters to learn in the word-level convolutional neural network. Thus, we propose a matrix factorization algorithm which utilizes a character-level convolutional neural network with which to extract the character-level features from the text data. We also conducted a performance study with real-life datasets to show the effectiveness of the proposed matrix factorization algorithm.

• Journal of KIISE
• /
• v.44 no.8
• /
• pp.774-781
• /
• 2017

Mention detection systems use nouns or noun phrases as a head and construct a chunk of text that defines any meaning, including a modifier. The term "mention detection" relates to the extraction of mentions in a document. In the mentions, a coreference resolution pertains to finding out if various mentions have the same meaning to each other. A pointer network is a model based on a recurrent neural network (RNN) encoder-decoder, and outputs a list of elements that correspond to input sequence. In this paper, we propose the use of mention detection using pointer networks. Our proposed model can solve the problem of overlapped mention detection, an issue that could not be solved by sequence labeling when applying the pointer network to the mention detection. As a result of this experiment, performance of the proposed mention detection model showed an F1 of 80.07%, a 7.65%p higher than rule-based mention detection; a co-reference resolution performance using this mention detection model showed a CoNLL F1 of 52.67% (mention boundary), and a CoNLL F1 of 60.11% (head boundary) that is high, 7.68%p, or 1.5%p more than coreference resolution using rule-based mention detection.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.11 no.11
• /
• pp.1083-1092
• /
• 2016

In this paper, the performance of the optical code-division multiple access (OCDMA) technology on a plastic optical fiber (POF) access network, which had received much attention due to its low weight, large core diameter, flexibility, easy installation, and especially its high bandwidth, is analyzed. Recently, POF was a very attractive candidate for transmission media in an access network based on OCDMA technology. But the conventional OCDMA system only allows finite units to transmit and access simultaneously according to the number of channels which are restricted by BER, and so, in this paper, to resolve this problem a novel multi-priority reservation protocol is also proposed. By using this reservation scheme and a distributed arbitration algorithm, channel collision and destination conflict could be avoided. And this protocol can efficiently support the transmission of multimedia messages that require the different time-delay. The network throughput and average delay using various system parameters have been investigated by numerical analysis and simulation experiments. These results shows that the multi-priority reservation protocol in this POF access network based on OCDMA technology is valid and efficient.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.11 no.6
• /
• pp.73-80
• /
• 2011

In this paper, we design and implementation of effective remote management system for VoIP(Voice over IP) terminal. Ordinary VoIP terminal is connected inside of NAT(Network Address Translation) assigned private IP address. NAT provides address mapping between outside public IP address and inside private IP address, for outside system telephone call is connected to inside VoIP terminal. In this paper, we also design and implementation of UDP hole punching function that the outside calls pass through the NAT, for outside management system call is connected to internal VoIP terminal, and gathered to management information of the VoIP terminal. Also, we evaluated and analysed of developed system in the test environment. As a result of test, It showed that it was well performed without any data error and data loss in the connectivity, and It showed that it was well gathered management information of the VoIP terminal.

• Journal of Digital Contents Society
• /
• v.18 no.8
• /
• pp.1509-1516
• /
• 2017

In first-person perspective training system, the users needs realistic experience. For providing this experience, the system should offer the users virtual and real images at the same time. We propose an automatic a persons's arm segmentation and image composition method. It consists of arm segmentation part and image composition part. Arm segmentation uses an arbitrary image as input and outputs arm segment or alpha matte. It enables end-to-end learning because we make use of FCN in this part. Image composition part conducts image combination between the result of arm segmentation and other image like road, building, etc. To train the network in arm segmentation, we used arm images through dividing the videos that we took ourselves for the training data.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2014.04a
• /
• pp.967-970
• /
• 2014

인터넷 게시판이나 트위터 같은 온라인 매체는 쉬운 접근성과 실시간 특성으로 어떤 사건에 대한 사용자들의 반응이 즉각적으로 나타난다. 또한, 실시간으로 엄청난 양의 데이터가 생성되고 있어 이 데이터를 잘 분석한다면 실제 사회에서 나타나는 다양한 현상들에 대해 파악할 수 있다. 최근 주식 시장에서도 이러한 온라인 데이터들을 분석하여 주가 변동이나 주식 시장 상황을 이해하려는 연구가 시도되고 있다. 이 논문에서는 주식 토론방의 게시물과 주가 사이에 어떤 상관관계가 있는지를 분석하고, 이를 이용한 주식 투자 종목 추천 시스템을 제안하고자 한다. 먼저 주가와 주식 토론방 게시물들 사이의 상관관계를 분석하기 위해서 KOSPI200에 속한 회사 중 55개의 회사를 대상으로 주가와 주식 토론방 게시물을 분석하였다. 2008년부터 2013년까지 6년 동안 각 회사의 주가와 게시물의 상관관계를 분석한 결과 개별 주가와 게시물 수 사이에는 특별한 상관관계가 나타나지 않았다. 하지만 주가와 게시물 수의 상관관계가 높을수록 주식 수익률이 높은 경향을 보였다. 이 논문에서는 주가와 게시물 수의 상관관계 정보를 이용한 투자 종목 추천 알고리즘을 제안하였고, 모의투자 실험을 통해 제안 방법의 효율성을 보였다. 2008년 1월부터 2013년 12월까지의 주가와 주식 토론방 데이터를 이용한 모의투자 실험에서 제안 방법으로 구성한 포트폴리오의 1개월 평균 수익률은 약 1.82%로, 주식 네트워크 특성을 이용한 기존 방법보다 약 0.64% 높은 수익률을 보였다. 또한, 마코위츠의 효율적 포트폴리오와 KOSPI200 수익률보다 각각 약 0.85%와 1.48% 높게 나타났다.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.37 no.4
• /
• pp.569-579
• /
• 2013

Robot control and image processing using a smartphone and Wi-Fi communication is introduced. The robot has a wheel and quadruped mechanism that is transformed according to the environment and is mainly used for drug bottle delivery. The captured image on the camera is transmitted to the smartphone in the form of stream data, and the image data is processed in the smartphone to enable the robot to identify an object and to control the robot itself. A network was constructed so that only image data from the stream data was used, and an image processing scheme to identify the drug bottle and deliver it to a person using a robot arm is also presented. In this study, image processing techniques and algorithms were purely implemented on a smartphone with considerable computational power and multiple functions rather than a computer, which contributes to the intelligence and miniaturization of the robot system.

• Journal of the Korea Computer Industry Society
• /
• v.5 no.3
• /
• pp.393-404
• /
• 2004

Recently, public home have used a more than two computer connected with network, and several home appliances using independently with internet or network are developing to be related closely with the network. Therefore, the home utilized for a simple terminal of the global network in the past is being expanded to another part of the sub network. For a variety of connecting home-area protocols with the existing existing network, we require a new Residential Gateway(RG) that does not only make the home-area network operating in the sub network but also connects to the external network. In this paper, RG has intrinsic limits against flexible service due to IP address assignment and hardware capacity. In order to solve this problem in the RG, we propose a Management Server(MS). The MS that offers the integrated managements and control services for a variety of devices connected the RG in the home-area. It can not only solve the dynamic IP address assigning problem but also assigns private IP addresses to the home network devices through the Network Address Translation(NAT). It also provides somewhat useful functions for the home network and the RG for other additional services. <중략> The MS is using a SNMP protocol for managing the RG in the domain, a polling method of the MS and the RG compose a sequence polling method, a polling method using a multi-process and a multi-thread. In this paper, we introduce a problem with polling method separately, show a polling method between the MS and the RG using a multi-thread.

• Journal of Fisheries and Marine Sciences Education
• /
• v.12 no.2
• /
• pp.176-190
• /
• 2000

It is possible to construct the more improved communication network quality due to practical use of data network using the redundant frequency bandwidth of CATV network. And the multimedia remote educations under the CATV network environment are being tried in the secondary schools. In general, CATV network is able to support not only the remote education using multimedia contents but also real-time responses because the network of CATV has capability to have transmission speed from 256Kbps to l0Mbps. In this paper, we design a new model of the efficient remote education by analysis of the multimedia data transmission capability using CATV network and suggest a method which can be applied specifically.