• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.4
• /
• pp.145-152
• /
• 2020

MRC (Machine reading comprehension) is the AI NLP task that predict the answer for user's query by understanding of the relevant document and which can be used in automated consult services such as chatbots. Recently, the BERT (Pre-training of Deep Bidirectional Transformers for Language Understanding) model, which shows high performance in various fields of natural language processing, have two phases. First phase is Pre-training the big data of each domain. And second phase is fine-tuning the model for solving each NLP tasks as a prediction. In this paper, we have made the Patent MRC dataset and shown that how to build the patent consultation training data for MRC task. And we propose the method to improve the performance of the MRC task using the Pre-trained Patent-BERT model by the patent consultation corpus and the language processing algorithm suitable for the machine learning of the patent counseling data. As a result of experiment, we show that the performance of the method proposed in this paper is improved to answer the patent counseling query.

• Journal of Broadcast Engineering
• /
• v.18 no.5
• /
• pp.680-690
• /
• 2013

In this paper, we propose a real-time temporal synchronization and compensation algorithm in stereoscopic video. Many temporal asynchronies are caused in the video editing stage and due to different transmission delays. These temporal asynchronies can degrade the perceived 3D quality. The goal of temporal alignment is to detect and to measure the temporal asynchrony and recover synchronization of the two video streams. In order to recover synchronization of the two video streams, we developed a method to detect asynchronies between the left and the right video streams based on a novel spatiogram information, which is a richer representation, capturing not only the values of the pixels but their spatial relationships as well. The proposed novel spatiogram additionally includes the changes of the spatial color distribution. Furthermore, we propose a block-based method for detection of the pair frame instead of one frame-based method. Various 3D experiments demonstrate the effectiveness of the proposed method.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.11A
• /
• pp.1092-1097
• /
• 2005

In this paper, we propose a low-power Booth multiplication which reduces the switching activities of partial products during multiplication process. Radix-4 Booth algorithm has a characteristic that produces the Booth encoded products with zero when input data have sequentially equal values (0 or 1). Therefore, partial products have higher chances of being zero when an input with a smaller effective dynamic range of two multiplication inputs is used as a multiplier data instead of a multiplicand. The proposed multiplier divides a multiplication expression into several multiplication expressions with smaller bits than those of an original input data, and each multiplication is computed independently for the Booth encoding. Finally, the results of each multiplication are added. This means that the proposed multiplier has a higher chance to have zero encoded products so that we can implement a low power multiplier with the smaller switching activity. Implementation results show the proposed multiplier can save maximally about $20\%$ power dissipation than a previous Booth multiplier.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.1A
• /
• pp.142-155
• /
• 2000

In this paper, we propose an efficient policing mechanism for Markov model-based traffic descriptors of VBR MPEG video traffic. A VBR video sequence is described by a set of traffic descriptors using a scene-basedMarkov model to the network for the effective resource allocation and accurate QoS prediction. The networkmonitors the input traffic from the source using a proposed new policing method. for policing the steady statetransition probability of scene states, we define two representative monitoring parameters (mean holding andrecurrence time) for each state. For frame level cell rate policing of each scene state, accumulated average cellrates for the frame types are compared with the model parameters. We propose an exponential bounding functionto accommodate dynanic behaviors during the transient period. Our simulation results show that the proposedpolicing mechanism for Markovian traffic descriptors monitors the sophisticated traffic such as MPEG videoeffectively and well protects network resources from the nalicious or misbehaved traffic.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.28 no.5B
• /
• pp.435-447
• /
• 2003

Current existing network dimensioning method simply relies on long-time measurement data or uses average traffic characteristics of users. In this paper, we propose an efficient network dimensioning method for DiffServ over MPLS networks. First, User's SLA information is distributed from edge nodes to core nodes according to the proposed algorithm. Then at each link, class-based capacity planning is performed. For capacity planning, we proposed sets of network design parameters for DiffServ classes and bandwidth allocation schemes that are most suitable for each class. We have developed a DiffServ over MPLS network design tool using the proposed method. Simulation results show that our proposed method is able to design efficiently the required class link resources.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.36 no.9A
• /
• pp.758-766
• /
• 2011

Cooperative communications are adopted as a promising solution to achieve high data rate over large areas in the future 40 wireless system and the relay station (RS) is the key concept in cooperative communications. However, most existing works in this area focus only on optimal RS selections. In addition, there are only few works consider another crucial issue: how many relay stations we need to place. Only when the number of relay stations is defined, the relay station selection can be performed well. In this paper we derive a formula which describes the impact of varying number of RS on end-to-end link throughput assuming a clustering scheme which is based on Voronoi tessellation. In addition to mathematical analysis on the feasibility of the formula, we also examine its performance through a set of simulations under the Erceg path loss model. Simulation results verify that the link throughput gain of our proposed scheme is promising.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.42 no.1
• /
• pp.50-63
• /
• 2017

In the wireless sensor networks, a sensor node has limited resources such as battery or memory. Hence, the routing protocol used in the wireless sensor networks is required efficient use of resources. In this paper, the VAP(Virtual Address Protocol), ensures the shortest transmission path for data transmission between the sink node and sensor node, is proposed. The VAP is assigned each node virtual address and sub address related with its neighboring nodes. The shortest transmission path to the destination node will be choose by comparing to each other address of its neighboring node. The comparison with other existing routing protocol shows that transmission path and usage of memory resource are minimized. Also transmission delay time of data was reduced.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.33 no.4
• /
• pp.1-6
• /
• 2005

The ability to predict radar return from flying vehicles becomes a critical technology issue in the development of stealth configurations. Toward developing a CEM code based on Maxwell's equations for analysis of RCS reduction schemes, an explicit upwind scheme suitable for multidisciplinary design is presented. The DFFT algorithm is utilized to convert the time-domain field values to the frequency-domain. A Green's function based on near field-to-far field transformation is also employed to calculate the bistatic RCS. To verify the numerical calculation the two-dimensional field around a perfectly conducting cylinder is considered. Finally results are obtained for the scattering electromagnetic field around an airfoil in order to illustrate the feasibility of applying CFD based methods to CEM.

• Annual Conference on Human and Language Technology
• /
• 1999.10e
• /
• pp.105-109
• /
• 1999

영-한 기계번역 중 변환 단계에서 한국어 문장을 생성하기 위해서는 구구조 변환 후 조사 및 대역어 선택으로 이루어진다. 그러나 하나의 영어 단어는 여러 개의 한국어 의미들을 가지고 있기 때문에 문장에서 사용된 영어의 정확한 의미에 해당하는 한국어 대역어를 선택하는 것은 번역의 질을 높이고 시스템의 성능에 매우 중요한 역할을 한다. 특히 용언 및 체언의 대역어 선택은 문장에서 서로 간의 의미적인 관계를 고려하여야 올바른 대역어를 선택할 수 있다. 기존에는 전자 사전에 용언과 체언간의 연어 정보(collocation information)를 구축하여 대역어 선택의 문제를 해결하려고 하였으나 연어 정보가 사전에 존재하지 않을 때 올바른 대역어를 선택할 수 없었다. 또한 용언과 체언의 관계를 나타내는 조사를 선택하기 위하여 격(case)을 세분화하여 사전을 구축하였으나 격의 분류 및 사전을 구축할 경우 격을 선택하는 어려움이 있었다. 이에 따라 본 논문에서는 문형(sentence pattern)에 의한 방법으로 용언의 대역어 및 용언이 갖는 필수격 체언의 조사와 대역어 선택방법을 제안한다. 문형의 구조적인 정보에는 용언과 체언의 의미적 역할(thematic role)을 하는 조사 및 용언이 갖는 필수격 체언의 의미 자질(semantic feature)을 갖고 있다. 이러한 의미 자질을 wordnet과 한/영 및 영/한 사전을 이용하여 의미 지표(semantic marker)를 갖는 문형 사전을 구축한다. 또한 의미 지표를 갖는 문형 사전을 기반으로 조사 및 대역어 선택 알고리즘을 개발한다.

• Computational Structural Engineering
• /
• v.11 no.1
• /
• pp.205-216
• /
• 1998

A solution method is presented to solve the eigenvalue problem arising in the dynamic analysis of nonclassicary damped structural systems with multiple eigenvalues. The proposed method is obtained by applying the modified Newton-Raphson technique and the orthonormal condition of the eigenvectors to the linear eigenproblem through matrix augmentation of the quadratic eigenvalue problem. In the iteration methods such as the inverse iteration method and the subspace iteration method, singularity may be occurred during the factorizing process when the shift value is close to an eigenvalue of the system. However, even though the shift value is an eigenvalue of the system, the proposed method provides nonsingularity, and that is analytically proved. Since the modified Newton-Raphson technique is adopted to the proposed method, initial values are need. Because the Lanczos method effectively produces better initial values than other methods, the results of the Lanczos method are taken as the initial values of the proposed method. Two numerical examples are presented to demonstrate the effectiveness of the proposed method and the results are compared with those of the well-known subspace iteration method and the Lanczos method.