• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.188-191
• /
• 2020

본 연구는 COVID-19 질의 응답 태스크를 위한 Poly-encoder 기반의 태스크를 제안하였다. COVID-19 질의 응답 시스템은 사람들에게 최신 정보에 대해 빠르고 신뢰성이 높은 정보를 전달하는 특성을 가져야한다. 검색 기반 질의 응답 시스템은 pairwise 연산을 기반으로 수행되는데, Poly-encoder는 사전 학습된 트랜스포머(transformer)기반의 pairwise 연산 방법론 중 기존 Cross-encoder와 Bi-encoder보다 실사용 및 성능이 뛰어남을 보였다 [1]. 특히, Poly-encoder는 정확도가 높으면서도 빠른 응답속도를 가지며 검색기반의 각종 태스크에서 좋은 성능을 보였다. 따라서 본 연구는 COVID-19를 위한 Poly-encoder기반의 질의 응답 태스크를 위하여 기존 질의 응답 태스크와 페르소나 기반의 질의 응답 태스크로 두 가지 유형의 태스크를 생성하여 모델을 학습하였다. 또한 신뢰성 있는 리소스정보로부터 모델에 최신 정보 반영을 위하여 자동 크롤러를 구축하여 데이터를 수집하였다. 마지막으로 전문가를 통한 데이터셋을 구축하여 질문-응답과 질의어-질문에 대한 모델 검증을 수행하였다.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.23 no.1
• /
• pp.154-162
• /
• 2019

Large scale damage has been globally increased due to natural disasters such as earthquake. Although a variety of studies secured seismic performance of buildings, casualties and economic loss have occurred because of poor security of seismic performance in non-structural components. Structure's location on which non-structural components are installed and characteristics of vibration occurring on each position of structures are varied, so a response spectrum is required for each position of structures. In addition, a response spectrum occurring in a structure is different, depending on the form of it and positions on which it is installed. Therefore, selection of a response spectrum is important, so a definite method for calculating the response spectrum which acts on non-structural components is necessary. A method for choosing a response spectrum is suggested in this paper, and a structural analysis was conducted with the suggested method, by selecting a ground response spectrum and a structural system, which may occur in Korea. Moreover, it helps create a response spectrum necessary for a seismic test of non-structural components, by suggesting the method for deduction it, with a simple formula.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.4
• /
• pp.31-38
• /
• 2024

This study proposes hyper-parameter settings for developing a generative AI-based learning support tool to facilitate programming education in online distance learning. We implemented an experimental tool that can set research hyper-parameters according to three different learning contexts, and evaluated the quality of responses from the generative AI using the tool. The experiment with the default hyper-parameter settings of the generative AI was used as the control group, and the experiment with the research hyper-parameters was used as the experimental group. The experiment results showed no significant difference between the two groups in the "Learning Support" context. However, in other two contexts ("Code Generation" and "Comment Generation"), it showed the average evaluation scores of the experimental group were found to be 11.6% points and 23% points higher than those of the control group respectively. Lastly, this study also observed that when the expected influence of response on learning motivation was presented in the 'system content', responses containing emotional support considering learning emotions were generated.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.399-403
• /
• 2023

본 논문에서는 다중 문서 기반의 대화 시스템을 통한 효율적인 정보 검색과 응답 생성에 중점을 둡니다. 대규모 데이터 집합에서 정확한 문서를 선택하는 데 필요한 검색의 중요성을 강조하며, 현재 검색 방법의 한계와 문제점을 지적합니다. 또한 더 자연스러운 답변을 생성하기 위해 대규모 언어 모델을 사용하게 되면서 fine-tuning 시에 발생하는 제약과 낭비를 모델의 제로샷 생성 능력을 활용하여 개선하려는 방안을 제안하며, 모델의 크기와 자원의 효율성에 대한 고려사항을 논의합니다. 우리의 접근 방식은 대규모 언어 모델을 프롬프트와 함께 다중 문서로 학습 없이 정보를 검색하고 응답을 생성하는 방향으로 접근하여 대화 시스템의 효율성과 유용성을 향상시킬 수 있음을 제시합니다.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.36 no.1
• /
• pp.1-11
• /
• 2016

The domestic and foreign seismic design codes have specified that time history covers design response spectrum when the response spectrum, which calculated from the time history, is smaller than the design response spectrum at five points or less. In order to verify the design codes, time history analysis for a pier was performed by using five artificial time histories conforming design code with various characteristics and its member forces were evaluated according to them. It was confirmed from analysis results that, regardless of the conformity to design code requirement, seismic design using the artificial time histories could not guarantee earthquake resistant design if the response spectrum from them is lower than design response spectrum at the similar period to the natural frequency of structure. Thus, the time history generating method to make its acceleration response spectrum to be greater than design response spectrum at all period was proposed by two modification function in this study. It was also verified whether time histories from the proposed method satisfy the seismic design codes or not.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.34 no.2
• /
• pp.101-111
• /
• 2021

In general, several computational resources are required to perform multiple-loading transient response analyses. In this paper, we present the procedure for multiple-loading transient response analysis using the Krylov subspace model order reduction and Newmark's time integration scheme. We utilized ANSYS MAPDL, Python, and ANSYS ACT to automate the transient response analysis procedure in the ANSYS Workbench environment and studied several engineering numerical examples to demonstrate the feasibility and efficiency of the proposed approach.

• Journal of Digital Contents Society
• /
• v.11 no.3
• /
• pp.375-383
• /
• 2010

In developing question-answering (QA) systems, it is hard to analyze natural language questions syntactically and semantically and to find exact answers to given query questions. In order to avoid these difficulties, we propose a new style of question-answering system that automatically generate natural language queries and can allow to search queries fit for given keywords. The key idea behind generating natural queries is that after significant sentences within text documents are applied to the named entity recognition technique, we can generate a natural query (interrogative sentence) for each named entity (such as person, location, and time). The natural query is divided into two types: simple type and sentence structure type. With the large database of question-answer pairs, the system can easily obtain natural queries and their corresponding answers for given keywords. The most important issue is how to generate meaningful queries which can present unambiguous answers. To this end, we propose two principles to decide which declarative sentences can be the sources of natural queries and a pattern-based method for generating meaningful queries from the selected sentences.

• Journal of the Korean GEO-environmental Society
• /
• v.10 no.7
• /
• pp.111-115
• /
• 2009

In Korea, the probabilistically developed seismic hazard maps are used with deterministically derived seismic site coefficients in developing the design response spectrum of a specific site. Even though the seismic hazard maps and seismic site coefficients are incompatible, the current design code ignores such incompatibility. If the seismic hazard map and seismic coefficients are both developed in identical probabilistic framework, such problems can be solved. Unfortunately, the available method cannot be use to derive "true" probabilistic site coefficients. This study uses the ground motion time histories, which were developed as the result of a new probabilistic seismic hazard analysis in the companion paper, as input motions in performing one-dimensional equivalent linear site response analyses, from which the uniform hazard response spectra are generated. Another important characteristic of the hazard response spectra are that the uncertainties and randomness of the ground properties are accounted for. The uniform hazard spectra are then used to derive probabilistic site coefficients. Comparison of probabilistic and deterministically site coefficients demonstrate that there is a distinct discrepancy between two coefficients.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.125-130
• /
• 2022

최근 자연어처리 연구로 지식 기반 대화에서 대화 내용에 자유로운 주제와 다양한 지식을 포함하는 연구가 활발히 이루어지고 있다. 지식 기반 대화는 대화 내용이 주어질 때 특정 지식 정보를 포함하여 이어질 응답을 생성한다. 이때 대화에 필요한 지식이 검색 가능하여 선택에 제약이 없는 오픈 도메인(Open-domain) 지식 기반 대화가 가능하도록 한다. 오픈 도메인 지식 기반 대화의 성능 향상을 위해서는 대화에 이어지는 자연스러운 답변을 연속적으로 생성하는 응답 생성 모델의 성능 뿐만 아니라, 내용에 어울리는 응답이 생성될 수 있도록 적합한 지식을 선택하는 지식 검색 모델의 성능 향상도 매우 중요하다. 본 논문에서는 오픈 도메인 지식 기반 한국어 대화에서 지식 검색 성능을 높이기 위해 밀집 벡터 기반 검색 방식과 주제어(Keyword) 기반의 검색 방식을 함께 사용하는 것을 제안하였다. 먼저 밀집 벡터 기반의 검색 모델을 학습하고 학습된 모델로부터 고난도 부정(Hard negative) 지식 후보를 생성하고 주제어 기반 검색 방식으로 고난도 부정 지식 후보를 생성하여 각각 밀집 벡터 기반의 검색 모델을 학습하였다. 성능을 측정하기 위해 전체 지식 중에서 하나의 지식을 검색했을 때 정답 지식인 경우를 계산하였고 고난도 부정 지식 후보로 학습한 주제어 기반 검색 모델의 성능이 6.175%로 가장 높은 것을 확인하였다.

• The Journal of the Acoustical Society of Korea
• /
• v.32 no.2
• /
• pp.116-123
• /
• 2013

This paper represents generation of time-varying underwater acoustic channels by performing scattering simulation with time-varying ocean surface and Kirchhoff approximation. In order to estimate the time-varying ocean surface, 1D Pierson-Moskowitz ocean power spectrum and Gaussian correlation function were used. The computed scattering coefficients are applied to the amplitudes of each impulse of BELLHOP simulation result. The scattering coefficients are then compared with measured doppler spectral density of signal components which were scattered from ocean surface and the correlation time used in the Gaussian correlation function was estimated by the comparison. Finally, bit-error-rate and channel correlation simulations were performed with the generated time-varying channel based on passive time-reversal communication scenario.