• 전자통신동향분석
• /
• 제35권3호
• /
• pp.9-19
• /
• 2020

Recently, a technique for applying a deep learning language model pretrained from a large corpus to fine-tuning for each application task has been widely used as a language processing technology. The pretrained language model shows higher performance and satisfactory generalization performance than existing methods. This paper introduces the major research trends related to deep learning pretrained language models in the field of language processing. We describe in detail the motivations, models, learning methods, and results of the BERT language model that had significant influence on subsequent studies. Subsequently, we introduce the results of language model studies after BERT, focusing on SpanBERT, RoBERTa, ALBERT, BART, and ELECTRA. Finally, we introduce the KorBERT pretrained language model, which shows satisfactory performance in Korean language. In addition, we introduce techniques on how to apply the pretrained language model to Korean (agglutinative) language, which consists of a combination of content and functional morphemes, unlike English (refractive) language whose endings change depending on the application.

• 대한산업공학회지
• /
• 제31권4호
• /
• pp.334-340
• /
• 2005

A mouse is an important input device that is used in most of all computer works. A mouse control time prediction model was proposed in this study. Especially, the model described the time of mouse control that made a cursor to move within path constraints. The model was developed by a laboratory experiment. Cursor movement times were measured in 36 task conditions; 3 levels of path length, 3 levels of path width and 4 levels of target's width. 12 subjects participated in all conditions. The time of cursor movement with path constraints could be better explained by the combination of Fitts' law with steering law($r^2=0.947$) than by the other models; Fitts' law($r^2=0.740$), Steering law($r^2=0.633$) and Crossman's model($r^2=0.897$). The proposed model is expected to be used in menu design or computer game design.

• Communications for Statistical Applications and Methods
• /
• 제22권3호
• /
• pp.305-312
• /
• 2015

We extend a generalized partially linear single-index model and newly define a generalized partially double-index model (GPDIM). The philosophy of sufficient dimension reduction is adopted in GPDIM to estimate unknown coefficient vectors in the model. Subsequently, various combinations of popular sufficient dimension reduction methods are constructed with the best combination among many candidates determined through a bootstrapping procedure that measures distances between subspaces. Distinguishing values are newly defined to match the estimates to the corresponding population coefficient vectors. One of the strengths of the proposed model is that it can investigate the appropriateness of GPDIM over a single-index model. Various numerical studies confirm the proposed approach, and real data application are presented for illustration purposes.

• Asian-Australasian Journal of Animal Sciences
• /
• 제19권10호
• /
• pp.1490-1495
• /
• 2006

The purpose of this research was to develop an optimum composition model for a new synbiotic fermented dairy product with high probiotic cell counts, and to experimentally verify this model. The optimum composition model indicated the growth promoter ratio that could provide the highest growth rate for probiotics in this fermented product. Different levels of growth promoters were first blended with milk to improve the growth rates of probiotics, and the optimum composition model was determined. The probiotic viabilities and chemical properties were analyzed for the samples made using the optimal formula. The optimal combination of the growth promoters for the synbiotic fermented milk product was 1.12% peptides, 3% fructooligosaccharides (FOS), and 1.87% isomaltooligosaccharides (IMO). A product manufactured according to the formula of the optimum model was analyzed, showing that the model was effective in improving the viability of both Lactobacillus spp. and Bifidobacterium spp.

• Journal of Advanced Marine Engineering and Technology
• /
• 제38권5호
• /
• pp.550-556
• /
• 2014

In this study, for the control of a single water tank system, a fuzzy-PID controller design technique based on a fuzzy model is investigated. For this purpose, a water tank system is linearized as a number of submodels depending on the operating point, and a fuzzy model is obtained by fuzzy combining. Each submodel is approximated as a first order time delay model, and a PID controller is designed using several existing tuning techniques. Then, through the fuzzy combination of this controller using the same method as that of the fuzzy model, a fuzzy-PID controller is designed. For the proposed technique, a simulation is performed using the fuzzy model of a water tank system, and the validity is examined by comparing its performance with that of a PID controller.

• 음성과학
• /
• 제11권2호
• /
• pp.247-257
• /
• 2004

There has been many research efforts to overcome the problems of speech recognition in the noisy conditions. Among them, the model-based compensation methods such as the parallel model combination (PMC) and vector Taylor series (VTS) have been found to perform efficiently compared with the previous speech enhancement methods or the feature-based approaches. In this paper, a data-driven model compensation approach that adapts the HMM(hidden Markv model) parameters for the noisy speech recognition is proposed. Instead of assuming some statistical approximations as in the conventional model-based methods such as the PMC, the statistics necessary for the HMM parameter adaptation is directly estimated by using the Baum-Welch algorithm. The proposed method has shown improved results compared with the PMC for the noisy speech recognition.

• 소성∙가공
• /
• 제13권2호
• /
• pp.168-173
• /
• 2004

Microforming can be a good application for bulk metallic glasses. It is important to simulate the deformation behaviour of the bulk metallic glasses in a supercooled liquid region for manufacturing micromachine parts. For these purposes, a correct constitutive model which can reproduce viscosity results is essential for good predicting capability. In this paper, we studied deformation behaviour of the bulk metallic glasses using the finite element method in conjunction with the fictive stress constitutive model which can describe non-Newtonian as well as Newtonian behaviour. A combination of kinetic equation which describes the mechanical response of the bulk metallic glasses at a given temperature and evolution equations fur internal variables provide the constitutive equation of the fictive stress model. The internal variables are associated with fictive stress and relation time. The model has a modular structure and can be adjusted to describe a particular type of microforming process. Implementation of the model into the MARC software has shown its versatility and good predictive capability.

• 음성과학
• /
• 제14권2호
• /
• pp.115-126
• /
• 2007

There have been many research efforts to improve the performance of the speech recognizer in noisy conditions. Among them, the model compensation method and the speech enhancement approach have been used widely. In this paper, we propose to combine the two different approaches to further enhance the recognition rates in the noisy speech recognition. For the speech enhancement, the minimum mean square error-short time spectral amplitude (MMSE-STSA) has been adopted and the parallel model combination (PMC) and Jacobian adaptation (JA) have been used as the model compensation approaches. From the experimental results, we could find that the hybrid approach that applies the model compensation methods to the enhanced speech produce better results than just using only one of the two approaches.

• 한국CDE학회논문집
• /
• 제11권6호
• /
• pp.429-439
• /
• 2006

The purpose of this study is to support the integration from current 2D drawing-based design to future 3D model-based design. In this paper, an important theme is the combination between the STEP-based 2D drawing standards (i.e., AP202) and the IFC-based 3D building model standards. To achieve the purpose, two methodologies are proposed as follows: the development of IFC extension model for the 2D shape data representation by harmonizing ISO/STEP AP202; and the development of mapping solution between IFC 2D extension model and KOSDIC by constructing the exchange scenario for 2D shape data representation. It is expected that the proposed IFC2X2 2D extension model and mapping solution will offer the basis of development of the integrated standards model in AEC industry.

• 한국시스템다이내믹스연구
• /
• 제6권2호
• /
• pp.5-23
• /
• 2005

This paper investigates a dynamic mechanism underlying the process of knowledge creation and evolution with a focus on the SECI model(standing for Socialization, Externalization, Combination, Internalization) as proposed by Nonaka and Takeuchi(1991) and broadly accepted especially among the practitioners in knowledge management field. The SECI model provides with intuitive logic and clear delineation of knowledge types between the tacit and the explicit, and embodies an interaction dynamic. However explanations of the propelling forces for the knowledge transfer over the four quadrants of the model is yet to be made. And the transmission mechanisms are not prescribed though the model mentions knowledge is created and evolved in a spiral process. This paper, therefore attempts first to extend and elaborate it into a dynamic SECI model by identifying those propelling factors and their relationships(linkages) based on the systems thinking.