• ETRI Journal
• /
• v.46 no.1
• /
• pp.59-70
• /
• 2024

We introduce a high-performance named entity recognition (NER) model for written and spoken language. To overcome challenges related to labeled data scarcity and domain shifts, we use transfer learning to leverage our previously developed KorBERT as the base model. We also adopt a meta-pseudo-label method using a teacher/student framework with labeled and unlabeled data. Our model presents two modifications. First, the student model is updated with an average loss from both human- and pseudo-labeled data. Second, the influence of noisy pseudo-labeled data is mitigated by considering feedback scores and updating the teacher model only when below a threshold (0.0005). We achieve the target NER performance in the spoken language domain and improve that in the written language domain by proposing a straightforward rollback method that reverts to the best model based on scarce human-labeled data. Further improvement is achieved by adjusting the label vector weights in the named entity dictionary.

• Interaction and multiscale mechanics
• /
• v.2 no.4
• /
• pp.353-374
• /
• 2009

A new seamless multiscale simulation was developed for coupling the continuum model with its molecular dynamics. Kriging-based Finite Element Method (K-FEM) is employed to model the continuum base of the entire domain, while the molecular dynamics (MD) is confined in a localized domain of interest. In the coupling zone, where the MD domain overlaps the continuum model, the overall Hamiltonian is postulated by contributions from the continuum and the molecular overlays, based on a quartic spline scaling parameter. The displacement compatibility in this coupling zone is then enforced by the Lagrange multiplier technique. A multiple-time-step velocity Verlet algorithm is adopted for its time integration. The validation of the present method is reported through numerical tests of one dimensional atomic lattice. The results reveal that at the continuum/MD interface, the commonly reported spurious waves in the literature are effectively eliminated in this study. In addition, the smoothness of the transition from MD to the continuum can be significantly improved by either increasing the size of the coupling zone or expanding the nodal domain of influence associated with K-FEM.

• Bulletin of the Korean Mathematical Society
• /
• v.49 no.3
• /
• pp.517-528
• /
• 2012

This paper is estimating the domain of attraction for a class of susceptible-exposed-infectious-recovered (SEIR) epidemic dynamic models by using sum of squares optimization. First, the stability is analyzed for the equilibriums of SEIR model, and the domain of attraction in the endemic equilibrium is estimated by using sum of squares optimization. Finally, a numerical example is examined.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 1996.04a
• /
• pp.62-69
• /
• 1996

In this study, an equivalent domain integral (EDI) method is presented to estimate the track-till integral parameter, J-value, for two dimensional cracked elastic bodies which may quantify the severity of the crack-tit) stress fields. The conventional J-integral method based on line integral has been converted to equivalent area or domain integrals by using the divergence theorem. It is noted that the EDI method is very attractive because all the quantities necessary for computation of the domain integrals are readily available in a finite element analysis. The details and its implementation are extened to both h-version finite element model with 8-node isoparametric element and p-version finite element model with high order hierarchic element using Legendre type shape fuctions. The variations with respect to the different path of domain integrals from the crack-tip front and the choice of 5-function have been tested by several examples.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.24 no.7
• /
• pp.730-736
• /
• 2013

Time-domain clutter generation model for airborne pulse doppler phase-array radar is presented. Time-domain surface clutter signal is generated assuming earth of a sphere and considering geometry of a clutter patch, and generation of sub-array clutter signal is presented. The generated sub-array clutter signal can be used by simulation input signal in various radar applications of DBF(Digital Beamforming), ABF(Adaptive Beamforming), Stap(Space-Time Adaptive Processing) and etc.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2015.05a
• /
• pp.253-253
• /
• 2015

This research evaluated the effect of two main sources on the results of the ocean regional circulation model (ORCMs) during downscaling and nesting the results from the coarse data. The two sources should be the domain size, and temporal and spatial resolution different between driving and driven data. The Big-Brother Experiment is applied to examine the impact of them on the results of the ORCMs separately. Within resolution of 3km grid point ORCMs applying in the Big-Brother Experiment framework, it showed that the simulation results of the ORCMs depend on the domain size and specially the spatial and temporal resolution of lateral boundary conditions (LBCs). The domain size can be selected at 9.5 times larger than the interest area, and the spatial resolution between driving data and driven model can be up to 3 of ratio resolution and updating frequency of the LBCs can be up to every 6 hours per day.

• Journal of Information Technology Applications and Management
• /
• v.28 no.5
• /
• pp.41-50
• /
• 2021

As COVID-19 spread widely, and rapidly, the number of misinformation is also increasing, which WHO has referred to this phenomenon as "Infodemic". The purpose of this research is to develop detection and rectification of COVID-19 misinformation based on Open-domain QA system and text similarity. 9 testing conditions were used in this model. For open-domain QA system, 6 conditions were applied using three different types of dataset types, scientific, social media, and news, both datasets, and two different methods of choosing the answer, choosing the top answer generated from the QA system and voting from the top three answers generated from QA system. The other 3 conditions were the Closed-Domain QA system with different dataset types. The best results from the testing model were 76% using all datasets with voting from the top 3 answers outperforming by 16% from the closed-domain model.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.9 no.3 s.43
• /
• pp.33-42
• /
• 2005

Frequency domain Volterra model is applied to nonlinear parameter identification procedure for dynamic systems modeled by nonlinear function. The frequency domain Volterra kernels, which correspond io linear, quadratic, and cubic transfer functions in lime domain, are incorporated in nonlinear parametric identification procedure. The nonlinear transfer functions, which can be derived from the Volterra series representation of the nonlinear differential equation of the system by Schetzen's method(1980), are directly used for modeling input output relation. The error is defined by the difference between the observed output and the estimated output which is calculated by substituting the observed input to nonlinear frequency domain model. The system parameters are searched by minimizing the error. Volterra model guarantees enough accuracy and convergence and the estimated coefficients have a good agreement with their actual values not only in the linear frequency region but also in the legion where the $2^{nd}\;or\;3^{rd}$ order nonlinearity is dominant.

• Journal of Digital Convergence
• /
• v.14 no.1
• /
• pp.453-463
• /
• 2016

This study examined the relationship among dropout, organizational trust, and intention to transfer in the department affiliated with Physical Education. Data were collected from 480 students in the department affiliated with Physical Education from 6 institutions of a 4-year university located in Seoul, Gyeonggi-do, Incheon, and ChungChong-do. For the research model analysis, a structural equation model and bootstrapping were conducted. Results indicated that the academic domain, the social domain, and the environmental domain of dropout had a positive effect on organizational trust. But the only the academic domain in the relationship between dropout and intention to transfer affected intention to transfer positively. The social domain and the environmental domain were not statistically significant on intention to transfer. Moreover, organizational trust had a positive effect on intention to transfer. Finally, organizational trust partially mediated the path between the academic domain and intention to transfer. It fully mediated the paths between the social domain and the environmental domain and intention to transfer.

• Convergence Security Journal
• /
• v.17 no.3
• /
• pp.51-57
• /
• 2017

VOD service provides an additional revenue for broadcasting companies in addition to the existing subscription fees and advertisement-based revenue. Therefore, each broadcasting company develops its own VOD service and performs frequent improvement work. This leads to the development of new VOD services, so developers are considering ways to effectively handle the frequent development needs. In this background, we conducted an underlying research to apply the feature-oriented analysis model to the development of VOD service. The feature-oriented analysis model used in this study is the Feature-Oriented Domain Analysis (FODA) developed by SEI of Carnegie Mellon University. FODA provides a tool for specifying a feature model of a software domain, based on which developers determine the configuration of a software with customers. This study developed a feature model of the VOD service domain and devised the functionalities and testcases in an integrated manner with the feature model. Additionally, we proposed a VOD service development process utilizing the feature model, function specification, and testcases.