• ETRI Journal
• /
• v.46 no.1
• /
• pp.127-136
• /
• 2024

The Korean language has written (formal) and spoken (phonetic) forms that differ in their application, which can lead to confusion, especially when dealing with numbers and embedded Western words and phrases. This fact makes it difficult to automate Korean speech recognition models due to the need for a complete transcription training dataset. Because such datasets are frequently constructed using broadcast audio and their accompanying transcriptions, they do not follow a discrete rule-based matching pattern. Furthermore, these mismatches are exacerbated over time due to changing tacit policies. To mitigate this problem, we introduce a data-driven Korean spoken-to-written transcription conversion technique that enhances the automatic conversion of numbers and Western phrases to improve automatic translation model performance.

• Journal of Powder Materials
• /
• v.31 no.3
• /
• pp.213-219
• /
• 2024

Additive Manufacturing (AM) is a process that fabricates products by manufacturing materials according to a three-dimensional model. It has recently gained attention due to its environmental advantages, including reduced energy consumption and high material utilization rates. However, controlling defects such as melting issues and residual stress, which can occur during metal additive manufacturing, poses a challenge. The trial-and-error verification of these defects is both time-consuming and costly. Consequently, efforts have been made to develop phenomenological models that understand the influence of process variables on defects, and mechanical/ electrical/thermal properties of geometrically complex products. This paper introduces modeling techniques that can simulate the powder additive manufacturing process. The focus is on representative metal additive manufacturing processes such as Powder Bed Fusion (PBF), Direct Energy Deposition (DED), and Binder Jetting (BJ) method. To calculate thermal-stress history and the resulting deformations, modeling techniques based on Finite Element Method (FEM) are generally utilized. For simulating the movements and packing behavior of powders during powder classification, modeling techniques based on Discrete Element Method (DEM) are employed. Additionally, to simulate sintering and microstructural changes, techniques such as Monte Carlo (MC), Molecular Dynamics (MD), and Phase Field Modeling (PFM) are predominantly used.

• Journal of Information Technology Services
• /
• v.23 no.4
• /
• pp.55-70
• /
• 2024

In order to promote the AI applications in the financial industry, the financial sector has recently been paying attention to synthetic data technology. Synthetic data generates using a purpose-built mathematical model or algorithm, with the aim of solving a set of data science tasks. This study evaluates the utility of synthetic data by analyzing heterogeneous tabular data that is composed of discrete, categorical and continuous variables and has the feature of unbalanced data, which is commonly found in the financial sector. As a synthetic data generation technique, the TGAN and CTGAN models are applied by considering the feature of tabular data. As a result of evaluating the utility in terms of resemblance and machine learning efficiency, those of TGAN are confirmed to be high, while the quality of CTGAN are relatively poor. This is interpreted to be particularly due to the generation of categorical variables, and it suggests that how those with categorical properties especially are considered in the synthetic data generation model is a major factor in determining the utility of generation synthetic data.

• Water for future
• /
• v.29 no.4
• /
• pp.209-221
• /
• 1996

It may be difficult to make exact estimates of peak discharge or runoff depth of a flood and to establish the proper measurement for the flood protection since water stages or discharges have been rarely measured at small river basins in Korea. Three small catchments in the Su-Young river basin in Pusan were selected for the study areas. Various runoff parameters for the study areas were determined, and runoff analyses were performed using three different runoff models available in literatures; the storage function method, the discrete, linear, input-output model, and the linear reservoir model. The hydrographs calculated by three different methods showed good agreement with the observed flood hydrographs, indicating that the models selected are all capable of sucessfully modeling the flood events for small watersheds. The storage function method gave the best results in spite of its weakness that it could not be applicable to small floods, while the linear reservoir model was found to provide relatively good results with less parameters. The capabilities of simulating flood hydrographs were also evaluated based on the effective rainfall from the storage function parameters, the $\Phi$-index method, and the constant percentage method. For the On-Cheon stream watershed, the storage function parameters provided better estimates of effective rainfall for regenerating flood hydrographs than any others considered in the study. The $\Phi$-index method, however, resulted in better estimates of effective rainfall for the other two study areas.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.10
• /
• pp.723-730
• /
• 2013

This paper presents the design of an arm gesture recognition system using Kinect sensor. A variety of methods have been proposed for gesture recognition, ranging from the use of Dynamic Time Warping(DTW) to Hidden Markov Models(HMM). Our system learns a unique HMM corresponding to each arm gesture from a set of sequential skeleton data. Whenever the same gesture is performed, the trajectory of each joint captured by Kinect sensor may much differ from the previous, depending on the length and/or the orientation of the subject's arm. In order to obtain the robust performance independent of these conditions, the proposed system executes the feature transformation, in which the feature vectors of joint positions are transformed into those of angles between joints. To improve the computational efficiency for learning and using HMMs, our system also performs the k-means clustering to get one-dimensional integer sequences as inputs for discrete HMMs from high-dimensional real-number observation vectors. The dimension reduction and discretization can help our system use HMMs efficiently to recognize gestures in real-time environments. Finally, we demonstrate the recognition performance of our system through some experiments using two different datasets.

• Journal of Korean Society of Transportation
• /
• v.24 no.6 s.92
• /
• pp.89-102
• /
• 2006

The purpose of this paper is to develop a dynamic OD estimating model to overcome the limitation of depicting teal situations in dynamic simulation models based on static OD trip. To estimate dynamic OD matrix we used the hybrid discrete choice model(called the 'Demand Simulation Model'), which combines travel departure time with travel mode and travel path. Using this Demand Simulation Model, we deduced that the traveler chooses the departure time and mode simultaneously, and then choose his/her travel path over the given situation In this paper. we developed a hybrid simulation model by joining a demand simulation model and the supply simulation model (called LiCROSIM-P) which was Previously developed. We simulated the hybrid simulation model for dependent/independent networks which have two origins and one destination. The simulation results showed that AGtt(Average gap expected travel time and simulated travel time) did not converge, but average schedule delay gap converged to a stable state in transportation network consisted of multiple origins and destinations, multiple paths, freeways and some intersections controlled by signal. We present that the hybrid simulation model can estimate dynamic OD and analyze the effectiveness by changing the attributes or the traveler and networks. Thus, the hybrid simulation model can analyze the effectiveness that reflects changing departure times, travel modes and travel paths by demand management Policy, changing network facilities, traffic information supplies. and so on.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.14 no.4
• /
• pp.293-308
• /
• 2012

The shear behavior at the particle/surface interface such as rock joint can determine the mechanical behavior of whole structure. Therefore, a fundamental understanding of the mechanisms governing its behavior and accurately estimation of the interface strength is essential. In this paper, PFC, a numerical analysis program of discrete element method was used to investigate the effects of the surface roughness on interface strength. The surface roughness was characterized by smooth, intermediate, and rough surface, respectively. In order to investigate the effects of particle shape and crushing on particle/surface interface behavior, one ball, clump, and cluster models were created and their results were compared. The shape of particle was characterized by circle, triangle, square, and rectangle, respectively. The results showed that as the surface roughness increases, interface strength and friction angle increase and the void ratio increases. The one ball model with smooth surface shows lower interface strength and friction angle than the clump model with irregular surface. In addition, a cluster model has less interface strength and friction angle than the clump model. The failure envelope of the cluster model shows non-linear characteristic. From these findings, it is verified that the surface roughness and particle shape effect on the particle/surface interface shear behavior.

• 한국지형공간정보학회:학술대회논문집
• /
• 2004.10a
• /
• pp.5-10
• /
• 2004

The purpose of this research is to measure a gravitational attraction about urban pulbic service facilities and to evaluate the location of public service facilities. It is important in that these facilities should provide more inexpensive and convenience public service to users. To do this, the GIS's spatial analysis and gravity model were used to analyze the efficiency of the public facilities. The gravity model was conducted as the main analysis method, and another model for this analysis was the discrete model. The gravity model is originally to anticipate migration flows, traffic flows and other types of movements so that this model compares the gravitational attraction between places. The discrete model is to find the optimal location and to evaluate a location of facility regarding urban areas as the combination of node and link. In this research, these two models were adopted to compare and analyze location of Mapo-Gu Office. The results of this research indicated that the locational evaluation of urban public service facilities discovered the appropriateness of those facilities, and the public facilities was necessary to displace to other site.

• Journal of Internet Computing and Services
• /
• v.22 no.2
• /
• pp.29-39
• /
• 2021

As damages to individuals, private sectors, and businesses increase due to newly occurring cyber attacks, the underlying network security problem has emerged as a major problem in computer systems. Therefore, NIDS using machine learning and deep learning is being studied to improve the limitations that occur in the existing Network Intrusion Detection System. In this study, a deep learning-based NIDS model study is conducted using the Convolution Neural Network (CNN) algorithm. For the image classification-based CNN algorithm learning, a discrete algorithm for continuity variables was added in the preprocessing stage used previously, and the predicted variables were expressed in a linear relationship and converted into easy-to-interpret data. Finally, the network packet processed through the above process is mapped to a square matrix structure and converted into a pixel image. For the performance evaluation of the proposed model, NSL-KDD, a representative network packet data, was used, and accuracy, precision, recall, and f1-score were used as performance indicators. As a result of the experiment, the proposed model showed the highest performance with an accuracy of 85%, and the harmonic mean (F1-Score) of the R2L class with a small number of training samples was 71%, showing very good performance compared to other models.

• Knowledge Management Research
• /
• v.22 no.4
• /
• pp.27-40
• /
• 2021

Click-Through Rate (CTR) prediction is a key function that determines the ranking of candidate items in the recommendation system and recommends high-ranking items to reduce customer information overload and achieve profit maximization through sales promotion. The fields of natural language processing and image classification are achieving remarkable growth through the use of deep neural networks. Recently, a transformer model based on an attention mechanism, differentiated from the mainstream models in the fields of natural language processing and image classification, has been proposed to achieve state-of-the-art in this field. In this study, we present a method for improving the performance of a transformer model for CTR prediction. In order to analyze the effect of discrete and categorical CTR data characteristics different from natural language and image data on performance, experiments on embedding regularization and transformer normalization are performed. According to the experimental results, it was confirmed that the prediction performance of the transformer was significantly improved when the L2 generalization was applied in the embedding process for CTR data input processing and when batch normalization was applied instead of layer normalization, which is the default regularization method, to the transformer model.