• Journal of The Korean Association of Information Education
• /
• v.26 no.3
• /
• pp.197-207
• /
• 2022

In this paper, we develop a machine learning based block code generation and recommendation model for the purpose of reducing cognitive load of learners during coding education that learns the learner's block that has been made in the block programming environment using natural processing model and fine-tuning and then generates and recommends the selectable blocks for the next step. To develop the model, the training dataset was produced by pre-processing 50 block codes that were on the popular block programming language web site 'Entry'. Also, after dividing the pre-processed blocks into training dataset, verification dataset and test dataset, we developed a model that generates block codes based on LSTM, Seq2Seq, and GPT-2 model. In the results of the performance evaluation of the developed model, GPT-2 showed a higher performance than the LSTM and Seq2Seq model in the BLEU and ROUGE scores which measure sentence similarity. The data results generated through the GPT-2 model, show that the performance was relatively similar in the BLEU and ROUGE scores except for the case where the number of blocks was 1 or 17.

• The Korean Journal of Applied Statistics
• /
• v.35 no.3
• /
• pp.407-419
• /
• 2022

The semiconductor fabrication process is complex and time-consuming. There are sometimes errors in the process, which results in defective die on the wafer bin map (WBM). We can detect the faulty WBM by finding some patterns caused by dies. When one manually seeks the failure on WBM, it takes a long time due to the enormous number of WBMs. We suggest a two-step approach to discover the probable pattern on the WBMs in this paper. The first step is to separate the normal WBMs from the defective WBMs. We adapt a hierarchical clustering for de-noising, which nicely performs this work by wisely tuning the number of minimum points and the cutting height. Once declared as a faulty WBM, then it moves to the next step. In the second step, we classify the patterns among the defective WBMs. For this purpose, we extract features from the WBM. Then machine learning algorithm classifies the pattern. We use a real WBM data set (WM-811K) released by Taiwan semiconductor manufacturing company.

• Advances in nano research
• /
• v.12 no.6
• /
• pp.605-616
• /
• 2022

We prepared two samples of ultrafine ferrihydrite (FH) nanoparticle ensembles of quite a different origin. First is the biosynthesized sample (as a product of the vital activity of bacteria Klebsiella oxytoca (hereinafter marked as FH-bact) with a natural organic coating and negligible magnetic interparticle interactions. And the second one is the chemically synthesized ferrihydrite (hereinafter FH-chem) without any coating and high level of the interparticle interactions. The interparticle magnetic interactions have been tuned by modifying the nanoparticle surface in both samples. The coating of the FH-bact sample has been partially removed by annealing at 150℃ for 24 h (hereinafter FH-annealed). The FH-chem sample, vice versa, has been coated (1.0 g) with biocompatible polysaccharide (arabinogalactan) in an ultrasonic bath for 10 min (hereinafter FH-coated). The changes in the surface properties of nanoparticles have been controlled by XPS. According to the electron microscopy data, the modification of the nanoparticle surface does not drastically change the particle shape and size. A change in the average nanoparticle size in sample FH-annealed to 3.3 nm relative to the value in the other samples (2.6 nm) has only been observed. The estimated particle coating thickness is about 0.2-0.3 nm for samples FH-bact and FH-coated and 0.1 nm for sample FH-annealed. Mössbauer and magnetization measurements are definitely shown that the drastic change in the blocking temperature is caused by the interparticle interactions. The experimental temperature dependences of the hyperfine field hf>(T) for samples FH-bact and FH-coated have not revealed the effect of interparticle interactions. Otherwise, the interparticle interaction energy E_int estimated from the hf>(T) for samples FH-chem and FH-annealed has been found to be 121k_B and 259k_B, respectively.

• Journal of Intelligence and Information Systems
• /
• v.28 no.4
• /
• pp.309-328
• /
• 2022

This study empirically analyzed a Korean pre-trained language models (PLMs) designed for natural language generation. The performance of two PLMs - BART and GPT - at the task of abstractive text summarization was compared. To investigate how performance depends on the characteristics of the inference data, ten different document types, containing six types of informational content and creation content, were considered. It was found that BART (which can both generate and understand natural language) performed better than GPT (which can only generate). Upon more detailed examination of the effect of inference data characteristics, the performance of GPT was found to be proportional to the length of the input text. However, even for the longest documents (with optimal GPT performance), BART still out-performed GPT, suggesting that the greatest influence on downstream performance is not the size of the training data or PLMs parameters but the structural suitability of the PLMs for the applied downstream task. The performance of different PLMs was also compared through analyzing parts of speech (POS) shares. BART's performance was inversely related to the proportion of prefixes, adjectives, adverbs and verbs but positively related to that of nouns. This result emphasizes the importance of taking the inference data's characteristics into account when fine-tuning a PLMs for its intended downstream task.

• Korean Journal of Construction Engineering and Management
• /
• v.23 no.6
• /
• pp.111-118
• /
• 2022

Recently, as underground facilities buried under roads in Korea are aging, the amount of underground facility maintenance work is rapidly increasing. For the maintenance and management of such underground facilities, the cutting work of the road pavement should be preceded. However, the conventional road pavement cutters used in Korea are not eco-friendly, and the reality is that they generate a lot of noise and cutting sludge (scattering dust). Therefore, in this study, the concept of the cutting sludge recovery device was derived, and an eco-friendly pavement cutter including this function was designed and manufactured. The developed equipment took about 20 to 30 seconds to cut 1m to a depth of 100 to 150mm. Also, the sludge suction performance was good in most sections, and the noise level of the equipment briefly measured at a distance of 2m was 82.7dB on average. However, due to the limitation that the developed equipment was at the level of the first prototype, the driving stability was somewhat low, and equipment abnormalities such as engine shutdown and sludge recovery performance decreased in some cases. The cutting performance and sludge recovery function will be more stable through tuning and improvement of the developed prototype in the future. In addition, we plan to quantitatively compare and analyze productivity by applying the improved prototype to actual field conditions.

• Journal of Adhesion and Interface
• /
• v.23 no.2
• /
• pp.53-58
• /
• 2022

Here, the surface characteristics of the dielectric were controlled by introducing the self-assembled monolayers (SAMs) as the intermediate layers on the surface of the AlOx dielectric, and the electrical performances of the organic charge modulated transistor (OCMFET) were significantly improved. The organic intermediate layer was applied to control the surface energy of the AlOx gate dielectric acting as a capacitor plate between the control gate (CG) and the floating gate (FG). By applying the intermediate layers on the gate dielectric surface, and the field-effect mobility (μ_OCMFET) of the OCMFET devices could be efficiently controlled. We used the four kinds of SAM materials, octadecylphosphonic acid (ODPA), butylphosphonic acid (BPA), (3-bromopropyl)phosphonic acid (BPPA), and (3-aminopropyl)phosphonic acid (APPA), and each μ_OCMFET was measured at 0.73, 0.41, 0.34, and 0.15 cm2V-1s-1, respectively. The results could be suggested that the characteristics of each organic SAM intermediate layer, such as the length of the alkyl chain and the type of functionalized end-group, can control the electrical performances of OCMFET devices and be supported to find the optimized fabrication conditions, as an efficient sensing platform device.

• Journal of the Korea Convergence Society
• /
• v.12 no.11
• /
• pp.35-43
• /
• 2021

Recently, there has been a growing interest in zero-shot cross-lingual transfer, which leverages cross-lingual language models (CLLMs) to perform downstream tasks that are not trained in a specific language. In this paper, we point out the limitations of the data-centric aspect of quality estimation (QE), and perform zero-shot cross-lingual transfer even in environments where it is difficult to construct QE data. Few studies have dealt with zero-shots in QE, and after fine-tuning the English-German QE dataset, we perform zero-shot transfer leveraging CLLMs. We conduct comparative analysis between various CLLMs. We also perform zero-shot transfer on language pairs with different sized resources and analyze results based on the linguistic characteristics of each language. Experimental results showed the highest performance in multilingual BART and multillingual BERT, and we induced QE to be performed even when QE learning for a specific language pair was not performed at all.

• Journal of The Korean Society of Integrative Medicine
• /
• v.10 no.3
• /
• pp.199-208
• /
• 2022

Purpose : The study aimed to examine gender-differences in muscle tone, elasticity, and stiffness of the erector spinae in young and old subjects. This study also aimed to assess the effect of aging on muscle tone, elasticity, and stiffness of the erector spinae in men and women, and compare the trend of the aging effect between the two gender groups. Methods : With the muscle in the relaxed state and subjects in the prone position, a myotonometer was used to quantify muscle tone, elasticity, and stiffness, of the erector spinae in 102 participants[46 males (29 young subjects, aged 22.48±2.23 years and 17 old subjects, aged 76.35±3.71 years), 56 females (40 young subjects, aged 20.38±1.43 years and 16 old subjects, aged 74.56±5.40 years)]. Results : The tone and stiffness of the erector spinae muscles were greater in men than in women for both age groups (p<.001-.01), while elasticity did not show a significant difference between men and women. For the direction of change, both male and female groups showed significantly increased tone and stiffness, and decreased elasticity with increasing age (p<.001). For age-related changes, a different tendency was observed between men and women. Men showed a greater increase in tone than women with aging. in contrast, both men and women exhibited a similar decrease or increase in elasticity and stiffness. Conclusion : Gender-differences in the erector spinae in terms of muscle tone and stiffness were observed. Regardless of the age, men had higher muscle tone and stiffness than women, but not elasticity. The erector spinae muscles showed age-related changes in all aspects of muscle tone, elasticity, and stiffness, in both men and women. Notably, men presented greater variation than women in the amount of increase of muscle tone with aging. These findings have implications for musculoskeletal therapeutic approaches, and gender-customized tuning may be indicated for designing exercise interventions to prevent and manage gender-sensitive muscular injuries or diseases and frailty.

• Economic Analysis
• /
• v.27 no.4
• /
• pp.1-42
• /
• 2021

The outlook for Korea's consumer price inflation rate has a profound impact not only on the Bank of Korea's operation of the inflation target system but also on the overall economy, including the bond market and private consumption and investment. This study presents the prediction results of consumer price inflation in Korea for the next three years. To this end, first, model selection is performed based on the out-of-sample predictive power of autoregressive distributed lag (ADL) models, AR models, small-scale vector autoregressive (VAR) models, and large-scale VAR models. Since there are many potential predictors of inflation, a Bayesian variable selection technique was introduced for 12 macro variables, and a precise tuning process was performed to improve predictive power. In the case of the VAR model, the Minnesota prior distribution was applied to solve the dimensional curse problem. Looking at the results of long-term and short-term out-of-sample predictions for the last five years, the ADL model was generally superior to other competing models in both point and distribution prediction. As a result of forecasting through the combination of predictions from the above models, the inflation rate is expected to maintain the current level of around 2% until the second half of 2022, and is expected to drop to around 1% from the first half of 2023.

• Composites Research
• /
• v.21 no.5
• /
• pp.9-14
• /
• 2008

Carbon nanofibers (CNFs) were used as dielectric lossy materials and NiFe particles were used as magnetic lossy materials. Total twelve specimens for the three types such as dielectric, magnetic and mixed radar absorbing materials (RAMs) were fabricated. Their complex permittivities and permeabilities in the range of $2{\sim}18$ GHz were measured using the transmission line technique. The parametric studios for reflection loss characteristics of each specimen to design the single-layered RAMs were performed. The mixed RAMs generally showed the improved absorbing characteristics with thinner matching thickness. One of the mixed RAMs, MD3with the thickness of 2.00 mm had the 10 dB absorbing bandwidth of 4.0 GHz in the X-band ($8.2{\sim}12.4$ GHz). It also showed very broad 10 dB absorbing bandwidth as wide as 6.0 GHz in the Ku-band ($12.0{\sim}18.0$ GHz) with the thickness tuning to 1.49 mm. The experimental results for selected several specimens were in very good agreements with simulation ones in terms of the overall reflection loss characteristics and 10 dB absorbing bandwidth.