• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.722-725
• /
• 2023

대규모 언어 모델(LLM)은 대규모의 데이터를 학습하여 얻은 지식을 기반으로 텍스트와 다양한 콘텐츠를 인식하고 요약, 번역, 예측, 생성할 수 있는 딥러닝 알고리즘이다. 초기 공개된 LLM은 영어 기반 모델로 비영어권에서는 높은 성능을 기대할 수 없었으며, 이에 한국, 중국 등 자체적 LLM 연구개발이 활성화되고 있다. 본 논문에서는 언어가 LLM의 성능에 영향을 미치는가에 대하여 한국어 기반 LLM과 영어 기반 LLM으로 KoBEST의 4가지 Task에 대하여 성능비교를 하였다. 그 결과 한국어에 대한 사전 지식을 추가하는 것이 LLM의 성능에 영향을 미치는 것을 확인할 수 있었다.

• The Journal of Korean Academic Society of Nursing Education
• /
• v.1 no.1
• /
• pp.46-61
• /
• 1995

The purpose of this study was to identify the degree of learning desire in the health education of military drill curriculum and to furnish basic data for the effective health education at high school. A total of 594 high school students were participated in the questionnaires and the 12 domains & the 55 questions were utilized for this study. The survey was conducted from March 15 to March 27, 1993 and the collected date were analized by T-test and F-test. The results of this study are as followings : 1. In the 12 domains, respondents indicated high degree of learning desire in 'human & sex', 'nutrition' and 'disease of adult' in that order. However 'accident & disaster', 'transport & management' and 'nursing' were not highly ranked. 2. In the 55 questions, respondents showed high degree of learning desire in 'artificial respiration' & 'cardiopulmonary resuscitation', 'precautionary of adult disease' and 'sex & sex moral' in that order, but low degree of learning desire in 'the management of many wounded persons' and the 'synopsis of nursing'. 3. Comparing the degree of learning desire by grade, the 1st, the 3rd and the End grade were ranked in that order. The 1st and 3rd graders showed higher degree of learning desire in 'human & sex', and the 2nd graders in 'nutrition'. 4. Also, female students showed higher degree of learning desire than male students in general. Female students indicated it in 'nutrition', 'human & sex' and 'adult disease' in that order while male students in 'human & sex', 'adult disease' and 'nutrition' in that order. 5. The academic high school students showed higher degree of learning desire than the vocational high school students. 'Human & sex' was highest ranked at both academic and vocational high school students.

• Asia pacific journal of information systems
• /
• v.34 no.1
• /
• pp.209-225
• /
• 2024

This research aims to use machine learning technology in human resource management to predict employees' work-life balance. The study utilized a dataset from IBM Watson Analytics in the IBM Community for the machine learning analysis. Multinomial dependent variables concerning workers' work-life balance were examined, categorized into continuous and categorical types using the Generalized Linear Model. The complexity of assessing variable roles and their varied impact based on the type of model used was highlighted. The study's outcomes are academically and practically relevant, showcasing how machine learning can offer further understanding of psychological variables like work-life balance through analyzing employee profiles.

• Journal of Information Processing Systems
• /
• v.4 no.4
• /
• pp.159-168
• /
• 2008

Human-Robot Interaction (HRI), based on already well-researched Human-Computer Interaction (HCI), has been under vigorous scrutiny since recent developments in robot technology. Robots may be more successful in establishing common ground in project-based education or foreign language learning for children than in traditional media. Backed by its strong IT environment and advances in robot technology, Korea has developed the world's first available e-Learning home robot. This has demonstrated the potential for robots to be used as a new educational media - robot-learning, referred to as 'r-Learning'. Robot technology is expected to become more interactive and user-friendly than computers. Also, robots can exhibit various forms of communication such as gestures, motions and facial expressions. This study compared the effects of non-computer based (NCB) media (using a book with audiotape) and Web-Based Instruction (WBI), with the effects of Home Robot-Assisted Learning (HRL) for children. The robot gestured and spoke in English, and children could touch its monitor if it did not recognize their voice command. Compared to other learning programs, the HRL was superior in promoting and improving children's concentration, interest, and academic achievement. In addition, the children felt that a home robot was friendlier than other types of instructional media. The HRL group had longer concentration spans than the other groups, and the p-value demonstrated a significant difference in concentration among the groups. In regard to the children's interest in learning, the HRL group showed the highest level of interest, the NCB group and the WBI group came next in order. Also, academic achievement was the highest in the HRL group, followed by the WBI group and the NCB group respectively. However, a significant difference was also found in the children's academic achievement among the groups. These results suggest that home robots are more effective as regards children's learning concentration, learning interest and academic achievement than other types of instructional media (such as: books with audiotape and WBI) for English as a foreign language.

• Journal of the Semiconductor & Display Technology
• /
• v.20 no.3
• /
• pp.120-124
• /
• 2021

Recently, human attribute recognition (HAR) attracts a lot of attention due to its wide application in video surveillance systems. Recent deep-learning-based solutions for HAR require time-consuming training processes. In this paper, we propose a post-processing technique that utilizes the historical video frames to improve prediction results without invoking re-training or modifying existing deep-learning-based classifiers. Experiment results on a large-scale benchmark dataset show the effectiveness of our proposed method.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.382-387
• /
• 2023

현재 GPT-4와 같은 거대한 언어 모델이 기계 번역, 요약 및 대화와 같은 다양한 작업에서 압도적인 성능을 보이고 있다. 그러나 이러한 거대 언어 모델은 학습 및 적용에 상당한 계산 리소스와 도메인 특화 미세 조정이 어려운 등 몇 가지 문제를 가지고 있다. In-Context learning은 데이터셋에서 추출한 컨택스트의 정보만으로 효과적으로 작동할 수 있는 효율성을 제공하여 앞선 문제를 일부 해결했지만, 컨텍스트의 샷 개수와 순서에 민감한 문제가 존재한다. 이러한 도전 과제를 해결하기 위해, 우리는 Super In-Context Learning (SuperICL)을 활용한 새로운 방법론을 제안한다. 기존의 SuperICL은 적용한 플러그인 모델의 출력 정보를 이용하여 문맥을 새로 구성하고 이를 활용하여 거대 언어 모델이 더욱 잘 분류할 수 있도록 한다. Super In-Context Learning for Generation은 다양한 자연어 생성 작업에 효과적으로 최적화하는 방법을 제공한다. 실험을 통해 플러그인 모델을 교체하여 다양한 작업에 적응하는 가능성을 확인하고, 자연어 생성 작업에서 우수한 성능을 보여준다. BLEU 및 ROUGE 메트릭을 포함한 평가 결과에서도 성능 향상을 보여주며, 선호도 평가를 통해 모델의 효과성을 확인했다.

• Journal of the Korea Society for Simulation
• /
• v.31 no.1
• /
• pp.11-18
• /
• 2022

Recently, research to classify human activity using imagery has been actively conducted for the purpose of crime prevention and facility safety in public places and facilities. In order to improve the performance of human activity classification, most studies have applied deep learning based-transfer learning. However, despite the increase in the number of backbone network models that are the basis of deep learning as well as the diversification of architectures, research on finding a backbone network model suitable for the purpose of operation is insufficient due to the atmosphere of using a certain model. Thus, this study applies the transfer learning into recently developed deep learning backborn network models to build an intelligent system that classifies human activity using imagery. For this, 12 types of active and high-contact human activities based on sports, not basic human behaviors, were determined and 7,200 images were collected. After 20 epochs of transfer learning were equally applied to five backbone network models, we quantitatively analyzed them to find the best backbone network model for human activity classification in terms of learning process and resultant performance. As a result, XceptionNet model demonstrated 0.99 and 0.91 in training and validation accuracy, 0.96 and 0.91 in Top 2 accuracy and average precision, 1,566 sec in train process time and 260.4MB in model memory size. It was confirmed that the performance of XceptionNet was higher than that of other models.

• Journal of The Korean Society of Integrative Medicine
• /
• v.9 no.1
• /
• pp.141-150
• /
• 2021

Purpose : The purpose of this study was to investigate the effect of case-based learning application in human growth development classes on nursing students' problem-solving ability, self-directed learning ability, and academic self-efficacy. Methods : The research method was a self-report questionnaire before and after case-based learning for second-year nursing students who took the human growth development course at U University in K city. The collected data were statistically processed using SPSS WIN 21.0. Results : The results of the study showed that after case-based learning, problem-solving ability, self-directed learning ability, and academic self-efficacy were all significantly improved. In addition, as a result of examining the correlation between each variable after case-based learning, problem solving ability score and self-directed learning ability score (r=.54, ｐ<.01), and problem solving ability scores and academic self-efficacy scores (r=.44, ｐ<.01), were significantly correlated with self-directed learning ability scores and the academic self-efficacy reduction scores (r=.76, ｐ<.01). Conclusion : The results of this study suggested the need for various learning programs such as case-based learning to improve nursing students' problem-solving abilities and self-directed learning abilities and their application. In addition, to improve the learning self-efficacy of nursing students, a continuous and systematic study is suggested to develop and apply customized educational programs according to the learners' preferences. Since the sample group in this study was limited to one university, there were few cases and no control group, so there are limitations in generalizing the test effect, However, significant differences a were verified in the case-based learning pre-tests and post-tests.

• Journal of Engineering Education Research
• /
• v.26 no.6
• /
• pp.19-29
• /
• 2023

This study aims to develop and apply a metaverse-based instructional design model for the education in science and technology. It analyzed the concept and characteristics of metaverse, existing non-contact education models, and major teaching strategies systematically. Based on the prior researches, an instructional design model using metaverse is developed that presents metaverse-related teaching strategies and design principles for the before-, during-, and after-lesson phases. Then, this model was applied to a project-based learning program, conducted a perception survey on instructors and learners, and revised the metaverse instructional design model based on the results of the survey. In the Metaverse Instructional Design Model, before-lesson phase is a physical and psychological preparation stage for class participation, which includes familiarization with the Metaverse learning environment, formation of expectations for education, and self-directed pre-learning. During the lesson, to effectively deliver the lesson content, it is necessary to build confidence in the learning environment, promote learning participation, provide reference materials, perform team projects and provide feedback, digest learning content, and transfer learning content. The after-lesson phase provides strategies for ongoing interaction between learners and mentors. This study introduces a new instructional design model that utilizes metaverse and shows the potential of metaverse-based education in science and technology. It also has important implications in that it provides practical guidelines for the effective design and implementation of metaverse-based education.

• International Journal of Control, Automation, and Systems
• /
• v.1 no.2
• /
• pp.236-242
• /
• 2003

This paper presents a cooperation strategy between a human operator and autonomous robots for an interactive robot soccer game, The interactive robot soccer game has been developed to allow humans to join into the game dynamically and reinforce entertainment characteristics. In order to make these games more interesting, a cooperation strategy between humans and autonomous robots on a team is very important. Strategies can be pre-programmed or learned by robots themselves with learning or evolving algorithms. Since the robot soccer system is hard to model and its environment changes dynamically, it is very difficult to pre-program cooperation strategies between robot agents. Q-learning - one of the most representative reinforcement learning methods - is shown to be effective for solving problems dynamically without explicit knowledge of the system. Therefore, in our research, a Q-learning based learning method has been utilized. Prior to utilizing Q-teaming, state variables describing the game situation and actions' sets of robots have been defined. After the learning process, the human operator could play the game more easily. To evaluate the usefulness of the proposed strategy, some simulations and games have been carried out.