• Title/Summary/Keyword: 모델 이해

Search Result 2,808, Processing Time 0.029 seconds

Evaluating the Impact of Training Conditions on the Performance of GPT-2-Small Based Korean-English Bilingual Models

  • Euhee Kim;Keonwoo Koo
    • Journal of the Korea Society of Computer and Information
    • /
    • v.29 no.9
    • /
    • pp.69-77
    • /
    • 2024
  • This study evaluates the performance of second language acquisition models learning Korean and English using the GPT-2-Small model, analyzing the impact of various training conditions on performance. Four training conditions were used: monolingual learning, sequential learning, sequential-interleaved learning, and sequential-EWC learning. The model was trained using datasets from the National Institute of Korean Language and English from BabyLM Challenge, with performance measured through PPL and BLiMP metrics. Results showed that monolingual learning had the best performance with a PPL of 16.2 and BLiMP accuracy of 73.7%. In contrast, sequential-EWC learning had the highest PPL of 41.9 and the lowest BLiMP accuracy of 66.3%(p < 0.05). Monolingual learning proved most effective for optimizing model performance. The EWC regularization in sequential-EWC learning degraded performance by limiting weight updates, hindering new language learning. This research improves understanding of language modeling and contributes to cognitive similarity in AI language learning.

3D-QSARs analyses for Tyrosinase Inhibitory Activity of 2-Phenyl-1,4-benzopyrone (Flavones) Analogues and Molecular Docking (2-Phenyl-1,4-benzopyrone 유도체 (Flavones)의 Tyrosinase 저해활성에 관한 3D-QSARs 분석과 분자도킹)

  • Park, Joon-Ho;Sung, Nack-Do
    • Journal of Applied Biological Chemistry
    • /
    • v.53 no.4
    • /
    • pp.225-231
    • /
    • 2010
  • To understand the inhibitory activity with changing hydroxyl substituents ($R_l-R_9$) of polyhydroxy substituted 2-phenyl-l,4-benzopyrone analogues (1-25) against tyrosinase (PDB ID: oxy-form; 1WX2), molecular docking and the three dimensional quantitative structure-activity relationships (3D-QSARs: Comparative molecular field analysis (CoMFA) & Comparative molecular similarity indices analysis (CoMSIA)) were studied quantitatively. The statistically best models were CoMFA 1 and CoMSIA 1 model from the results. The optimized CoMSIA 1 model with the sensitivity of the perturbation and the prediction produced ($dq^2'/dr_{yy'}^2$=1.009 & $q^2$=0.51l) by a progressive scrambling analysis were not dependent on chance correlation. The inhibitory activities with optimized CoMSIA 1 model were dependent upon electrostatic factor (51.4%) of substrate molecules. Contour mapping the 3D-QSAR models to the active site of tyrosinase provides new insight into the interaction between tyrosinase as receptor and 2-phenyl-l,4-benzopyrone analogues as inhibitor. Therefore, the results will he able to apply to the optimization of a new potent tyrosinase inhibitors.

Actantial Model-based Character Role Recognition using Emotional Flow Graph among Characters in Text Stories (텍스트 스토리에서 등장인물간 감정 흐름 그래프를 이용한 행위소 모델 기반의 등장인물 역할 인식)

  • Yu, Hye-Yeon;Kim, Moon-Hyun;Bae, Byung-Chull
    • Journal of Internet Computing and Services
    • /
    • v.22 no.1
    • /
    • pp.51-63
    • /
    • 2021
  • Identifying characters who appear in a story and analyzing their relationships is essential to understanding the story. This paper aims to identify the two actants (or character roles) as Helper and Opponent in Greimas's Actantial model by identifying Subject (i.e., protagonist) and analyzing the emotional interactions between the Subject and the two actants (Helper/Opponent). Our proposed method consists of three steps. First, we identify objects (i.e., characters) appearing in the text story. Next, we extract relational information through the interaction of the characters, and then classify emotions in the text expressed as relational information. Finally, we represent the flow of emotional relations among characters as a directed graph. The node with the highest degree is considered as the Subject because it includes the most relational information. The node that sends the most positive/negative emotions to the Subject is considered as the Helper/Oppenent, respectively. Our research contributes to the computer-based narrative understanding by providing a computational method that automatically extracts the three key character roles (Subject, Helper, and Opponent) from the text story.

Dimensionality of emotion suppression and psychosocial adaptation: Based on the cognitive process model of emotion processing (정서 처리의 인지 평가모델을 기반으로 한 정서 억제의 차원성과 심리 사회적 적응)

  • Woo, Sungbum
    • Korean Journal of Culture and Social Issue
    • /
    • v.27 no.4
    • /
    • pp.475-503
    • /
    • 2021
  • The purpose of this study is to clarify the constructs of emotion suppression and help understanding on the multidimensional nature of emotion suppression by classifying constructs for suppression according to the KMW model. Also, this study examined the gender differences of emotion suppression. For this purpose, 657 adult male and female subjects were evaluated for attitude toward emotions, and difficulty in emotional regulation, as well as depression, state anger and daily stress scale. As a result of the exploratory factor analysis on the scales related to the emotion suppression factors, the emotion suppression factors corresponding to each stage of the KMW model were found to be 'distraction against emotional information, 'difficulty in understanding and interpretation of emotions', 'emotion control beliefs', 'vulnerability on emotional expression beliefs'. Next, the study participants were classified by performing a cluster analysis based on each emotion suppression factor. As a result, four clusters were extracted and named 'emotional control belief cluster', 'emotional expression cluster', 'emotional attention failure cluster', and 'general emotional suppression cluster'. As a result of examining the average difference of male depression, depression, state anger, and daily stress for each group, significant differences were found in all dependent variables. As a result of examining whether there is a difference in the frequency of emotional suppression clusters according to gender, the frequency of emotional suppression clusters was high in men, and the ratio of emotional expression clusters was high in women. Finally, it was analyzed whether there was a gender difference in the effect of the emotional suppression cluster on psychosocial adaptation, and the implications were discussed based on the results of this study.

Analysis of Nursing College Students' Educational Needs about Program outcomes (간호대학생의 프로그램학습성과에 대한 교육요구 분석)

  • Nam, Soung-Mi
    • Journal of the Korea Academia-Industrial cooperation Society
    • /
    • v.19 no.7
    • /
    • pp.315-324
    • /
    • 2018
  • This study was conducted to identify and analyze the educational needs for program outcomes in nursing students. Participants included 256 nursing students at K college, and a survey was conducted from October 30 to December 22, 2017. Data was analyzed with SPSS WIN 20.0 using descriptive statistics, t-test, ANOVA, Borich's Needs Analysis, and the Locus for Focus Model. Results found that subjects recognized that the present level was high about 'Apply therapeutic communication skills,' 'Provide fundamental nursing skills' and 'Understanding nursing profession standards'. 'Provide fundamental nursing skills', 'Apply a nursing intervention', 'Understanding legal and ethical standards of nursing practice and implementation' were the most important sub-elements of program outcomes. In according with the Borich's needs value and the Locus for Focus Model, 'Apply a nursing intervention', 'Explain the importance of the role and cooperation within the healthcare team', 'Provide fundamental nursing skills', 'Execute the nursing process', 'Understanding the legal and ethical standards of nursing practice and implement' were identified as the top priorities of educational needs. Results of this study are expected to be used as basic data to develop outcome-based curriculum and nursing education programs.

Analysis on Factors Influencing the Path to Use Information Technology in Public Organization (공공부문 구성원의 정보기술 사용경로와 영향요인 분석: 심리적 대응이론을 적용한 구조적 분석)

  • Lee, Cheoul-Joo
    • Information Systems Review
    • /
    • v.12 no.2
    • /
    • pp.47-74
    • /
    • 2010
  • This study presents two dimensional views of IT-use behaviors. It examines public employee's IT-use behavior as manifested in two ways: acceptance on IT, utilization on IT. This study investigates levels of employee's the acceptance and utilization on IT in the Korean Central Government(ministry and agency) and analyses factors affecting public employee's the acceptance and utilization on task-related IT in the public sector. This study tries to examine the variables affecting the acceptance and utilization on IT; especially, the influences of such variables as leadership, task-related communication with people, task motivation, IT usability and ease of uses, task understanding in the context of Korean Central Government. Based on a questionnaire survey in Korean Central Government(8 ministries, 5 agencies) that have implemented Government Knowledge Management System(GKMS), 311 valid instruments were received and analyzed using various statistics and structural equation model(SEM). And this study will adopt try to delineate pre and intervening conditions for public employee's the acceptance and utilization. According to the results of interviews and SEM analysis suggests that the influence of task-understanding, organizational communication, IS usability were statistically significant in the acceptance and utilization on IT. And it is also hoped that the path-analytic model suggested in this study helps reveal the causal chains between the acceptance and the utilization. Especially, task-understanding proves to be important in encouraging public employee's the acceptance and utilization. This result illustrates why task-understanding people are the key to utilize the task-related IT in public organization.

A Study on the Health Information Management Practice Program Model for EMR Certification System Education -Focus on Patient Information Management- (EMR 인증제 교육을 위한 보건의료정보관리 실습 프로그램 모델 연구 -환자정보관리 중심-)

  • Choi, Joon-Young
    • Journal of the Health Care and Life Science
    • /
    • v.9 no.1
    • /
    • pp.1-9
    • /
    • 2021
  • In this study, a model in which certification standards were added to the health information management practice program was studied and presented in order to understand the EMR certification standards implemented by the Korea Health and Medical Information Service. In the practice program, the certification standard function for patient information management was added to the health information management education system to practice and understand patient information management that corresponds to the functional standard of the EMR certification system. The EMR certification standard practice program for patient information management is composed of the following certification standards. registration number and personal information management, treatment reservation schedule management, personal information revision history management, identification of people with the same name, integrated management of multiple registration numbers, patient search by identification information, patient search by health care type, surgical procedure consent record and inquiry, record/inquiry of consent form for personal information use, display of life-sustaining medical decision information, registration/inquiry of external medical institution documents, registration and inquiry of external examination results. In this way, by operating and practicing the functions of the health information system according to the certification standards, it is possible to understand and practice the certification standards and details of patient information management in the functional area of the certification standards. In addition, since the function of the EMR certification standard can be checked, it will be possible to improve the management ability of the electronic medical record system of the health information manager in the medical institution.

Development and Application of a Science History Role-Playing Game for High School Students' Understanding of Nature of Science: Focus on Storytelling of the Continental Drift Theory (고등학생의 과학의 본성 이해를 위한 과학사 롤플레잉게임(SHRPG) 개발 및 적용 -대륙이동설 스토리텔링을 중심으로-)

  • Shim, Eun-Ji;Choe, Seung-Urn;Kim, Chan-Jong
    • Journal of The Korean Association For Science Education
    • /
    • v.39 no.1
    • /
    • pp.45-57
    • /
    • 2019
  • NOS education through the history of science is regarded effective. However, science teaching has been criticized for not considering the interest of the learners enough and providing the context of learning themes that hinder the understanding of NOS. This study intends to convey the NOS element through the rich context of storytelling. The theme of the story is the history of continental drift, in which, the debate of many scientists and Wegener's creativity are prominent. Of the various media that deliver storytelling, the most powerful medium that leads to personal immersion is computer games, and among many kinds of games, the main genre of storytelling is role-playing games (RPGs). We developed the science history role-playing game (SHRPG) focusing on continental drift. The game development procedure followed Kim's 4F process (2017), which consists of the Figure Out, Focus, Fun Design, and Finalize. The story was constructed based on the NOS elements of Lederman et al. (2002), namely creativity and imagination demand, subjectivity, socio-cultural personality and tentativeness, which are all present in the story of the continental drift theory. The mechanics and rules of the RPG included quests, rewards, quizzes, NOS scores, and rankings. In the final phase of development, the game developed was pilot tested four times. The results of the tests showed that students' understanding of NOS through SHRPG has increased, especially in the creativity domain. The students' satisfaction with the fun, sympathy, and immersion during the game was very high.

Learning from Instruction: A Comprehension-Based Approach (지시문을 통한 학습: 이해-기반 접근)

  • Kim, Shin-Woo;Kim, Min-Young;Lee, Jisun;Sohn, Young-Woo
    • Korean Journal of Cognitive Science
    • /
    • v.14 no.3
    • /
    • pp.23-36
    • /
    • 2003
  • A comprehension-based approach to learning assumes that incoming information and background knowledge are integrated to form a mental representation which is subsequently used to incorporate new knowledge. It is demonstrated that this approach can be validated by comparing human and computational model performance in the prompt learning context. A computational model (ADAPT-UNIX) based on the construction-integration theory of comprehension (Kintsch, 1988; 1998) predicted how users learn from help prompts which are designed to assist UNIX composite command production. In addition, the comparison also revealed high similarity in composite production task performance between model and human. Educational implications of present research are discussed on the basis of the fact that prompt instructions have differential effect on learning and application as background knowledge varies.

  • PDF

Teacher Knowledge Necessary to Address Student Errors and Difficulties about Ratio and Rate (비와 비율에 관한 학생의 오류와 어려움 해결을 위해 필요한 교사지식)

  • Kang, Hyangim;Choi, Eun Ah
    • School Mathematics
    • /
    • v.17 no.4
    • /
    • pp.613-632
    • /
    • 2015
  • In this study, we hope to reveal teacher knowledge necessary to address student errors and difficulties about ratio and rate. The instruments and interview were administered to 3 in-service primary teachers with various education background and teaching experiments. The results of this study are as follows. Specialized content knowledge(SCK) consists of profound knowledge about ratio and rate beyond multiplicative comparison of two quantities and professional knowledge about the definitions of textbook. Knowledge of content and students(KCS) is the ability to recognize students' understanding the concept and the representation about ratio and rate. Knowledge of content and teaching(KCT) is made up of knowledge about various context and visual models for understanding ratio and rate.