• Title/Summary/Keyword: 정보이론적 학습

Search Result 641, Processing Time 0.031 seconds

BERT-based Hateful Text Filtering System - Focused on University Petition System (BERT 기반 혐오성 텍스트 필터링 시스템 - 대학 청원 시스템을 중심으로)

  • Taejin Moon;Hynebin Bae;Hyunsu Lee;Sanguk Park;Youngjong Kim
    • Annual Conference of KIPS
    • /
    • 2023.05a
    • /
    • pp.714-715
    • /
    • 2023
  • 최근들어 청원 시스템은 사람들의 다양한 의견을 반영하고 대응하기 위한 중요한 수단으로 부상하고 있다. 그러나 많은 양의 청원 글들을 수작업으로 분류하는 것은 매우 시간이 많이 소요되며, 인적 오류가 발생할 수 있는 문제점이 존재한다. 이를 해결하기 위해 자연어처리(NLP) 기술을 활용한 청원 분류 시스템을 개발하는 것이 필요하다. 본 연구에서는 BERT(Bidirectional Encoder Representations from Transformers)[1]를 기반으로 한 텍스트 필터링 시스템을 제안한다. BERT 는 최근 자연어 분류 분야에서 상위 성능을 보이는 모델로, 이를 활용하여 청원 글을 분류하고 분류된 결과를 이용해 해당 글의 노출여부를 결정한다. 본 논문에서는 BERT 모델의 이론적 배경과 구조, 그리고 미세 조정 학습 방법을 소개하고, 이를 활용하여 청원 분류 시스템을 구현하는 방법을 제시한다. 우리가 제안하는 BERT 기반의 텍스트 필터링 시스템은 청원 글 분류를 자동화하고, 이에 따른 대응 속도와 정확도를 향상시킬 것으로 기대된다. 또한, 이 시스템은 다양한 분야에서 응용 가능하며, 대용량 데이터 처리에도 적합하다. 이를 통해 대학 청원 시스템에서 혐오성 발언 등 부적절한 내용을 사전에 방지하고 학생들의 의견을 효율적으로 수집할 수 있는 기능을 제공할 수 있다는 장점을 가지고 있다.

Implementing Neural Network and measuring execution speed using CUDA based on Parallel Computing (CUDA를 사용한 병렬 컴퓨팅 기반 신경망 구현 및 수행 속도 측정)

  • Jang, Yong-Seok;Jeon, Woong-Gi;O, Byeong-Jin;Choi, Heung-Kook
    • Proceedings of the Korea Multimedia Society Conference
    • /
    • 2012.05a
    • /
    • pp.275-278
    • /
    • 2012
  • 신경망 이론은 그 특성상 각각의 뉴런과 신경들 사이의 병렬적인 처리에 의해 Input에 대한 Output을 계산해 낸다. 하지만, 현대 컴퓨터들은 CPU를 통한 순차처리 방식으로 정보를 취급하기에 그 근본 구조가 달라 병렬구조를 모사하기 위해 계산하는 과정에 많은 시간이 소요된다. 본 논문에서는 신경망 학습을 NVIDIA사에서 제공한 CUDA를 사용하여 병렬 컴퓨팅 구조로 수행함으로서 시간을 단축시키는 것을 확인하고자 한다.

  • PDF

A Rule Extraction Method Using Relevance Factor for FMM Neural Networks (FMM 신경망에서 연관도요소를 이용한 규칙 추출 기법)

  • Lee, Seung Kang;Lee, Jae Hyuk;Kim, Ho Joon
    • KIPS Transactions on Software and Data Engineering
    • /
    • v.2 no.5
    • /
    • pp.341-346
    • /
    • 2013
  • In this paper, we propose a rule extraction method using a modified Fuzzy Min-Max (FMM) neural network. The suggested method supplements the hyperbox definition with a frequency factor of feature values in the learning data set. We have defined a relevance factor between features and pattern classes. The proposed model can solve the ambiguity problem without using the overlapping test process and the contraction process. The hyperbox membership function based on the fuzzy partitions is defined for each dimension of a pattern class. The weight values are trained by the feature range and the frequency of feature values. The excitatory features and the inhibitory features can be classified by the proposed method and they can be used for the rule generation process. From the experiments of sign language recognition, the proposed method is evaluated empirically.

Learning Multidimensional Sequential Patterns Using Hellinger Entropy Function (Hellinger 엔트로피를 이용한 다차원 연속패턴의 생성방법)

  • Lee, Chang-Hwan
    • The KIPS Transactions:PartB
    • /
    • v.11B no.4
    • /
    • pp.477-484
    • /
    • 2004
  • The technique of sequential pattern mining means generating a set of inter-transaction patterns residing in time-dependent data. This paper proposes a new method for generating sequential patterns with the use of Hellinger measure. While the current methods are generating single dimensional sequential patterns within a single attribute, the proposed method is able to detect multi-dimensional patterns among different attributes. A number of heuristics, based on the characteristics of Hellinger measure, are proposed to reduce the computational complexity of the sequential pattern systems. Some experimental results are presented.

Human Ear Detection for Biometries (생체인식을 위한 귀 영역 검출)

  • Kim Young-Baek;Rhee Sang-Yong
    • Journal of the Korean Institute of Intelligent Systems
    • /
    • v.15 no.7
    • /
    • pp.813-816
    • /
    • 2005
  • Ear detection is an important part of an non-invasive ear recognition system. In this paper we propose human ear detection from side face images. The proposed method is made by imitating the human recognition process using feature information and color information. First, we search face candidate area in an input image by using 'skin-color model' and try to find an ear area based on edge information. Then, to verify whether it is the ear area or not, we use the SVM (Support Vector Machine) based on a statistical theory. The method shows high detection ratio in indoors environment with stable illumination.

The Analysis of Difficulties of Pre-service Teachers in Process of Instructional Design based on Programming (프로그래밍 기반 수업 설계 과정에서 예비 교사의 어려움 분석)

  • Kim, Seong-Won;Lee, Youngjun
    • Proceedings of the Korean Society of Computer Information Conference
    • /
    • 2019.07a
    • /
    • pp.209-210
    • /
    • 2019
  • 정보 교과가 아닌 타 교과에서 교육 도구로써 프로그래밍의 교육적 효과가 증명됨에 따라 프로그래밍을 수업에 도입하기 위한 다양한 연구가 진행되었다. 김성원과 이영준(2017)은 TPACK 이론을 도입하여, 테크놀로지 도구로 프로그래밍 언어를 도입하고, 예비 교사의 수업 전문성 향상을 위한 프로그래밍 기반 TPACK 교육 프로그램을 개발하였다. 개발한 교육 프로그램을 수업에 도입하였지만, 예비 교사는 여러 가지 어려움을 느끼고 있었다. 따라서 본 연구에서는 프로그래밍 기반 TPACK 교육 프로그램을 개선하기 위하여 예비 교사에게 선행 연구에서 개발한 교육 프로그램을 적용하고, 프로그래밍 기반 수업 설계를 하는 과정에서 예비 교사가 겪는 어려움을 조사하였다. 연구 결과, 예비 교사는 예비 교사는 프로그래밍 학습, 프로그램의 설계 및 구현, 교과 특성에 맞는 프로그램 개발, 프로그래밍 언어의 한계로 프로그래밍 기반 수업 설계에서 어려움을 겪고 있었다. 향후 연구에서는 이러한 어려움을 해결하기 위한 교육 프로그램 개선을 진행하고, 개선한 교육 프로그램을 예비 교사에게 적용하여 효과를 검증하고자 한다.

  • PDF

Iterative Feedback-based Personality Persona Generation for Diversifying Linguistic Patterns in Large Language Models (대규모 언어 모델의 언어 패턴 다양화를 위한 반복적 피드백 기반 성격 페르소나 생성법)

  • Taeho Hwang;Hoyun Song;Jisu Shin;Sukmin Cho;Jong C. Park
    • Annual Conference on Human and Language Technology
    • /
    • 2023.10a
    • /
    • pp.454-460
    • /
    • 2023
  • 대규모 언어 모델(Large Language Models, LLM)의 발전과 더불어 대량의 학습 데이터로부터 기인한 LLM의 편향성에 관심이 집중하고 있다. 최근 선행 연구들에서는 LLM이 이러한 경향성을 탈피하고 다양한 언어 패턴을 생성하게 하기 위하여 LLM에 여러가지 페르소나를 부여하는 방법을 제안하고 있다. 일부에서는 사람의 성격을 설명하는 성격 5 요인 이론(Big 5)을 이용하여 LLM에 다양한 성격 특성을 가진 페르소나를 부여하는 방법을 제안하였고, 페르소나 간의 성격의 차이가 다양한 양상의 언어 사용 패턴을 이끌어낼 수 있음을 보였다. 그러나 제한된 횟수의 입력만으로 목표하는 성격의 페르소나를 생성하려 한 기존 연구들은 세밀히 서로 다른 성격을 가진 페르소나를 생성하는 데에 한계가 있었다. 본 연구에서는 페르소나 부여 과정에서 피드백을 반복하여 제공함으로써 세세한 성격의 차이를 가진 페르소나를 생성하는 방법론을 제안한다. 본 연구의 실험과 분석을 통해, 제안하는 방법론으로 형성된 성격 페르소나가 다양한 언어 패턴을 효과적으로 만들어 낼 수 있음을 확인했다.

  • PDF

Fast 3D Model Extraction Algorithm with an Enhanced PBIL of Preserving Depth Consistency (깊이 일관성을 보존하는 향상된 개체군기반 증가 학습을 이용한 고속 3차원 모델 추출 기법)

  • 이행석;장명호;한규필
    • Journal of KIISE:Computer Systems and Theory
    • /
    • v.31 no.1_2
    • /
    • pp.59-66
    • /
    • 2004
  • In this paper, a fast 3D model extraction algorithm with an enhanced PBIL of preserving depth consistency is proposed for the extraction of 3D depth information from 2D images. Evolutionary computation algorithms are efficient search methods based on natural selection and population genetics. 2D disparity maps acquired by conventional matching algorithms do not match well with the original image profile in disparity edge regions because of the loss of fine and precise information in the regions. Therefore, in order to decrease the imprecision of disparity values and increase the quality of matching, a compact genetic algorithm is adapted for matching environments, and the adaptive window, which is controlled by the complexity of neighbor disparities in an abrupt disparity point is used. As the result, the proposed algorithm showed more correct and precise disparities were obtained than those by conventional matching methods with relaxation scheme.

A Convergence Work Experience of Nursing Students under Preceptorship (간호대학생의 프리셉터십 융복합 실습 경험)

  • Nam, Mun-Hee;Kim, Hae-Ok
    • Journal of Digital Convergence
    • /
    • v.15 no.9
    • /
    • pp.283-300
    • /
    • 2017
  • This study is to understand the work experiences of nursing students under preceptorship in 3+1 year program, and suggests effective practice to prevent stopping nursing profession. Four focus group interviews were made with 23 nursing students in 3+1 year program. Data were analyzed through the grounded theory of Corbin and Strauss' (2008). Through comparative analysis, central phenomenon of was 'looking for lighthouse'. The causal condition was 'opportunity to learning practice'. Contextual conditions were 'mood of being a true nurse', 'getting vivid information about clinical field', and 'lack of students perception about preceptorship'. Intervention conditions were 'work environment', 'proactive learning' and 'attachment to the hospital'. Action/interaction strategies were 'adjustment to diverse demands of patients and guardians', 'feeling as if it were one's own affair', and 'looking as much as knowing'. Consequence were 'pride' and 'itinerary of learning as a new graduate nurse'. This study is meaningful because it gives useful basis to nursing students in the process of 'looking for lighthouse'.

A Study on the Curriculum for Record Management Science Education - with focus on the Faculty of Cultural Information Resources, Surugadai University; Evolving Program, New Connections (기록관리학의 발전을 위한 교육과정연구 -준하태(駿河台)(스루가다이)대학(大學)의 경우를 중심(中心)으로-)

  • Kim, Yong-Won
    • Journal of Korean Society of Archives and Records Management
    • /
    • v.1 no.1
    • /
    • pp.69-94
    • /
    • 2001
  • The purpose of this paper is to provide an overview of the current status of the records management science education in Japan, and to examine the implications of the rapid growth of this filed while noting some of its significant issues and problems. The goal of records management science education is to improve the quality of information services and to assure an adequate supply of information professionals. Because records management science programs prepare students for a professional career, their curricula must encompass elements of both education and practical training. This is often expressed as a contrast between theory and practice. The confluence of the social, economic and technological realities of the environment where the learning takes place affects both. This paper reviews the historical background and current trends of records management science education in Japan. It also analyzes the various types of curriculum and the teaching staff of these institutions, with focus on the status of the undergraduate program at Surugadai University, the first comprehensive, university level program in Japan. The Faculty of Cultural Information Resources, Surugadai University, a new school toward an integrated information disciplines, was opened in 1994, to explore the theory and practice of the management diverse cultural information resources. Its purpose was to stimulate and promote research in additional fields of information science by offering professional training in archival science, records management, and museum curatorship, as well as librarianship. In 1999, the school introduced a master program, the first in Japan. The Faculty has two departments and each of them has two courses; Department of Sensory Information Resources Management; -Sound and Audiovisual Information Management, -Landscape and Tourism Information Management, Department of Knowledge Information Resources Management; -Library and Information Management, -Records and Archives Management The structure of the entire curriculum is also organized in stages from the time of entrance through basic instruction and onwards. Orientation subjects which a student takes immediately upon entering university is an introduction to specialized education, in which he learns the basic methods of university education and study, During his first and second years, he arranges Basic and Core courses as essential steps towards specialization at university. For this purpose, the courses offer a wide variety of study topics. The number of courses offered, including these, amounts to approximately 150. While from his third year onwards, he begins specific courses that apply to his major field, and in a gradual accumulation of seminar classes and practical training, puts his knowledge grained to practical use. Courses pertaining to these departments are offered to students beginning their second year. However, there is no impenetrable wall between the two departments, and there are only minor differences with regard requirements for graduation. Students may select third or fourth year seminars regardless of the department to which they belong. To be awarded a B.A. in Cultural Information Resources, the student is required to earn 34 credits in Basic Courses(such as, Social History of Cultural Information, Cultural Anthropology, History of Science, Behavioral Sciences, Communication, etc.), 16 credits in Foreign Languages(including 10 in English), 14 credits on Information Processing(including both theory and practice), and 60 credits in the courses for his or her major. Finally, several of the issues and problems currently facing records management science education in Japan are briefly summarized below; -Integration and Incorporation of related areas and similar programs, -Curriculum Improvement, -Insufficient of Textbooks, -Lack of qualified Teachers, -Problems of the employment of Graduates. As we moved toward more sophisticated, integrated, multimedia information services, information professionals will need to work more closely with colleagues in other specialties. It will become essential to the survival of the information professions for librarians to work with archivists, record managers and museum curators. Managing the changes in our increasingly information-intensive society demands strong coalitions among everyone in cultural Institutions. To provide our future colleagues with these competencies will require building and strengthening partnerships within and across the information professions and across national borders.