• 한국어정보학회:학술대회논문집
• /
• 2017.10a
• /
• pp.29-34
• /
• 2017

챗봇 혹은 대화 시스템은 특정 질문이나 발화에 대해 적절한 응답을 해주는 시스템으로 자연어처리 분야에서 활발히 연구되고 있는 주제 중 하나이다. 최근에는 대화 모델 학습에 딥러닝 방식의 시퀀스-투-시퀀스 프레임워크가 많이 이용되고 있다. 하지만 해당 방식을 적용한 모델의 경우 학습 데이터에 나타나지 않은 다양한 형태의 질의문에 대해 응답을 잘 못해주는 문제가 있다. 이 논문에서는 이러한 문제점을 해결하기 위하여 디노이징 응답 생성 모델을 제안한다. 제안하는 방법은 다양한 형태의 노이즈가 임의로 가미된 질의문을 모델 학습 시에 경험시킴으로써 강건한 응답 생성이 가능한 모델을 얻을 수 있게 한다. 제안하는 방법의 우수성을 보이기 위해 9만 건의 질의-응답 쌍으로 구성된 한국어 대화 데이터에 대해 실험을 수행하였다. 실험 결과 제안하는 방법이 비교 모델에 비해 정량 평가인 ROUGE 점수와 사람이 직접 평가한 정성 평가 모두에서 더 우수한 결과를 보이는 것을 확인할 수 있었다.

• Proceedings of the Korea Multimedia Society Conference
• /
• 1998.10a
• /
• pp.79-84
• /
• 1998

일반적으로 동영상 검색을 위해 주석 기반과 내용 기반 검색을 사용하는데, 주석 기반 검색은 사용자의 주관이 개입되어 일관성을 잃기 쉽고, 내용 기반 검색은 동영상 데이터가 담고 있는 의미가 추출되기 어렵다는 단점을 가지고 있다. 그래서, 본 논문에서는 이 두 검색 기법을 상호 보완하여 검색의 효율성과 정확성을 높이기 위해 통합 동영상 데이터 모델(IVDM)을 제안하고, 이것을 기반으로 뉴스 동영상 검색을 위한 데이터베이스 스키마와 질의 유형을 설계한다. 이 모델은 동영상 데이터를 계층적으로 구조화한 형태로 상위수준에서는 주제별로 부여된 메타 정보로 주석 기반 검색을 지원하고, 하위 수준에서는 동영상 데이터에서 색깔, 모양, 움직임, 질감 등의 특징 데이터를 추출하여 내용 기반 검색을 지원한다.

• Therapeutic Science for Rehabilitation
• /
• v.9 no.4
• /
• pp.33-43
• /
• 2020

Objective: This study used a mixed methods analytical approach to analyze the level of depression in and quality of life of patients who had been exposure to Agent Orange. Methods: Data on the general characteristics, depression level, and the quality of life of 29 patients who were hospitalized because of exposure to Agent Orange were collected. Focus group interviews were conducted with 17 of these 29 patients. Regarding mixed methods, the sequential integrated design model proposed by Teddlie and Tashakkori (2003) was employed. The analytical methods were as follows: (1) general characteristics were examined using frequencies and proportional distributions; (2) depression scores and quality of life scores were assessed using descriptive statistical analysis; and (3) content analysis was used to examine the focus group interview data. Results: The average K-GDS depression score was 19.24 points, and the average WHOQOL-BREF quality of life score was 57.66 points. The focus group interview data yielded 3 topics were named "deployment route", "the manner of exposure to Agent Orange", and "life after Agent Orange exposure". Conclusion: There is an urgent need to increase public awareness, develop a public policy response, and conduct additional research on ground occupational therapy programs that include physical therapy.

• Journal of dental hygiene science
• /
• v.13 no.4
• /
• pp.359-368
• /
• 2013

This study aimed to evaluate conditions and performance of quality improvement (QI) activity in Korean dental hospitals. Twenty one representative dental hospitals in Korea were selected by the selection criteria. A questionnaire was implemented to measure awareness, perceived needs, educations, conferences, works, activating/obstructive factors and satisfaction related to QI activities. Satisfaction score range from 0 (lowest) to 5 (highest) among 329 dental hygienists who worked in the selected 21 dental hospitals. Implementation rate of programs on external customer satisfaction was the highest 59.9% in QI activities, and reduction of expenses (59.4%), Improvement in health care quality (58.7), risk management (52.8) were followed, while improvement in work efficiency (46.5) showed the lowest implementation rate. The most influencing factor accelerating QI activities was 'active participation of hospital workers' (54.7%), and 'effort to improve customer satisfaction' (44.7%) followed. The most influencing QI related obstructive factor was 'lack of QI-related manpower and support' (47.1%). A balanced development including QI programs with lower implementation rates is necessary. Encouraging participation of workers may be the most important in developing dental QI activities further.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.14 no.5
• /
• pp.35-43
• /
• 2014

A data allocation technique is essential to improve the performance of data broadcast systems. This paper explores the issues for allocating data items on broadcast channels in the environment where multiple-data queries are submitted, and proposes a new data allocation scheme named DAMA. The proposed scheme employs the strategy that the broadcast frequency of each data is determined by the moving average of its access probability. DAMA could enhance the performance of query response time since it is capable of controlling the influence of access probability properly according to the query size. Simulation is performed to evaluate the performance of the proposed scheme. The simulation results show that the performance of DAMA is superior to other schemes in terms of the average response time.

• Journal of Digital Contents Society
• /
• v.19 no.5
• /
• pp.927-933
• /
• 2018

One of the factors that have the greatest influence on the error of the question answering system that finds and provides answers to natural language questions is the step of searching for documents or passages that contain correct answers. In order to improve the retrieval performance, it is necessary to understand the characteristics of documents and passages containing correct answers. This paper experimentally analyzes how many question words appear in the correct answer documents, how the location of the question word is distributed, and how the topic of the question and the correct answer document are similar using the corpus composed of the question, the documents with correct answer, and the documents without correct answer. This study explains the causes of previous search research results for question answer system and discusses the necessary elements of effective search step.

• Journal of the Korea Society of Computer and Information
• /
• v.16 no.7
• /
• pp.95-103
• /
• 2011

The actual utilization of Semantic Web technology which aims to provide more intelligent and automated service for information retrieval over the Web becomes gradually reality. RDF is widely used as the one of standard formats to present and manage the voluminous data on the Web. Efficient query processing on RDF data, therefore, is one of the ongoing research topics. Retrieving resources having a specific association from a given resource is the typical query processing type and several researches for this have done. However the most of previous researches have not fully considered discovering the complex relationship among resources such as returning the association between resources as the query processing result. This paper introduces the indexing and query processing for ${\rho}$-intersect operation which is one of the semantic association retrieval types. It includes an indexing scheme using suffix array and optimal processing approaches for handling ${\rho}$-intersect operation. The experimental evaluations shows that the average execution times for the proposed approach is 3~7 times faster than the previous approach.

• Annual Conference on Human and Language Technology
• /
• 2004.10d
• /
• pp.275-282
• /
• 2004

본 논문은 3단계 정답 추출 방법을 통해 백과사전 인물분야 질의응답 시스템을 구현하는 방법을 제안한다. 논문에서 제안한 3단계 정답 추출 방법은 1) 백과사전 문서 내에서 정형화 될 수 있는 지식들을 추출한 백과사전 KB 기반 정답 추출 방법, 2) 문장을 언어분석 하여 LF(Logical Form)구조를 추출하여 색인한 LF 기반 정답추출 방법, 3) 각 문장을 주제 태깅을 하여, 주제별로 묶어 의미적 단락으로 구분하고 단락 검색을 기반으로 정답을 추정하는 의미적 단락 기반 정답 추출 방법으로 구성되어 있다. 이러한 방법론은 백과사전이라는 문서 도메인의 특성을 반영하고. 사용자 질문의 난이도 또는 형태에 따라서 정답을 제공할 수 있는 백과사전 인물분야 질의응답 시스템에 적합하다.

• Journal of the Korean Society for information Management
• /
• v.28 no.1
• /
• pp.7-22
• /
• 2011

Internet portals have been revolutionized not only as simple search engines but also as a new space for the Internet users. They have developed to give satisfying search results for academic information users. academic fields. However, their attention was given to the quantity rather than the quality of the results. This tendency is now changing. This study addresses the problems in the search process using the current portal sites and presents an integrated scholarly information service where users can access more organized and trustworthy information linked with online technical keyword dictionary. When a user enter a keyword on a portal site, he/she can access to high quality scholarly information resources linked with keyword. This could assure the user to get an expanded knowledge with confirmation.

• The Journal of Korean Academy of Prosthodontics
• /
• v.62 no.3
• /
• pp.183-192
• /
• 2024

Purpose. This study aimed to evaluate the reliability and information quality of YouTube videos on implant overdenture searched in two languages (Korean and English). Materials and methods. Youtube, an online video sharing platform was searched using search terms in two different languages related to implant overdenture. A total of 120 videos were selected (60 videos for each search term), then the reliability and information quality of the videos were evaluated. Topic domain, DISCERN instrument, and JAMA benchmark were used to evaluate the reliability and information quality of the videos. Statistical analyses were performed by using the Mann-Whitney U test and Kruskal-Wallis test. Results. Out of a total of 120 videos, the topic domain scores of 78 (65.0%) videos were evaluated as 'poor', and the DISCERN scores of 104.5 (87.1%) videos were evaluated as 'very poor' and 'poor'. The Korean videos had significantly higher topic domain scores and DISCERN scores than the English videos (P < .05). 3.5 Korean videos and 4 English videos met the criteria for attribution of JAMA benchmark. Conclusion. The reliability and information quality of YouTube videos on implant overdenture were low.