• Annual Conference on Human and Language Technology
• /
• 2004.10d
• /
• pp.181-188
• /
• 2004

대다수의 개방영역 자연어 질의응답 시스템은 답을 선택할 수 있는 개념영역을 미리 정의하고 있기 때문에 시스템이 준비하지 못한 범주의 개념을 묻는 질의문에 대해서는 올바른 응답을 생성하지 못하거나 예외 처리 방식으로 응답을 생성해 낸다. 본 논문에서는 전형적인 범주에 속하지 않는 명사 개념에 관한 질의문에 대해 범용적으로 대응할 수 있는 개방영역 자연어 질의응답 시스템을 제안한다. 제안하는 시스템은 상위 개념 명사구(Hypernym)에 포함되는 하위 개념의 명사구(Hyponym)들을 추출할 수 있는 일반적인 패턴들을 그 신뢰도와 함께 가지고 있다. 따라서 질의문이 임의의 명사구 개념을 요청할 때 정답의 후보들을 동적으로 생성되는 가상의 is-a 의미관계 사전으로부터 신뢰 순위로 정렬하여 추출해 낼 수가 있다. 제안하는 시스템은 "What 명사구 동사구" 형태의 질의문들 중에서 개체명 인식기나 시소러스를 이용하여 정답 후보를 손쉽게 생성할 수 있는 질의문을 배제한 실험용 질의문 집합을 이용한 실험에서 42%의 재현율을 보였다.

• Annual Conference on Human and Language Technology
• /
• 2004.10d
• /
• pp.46-52
• /
• 2004

본 논문에서는 질의-응답 시스템의 질의에서 많이 나타나는 시간 표현을 인식하고, 인식한 시간 표현에 대해서 정규화 하는 방법을 제안한다. 본 논문에서 사용하는 질의-응답 시스템의 도메인은 TV방송 스케줄, 날씨 정보이며, 이러한 도메인에서는 시간 표현이 매우 빈번하게 사용되기 때문에 질의에 나타나는 시간 표현을 정확하게 인식해서 정규화 하는 것이 중요하다. 제안하는 방법은 시간 표현을 의미와 기능에 따라 분류하고 각 유형마다 적절한 인식 및 정규화 기법을 사용한다. 질의에서 시간 표현은 시간 개체명 태거. 품사 태거, 시간 파서를 사용하여 인식하고, 시간 추론기와 시간 표현 사전을 이용하여 정규화 한다. TV방송 스케줄과 날씨 정보 도메인의 280개 질의에서 184개의 시간표현을 이용하여 평가한 결과, 시간 표현의 인식과 정규화는 각각 93%와 96%의 정확률, 97%와 93%의 재현율을 보였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.05a
• /
• pp.389-392
• /
• 2021

문장 압축은 원본 문장의 중요한 의미를 보존하는 짧은 길이의 압축 문장을 생성하는 자연어처리 태스크이다. 문장 압축은 사용자가 텍스트로부터 필요한 정보를 빠르게 획득할 수 있도록 도울 수 있어 활발히 연구되고 있지만, 기존 연구들은 사람이 직접 정의한 압축 규칙이 필요하거나, 모델 학습을 위해 대량의 데이터셋이 필요하다는 문제점이 존재한다. 사전 학습된 언어 모델을 통한 perplexity 기반의 문장 점수 측정을 통해 문장을 압축하여 압축 규칙과 모델 학습을 위한 데이터셋이 필요하지 않은 연구 또한 존재하지만, 문장 점수 측정에 문장에 속한 단어들의 의미적 중요도를 반영하지 못하여 중요한 단어가 삭제되는 문제점이 존재한다. 본 논문은 언어 정보 중 품사 정보, 의존관계 정보, 개체명 정보의 중요도를 수치화하여 perplexity 기반의 문장 점수 측정에 반영하는 방법을 제안한다. 또한 제안한 문장 점수 측정 방법을 활용하였을 때 문장 점수 측정 기반 문장 압축 모델의 문장 압축 성능이 향상됨을 확인하였으며, 이를 통해 문장에 속한 단어의 언어 정보를 문장 점수 측정에 반영하는 것이 의미적으로 적절한 압축 문장을 생성하는 데 도움이 될 수 있음을 보였다.

• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.383-388
• /
• 2020

본 연구에서는 AI 어시스턴트의 음악청취 도메인 내 요청문을 인식 및 처리하기 위해 한국어와 중국어를 중심으로 도메인 사전 및 패턴문법 언어자원을 구축하고 그 결과를 비교분석 하였다. 이를 통해 향후 다국어 언어자원 구축의 접근 방법을 모색할 수 있으며, 궁극적으로 패턴 기반 문법으로 기술한 언어자원을 요청문 인식에 직접 활용하고 또한 주석코퍼스 생성을 통해 기계학습 성능 향상에 도움을 줄 수 있을 것으로 기대된다. 본 연구에서는 우선 패턴문법의 구체적인 양상을 살펴보기에 앞서, 해당 도메인의 요청문 유형의 카테고리를 결정하는 과정을 거쳤다. 이를 기반으로 한국어와 중국어 요청문의 실현 양상과 패턴유형을 LGG 프레임으로 구조화한 후, 한국어와 중국어 패턴문법 간의 통사적, 형태적, 어휘적 차이점을 비교분석 하여 음악청취 도메인 요청문의 언어별 생성 구조 차이점을 관찰할 수 있었다. 구축한 패턴문법은 개체명을 변수(X)로 설정하는 경우, 한국어에서는 약 2,600,600개, 중국어에서는 약 11,195,600개의 표현을 인식할 수 있었다. 결과적으로 본 연구에서 제안한 언어자원의 언어별 차이에 대한 통찰을 통해 다국어 차원의 요청문 인식 자원과 기계학습 데이터로서의 효용을 확인하였다.

• Journal of KIISE:Software and Applications
• /
• v.33 no.1
• /
• pp.114-129
• /
• 2006

Ontology mapping is the task of finding semantic correspondences between two ontologies. In order to improve the effectiveness of ontology mapping, we need to consider the characteristics and constraints of data models used for implementing ontologies. Earlier research on ontology mapping, however, has proven to be inefficient because the approach should transform input ontologies into graphs and take into account all the nodes and edges of the graphs, which ended up requiring a great amount of processing time. In this paper, we propose a multi-strategic mapping approach to find correspondences between ontologies based on the syntactic or semantic characteristics and constraints of the topic maps. Our multi-strategic mapping approach includes a topic name-based mapping, a topic property-based mapping, a hierarchy-based mapping, and an association-based mapping approach. And it also uses a hybrid method in which a combined similarity is derived from the results of individual mapping approaches. In addition, we don't need to generate a cross-pair of all topics from the ontologies because unmatched pairs of topics can be removed by characteristics and constraints of the topic maps. For our experiments, we used oriental philosophy ontologies, western philosophy ontologies, Yahoo western philosophy dictionary, and Yahoo german literature dictionary as input ontologies. Our experiments show that the automatically generated mapping results conform to the outputs generated manually by domain experts, which is very promising for further work.

• Journal of the Korean Society for information Management
• /
• v.37 no.2
• /
• pp.197-223
• /
• 2020

The purpose of this study is to analyze the issues resulted from the process of grouping KORMARC records using FRBR WORK concept and to suggest a new method. The previous studies did not sufficiently address the criteria or processes for identifying representative authors of records and their derivatives. Therefore, our study focused on devising a method of identifying the representative author when there are multiple contributors in a work. The study developed a method of identifying representative authors using an author role dictionary constructed by extracting role-terms from the statement of responsibility field (245). We also designed another way to group records as a work by calculating similarity measures of authors and titles. The accuracy rate of WORK grouping was the highest when blank spaces, parentheses, and controling processes were removed from titles and the measured similarity rates of authors and titles were higher than 80 percent. This was an experiment study where we developed an author-role dictionary that can be utilized in selecting a representative author and measured the similarity rate of authors and titles in order to achieve effective WORK grouping of KORMARC records. The future study will attempt to devise a way to improve the similarity measure of titles, incorporate FRBR Group 1 entities such as expression, manifestation and item data into the algorithm, and a method of improving the algorithm by utilizing other forms of MARC data that are widely used in Korea.

• Journal of Intelligence and Information Systems
• /
• v.24 no.3
• /
• pp.243-262
• /
• 2018

Database schema integration is a significant issue in information systems. Because schema integration is a time-consuming and labor-intensive task, many studies have attempted to automate it. Researchers typically use XML as the source schema and leave much of the work to be done through DBA intervention, e.g., there are various naming conflicts related to relationship names in schema integration. In the past, the DBA had to intervene to resolve the naming-conflict name. In this paper, we introduce an algorithm that automatically generates relationship names to resolve relationship name conflicts that occur during schema integration. This algorithm is based on an Internet collocation and English sentence example dictionary. The relationship between the two entities is generated by analyzing examples extracted based on dictionary data through natural language processing. By building a semi-automated schema integration system and testing this algorithm, we found that it showed about 90% accuracy. Using this algorithm, we can resolve the problems related to naming conflicts that occur at schema integration automatically without DBA intervention.

• Journal of KIISE
• /
• v.43 no.7
• /
• pp.773-780
• /
• 2016

Semantic information and features are very important for Semantic Role Labeling(SRL) though many SRL systems based on machine learning mainly adopt lexical and syntactic features. Previous SRL research based on semantic information is very few because using semantic information is very restricted. We proposed the SRL system which adopts semantic information, such as named entity, word sense disambiguation, filtering adjunct role based on sense, synonym cluster, frame extension based on synonym dictionary and joint rule of syntactic-semantic information, and modified verb-specific numbered roles, etc. According to our experimentations, the proposed present method outperforms those of lexical-syntactic based research works by about 3.77 (Korean Propbank) to 8.05 (Exobrain Corpus) F1-scores.

• Journal of The Korean Association For Science Education
• /
• v.35 no.4
• /
• pp.691-707
• /
• 2015

This study aims to develop biology teachers' education program based on argumentation activity about core concepts of evolution and to analyze the characteristics of core concepts of evolution learned during the program. The eight core concepts of evolution in this study were variation, heritability of variation, competition, natural selection, adaptation, differential reproductive rate of individuals, changes in genetic pool within a population, and macroevolution. The performances of teachers participating in the program were compared before and after argumentation activities; consisting of seven sessions on the eight core concepts of evolution. The process of the program was specially designed by learning cycle model for teacher education, consisting of seven phases: identification of the task, production of a tentative argument, small group's written argument, share arguments with the other groups, reflective discussion, final written argument, and organization by an instructor. Participants in the study were two pre-service biology teachers and four in-service biology teachers. The results suggest that biology teachers reduced the teleological explanation for biological evolution and improve its adequacy after the intervention. Teachers lacked the opportunity to discuss variation, heritability of variation, competition, and macroevolution because science textbooks lack information on the concepts of biological evolution. The results of this study suggest that because the argumentation program developed for teachers helps to improve understanding the concepts of evolution and to reduce inadequate conceptions in biology, teacher education programs using argumentation activity and eight core concepts of evolution will play a role for efficient evolution education for biology teachers.

• Journal of The Korean Association For Science Education
• /
• v.36 no.5
• /
• pp.801-814
• /
• 2016

The purpose of this study is to find out whether pre-service biology teachers' biological meanings of competition and adaptation is the same as everyday meanings to understand natural selection and evolution. Fifty three pre-service biology teachers answered written tests and thirteen of them were interviewed. The tests on general meanings on competition and adaptation, tests on biological meanings on competition and adaptation and tests on natural selection were administered. Questions on semi-structured interviews were composed of only biological meanings about competition and adaptation. Analyzing data through four steps, pre-service biological teachers' general meanings on competition and adaptation were compared with the definitions of Korean dictionary and compared with biological definitions. Pre-service teachers' alternative conceptions about biological competition and adaptation were classified and the alternative conceptual types were checked when explaining natural selection and evolution. They realized competition is battle for intention or purpose like everyday meaning of competition. However, they didn't realize adaptation happen in the population level. They thought that adaptation, like in everyday life meaning, is getting suitable to environment for survival on an individual level. By relating adaptation to evolution, they thought that long-lasting adaptation become evolution. This study has significance with respect to suggesting the alternative concepts of competition and adaptation based on effective evolution teaching from constructivism viewpoint and highlighting the importance of the concepts of biological competition and adaptation, which have been concerned rarely for a long time.