公开(公告)号：KR100815563B1

公开(公告)日：2008-03-20

申请号：KR1020060081785

申请日：2006-08-28

Applicant: 한국과학기술정보연구원

Inventor： 정한민 ,  강인수 ,  김평 ,  이승우 ,  이미경 ,  성원경 ,  박동인

IPC: G06F17/30 ,  G06F17/00

CPC classification number: G06F17/30424 ,  G06N5/02

Abstract: 본 발명은 DBMS 기반 지식 확장 및 추론 서비스 시스템 및 그 방법을 제공하기 위한 것으로, 온톨로지 스키마와 인스턴스들을 이용하여 트리플로 표현되는 지식을 생성하는 지식 생성부와; 지식 생성부에서 생성된 지식을 입력받고, 지식 확장 규칙을 이용하여 지식을 확장하는 지식 확장 처리부; 및 지식 확장 처리부에서 확장된 지식을 이용하고, 사용자 질의를 변환하고 질의를 처리하여 추론 서비스를 제공하는 추론 서비스 제공부;를 포함하여 구성함으로서, 온톨로지를 이용하여 지식을 생성하고, 확장 규칙을 이용하여 지식을 확장하고, 질의 처리를 통해 추론 서비스를 제공하여 시맨틱 웹 기술을 활용하고 추론 서비스를 제공할 필요가 있는 전 분야에 걸쳐서 안정적이고 효율적인 프레임워크를 제시함으로써 시맨틱 웹 서비스의 상용화를 가능하게 할 수 있게 되는 것이다.
DBMS, 온톨로지, 지식 확장, 추론 서비스, RDF 트리플

公开(公告)号：KR1020080019440A

公开(公告)日：2008-03-04

申请号：KR1020060081787

申请日：2006-08-28

Applicant: 한국과학기술정보연구원

Inventor： 정한민

IPC: G06F17/21

CPC classification number: G06F17/30707 ,  G06F17/2755 ,  G06F17/2795 ,  G06F17/30011 ,  G06F17/30737

Abstract: A thesaurus extension method using statistics-based document classification and a hybrid document classification method are provided to extend a thesaurus by obtaining a quality set through statistics-based document classification and using keywords obtained from the quality set as candidate terms, and classify document stably by combining a statistics-based and thesaurus-based document classification system. Statistics-based document classification is performed by extracting qualities from more than one of a test document set, a field classification system, and a correct answer set, performing machine training and document classification, forming a quality set, and repeating previous steps(ST1). A thesaurus is extended by obtaining the quality set from the statistics-based document classification, selecting keywords from the quality set, determining the selected keywords as candidate terms, constructing the thesaurus, and repeating the previous steps(ST2). A field classification system and the thesaurus are mapped by using taxonomy(ST3). Thesaurus-based document classification is performed(ST4). Hybrid document classification is performed by combining the statistics-based and thesaurus-based document classification(ST5). A classification result is evaluated, a correct answer set is made, and a process is repeated(ST7).

Abstract translation: 提供了使用基于统计的文档分类和混合文档分类方法的词库扩展方法，通过基于统计的文档分类获得质量集合并使用从质量集合获得的关键词作为候选词来扩展词库，并且通过 结合基于统计数据和基于词典的文档分类系统。 基于统计的文档分类是通过从测试文档集，场分类系统和正确答案集中的多个提取质量，执行机器训练和文档分类，形成质量集合以及重复前面的步骤（ST1）来执行的。 。 通过从基于统计的文档分类中获取质量集，从质量集中选择关键词，将所选择的关键词确定为候选词，构建辞典并重复前面的步骤（ST2）来扩展词库。 使用分类法（ST3）映射场分类系统和同义词库。 执行基于词库的文档分类（ST4）。 通过组合基于统计数据和基于词库的文档分类（ST5）来执行混合文档分类。 评估分类结果，进行正确的答案设定，重复处理（ST7）。

公开(公告)号：KR100659370B1

公开(公告)日：2006-12-19

申请号：KR1020060014749

申请日：2006-02-15

Applicant: 한국과학기술정보연구원

Inventor： 정한민 ,  성원경 ,  김평 ,  이미경 ,  구희관 ,  구남앙 ,  박동인

IPC: G06F17/40 ,  G06F17/30

Abstract: A document database forming method and an information retrieval method using thesaurus matching are provided to easily form a document database by automatically forming the document database with a program, and quickly update the document database in case that a new field and term is generated, as formation of the document database is automatically changed by update of a field classification system and a thesaurus. A document is read and contents of the document are extracted(S120). An index word of the document is extracted by performing morpheme analysis for the extracted document(S140). The index word is analyzed through the thesaurus by matching with the thesaurus(S150). An analysis result is stored to the document database(S160). Thesaurus matching is analyzed by comparing more than one condition selected from a term frequency, a document frequency, a field classification frequency, and a concept word depth of the index word.

Abstract translation: 提供了一种使用辞典匹配的文档数据库形成方法和信息检索方法，以通过利用程序自动形成文档数据库来容易地形成文档数据库，并且在产生新字段和词语的情况下快速更新文档数据库， 通过更新字段分类系统和同义词库来自动改变文档数据库的内容。 读取文档并提取文档的内容（S120）。 通过对所提取的文档执行语素分析来提取文档的索引词（S140）。 索引词通过与辞典匹配的辞典进行分析（S150）。 分析结果被存储到文档数据库（S160）。 通过比较从术语频率，文档频率，字段分类频率和索引词的概念词深度中选择的多于一个条件来分析词库匹配。

公开(公告)号：KR100644317B1

公开(公告)日：2006-11-10

申请号：KR1020050078461

申请日：2005-08-25

Applicant: 한국과학기술정보연구원

Inventor： 정한민 ,  김원용 ,  구남앙 ,  성원경

IPC: G06F15/16

Abstract: A system and a method for distribution information based on ontology are provided to offer high quality information, and track a document written by the user, and manage history of the document by converting meta information written from a user into the ontology and using the ontology. A format verifying module(302) checks whether the document includes virus or worm, and an extension of the document is correct or not in case that a document registration request signal for managing/distributing the information is received from a user computer. An ontology repository(306) converts the meta information received from the user computer into the ontology, and stores the information for the document with the information including URI(Uniform Resource Identifier), structure, concept, and additional information for the document. A program storing part(316) includes an ontology storing/managing program for converting the meta information into the ontology after the meta information is verified by linking with the ontology and assigning an ID to each ontology information by combining the IDs.

Abstract translation: 本发明提供了一种基于本体的分发信息的系统和方法，以提供高质量的信息并跟踪用户编写的文档，并通过将从用户写入的元信息转换为本体并使用本体来管理文档的历史。 格式验证模块（302）在从用户计算机接收到用于管理/分发信息的文档注册请求信号的情况下，检查文档是否包含病毒或蠕虫，并且文档的扩展是否正确。 本体库（306）将从用户计算机接收到的元信息转换为本体，并且利用包括URI（统一资源标识符），结构，概念和文档的附加信息的信息来存储文档的信息。 程序存储部分（316）包括本体存储/管理程序，用于在元信息通过与本体链接而被验证之后，将元信息转换为本体，并通过组合这些ID分配ID给每个本体信息。