-
公开(公告)号：KR1020160149660A
公开(公告)日：2016-12-28
申请号：KR1020150087120
申请日：2015-06-19
Applicant: 강원대학교산학협력단 , 한국전자통신연구원
Abstract: 본발명의한국어상호참조해결시스템은국어문자열을포함하는문서를입력받아, 문서내의문장들의의존구문구조와, 문장내에등장하는개체명및 어절들의품사를찾아내는과정을처리하는전 처리부, 상기전 처리부에서전 처리된상호참조해결의대상이되는명사구들을추출하기위한멘션탐지부, 상기멘션탐지부에서추출된명사구들에대한멘션간의참조여부를임시로결정하는모듈로서, 가이드자질을제공하기위한규칙기반모듈부, 상기멘션탐지부에서추출한멘션들을둘씩묶어멘션페어평가시 지표가되는자질들을추출하기위한자질추출부, 상기자질추출부에서추출된자질들을바탕으로 SVM(Support Vector Machine)을통해멘션페어의관계를평가하는 SVM 평가부및 상기 SVM 평가부에서서로참조하는멘션페어들로평가된멘션페어들을정리하여엔티티(Entity)로통합하는엔티티집성부를포함한다. 본발명에의하면한국어에대하여문서내에등장하는모든명사또는명사구를고려하고, 그것들에대하여상호참조해결을수행함으로써, 재현율을높일수 있는효과가있다.
-
公开(公告)号：KR1020130021944A
公开(公告)日：2013-03-06
申请号：KR1020110084528
申请日：2011-08-24
Applicant: 한국전자통신연구원
CPC classification number: G06F17/30867 , G06F17/30663
Abstract: PURPOSE: A descriptive question answer method and a device thereof are provided to allow a user to rapidly obtain desired information by automatically providing correct answers for descriptive questions. CONSTITUTION: A core keyword is extracted by analyzing descriptive questions(S302). The questions are classified according to the intention of the questions, and clue words are added to the questions in each type(306). A document related to a linked website document or the collected documents is searched based on the core keyword and the clue word(308). The documents including correct answers are ranked in the searched documents(310). [Reference numerals] (300) Inputting a natural language-based question; (302) Extracting a core keyword and determining the types of the question; (304) Is a clue word added?; (306) Adding the clue word corresponding to the determined type of the question; (308) Searching related document based on the extracted keyword; (310) Re-ranking a document determined a correct answer of the searched documents on the upper rank; (312) Is the correct answer extracted?; (314) Outputting re-ranked information; (AA) Start; (BB) End; (S316) Outputting information classified by the correct answer extraction
Abstract translation: 目的：提供描述性问题回答方法及其设备，以允许用户通过自动提供描述性问题的正确答案来快速获得所需信息。 构成：通过分析描述性问题提取核心关键词（S302）。 这些问题根据问题的意图进行分类，每个类型的问题都添加了线索（306）。 基于核心关键词和线索字搜索与链接的网站文档或收集的文档相关的文档（308）。 包括正确答案的文件在搜索到的文档中排列（310）。 （附图标记）（300）输入基于自然语言的问题; （302）提取核心关键字并确定问题的类型; （304）是否添加了一个线索？ （306）添加与确定的问题类型对应的线索字; （308）基于提取的关键字搜索相关文档; （310）重新排列文件，确定上级搜索文件的正确答案; （312）提取正确的答案？ （314）输出重排信息; （AA）开始; （BB）结束; （S316）按正确答案提取分类的信息输出
-
公开(公告)号：KR1020120064559A
公开(公告)日：2012-06-19
申请号：KR1020100125839
申请日：2010-12-09
Applicant: 한국전자통신연구원
CPC classification number: G06F17/30864 , G06F17/00 , G06F17/20 , G06F17/21 , G06F17/215 , G06F17/218 , G06F17/27 , G06F17/28 , G06F17/2881 , G06F17/30 , G06F17/30654 , G06F17/30657 , G06F19/00
Abstract: PURPOSE: An apparatus and a method for question analysis for open web question-answering are provided to pre-treat about external question, to analyze language, to recognize information of a question, to recognize the type of answer about the question, and extract keyword of the question. CONSTITUTION: Apparatus for question analysis for open web question-answering comprises: a query language analysis unit(110) conducting language analysis of provided question; a query recognition unit(120) recognizing at least one query type recognition result, address information, subject information and query type information from the language-analyzed question; a query answer type recognition unit(130) recognizing the answer type of the question; and a keyword recognition unit(140) extracting keyword from the question, and expanding the extracted keyword.
Abstract translation: 目的：提供一种用于开放式Web问答的问题分析的设备和方法，用于预处理外部问题，分析语言，识别问题的信息，识别问题的答案类型，并提取关键字 的问题。 构成：用于开放式网络问答的问题分析装置，包括：对提供的问题进行语言分析的查询语言分析单元（110） 从语言分析的问题中识别至少一个查询类型识别结果，地址信息，主题信息和查询类型信息的查询识别单元（120） 查询答案类型识别单元（130），识别问题的答案类型; 以及关键词识别单元（140）从所述问题中提取关键词，并扩展所提取的关键字。
-
公开(公告)号：KR1020120061133A
公开(公告)日：2012-06-13
申请号：KR1020100104145
申请日：2010-10-25
Applicant: 한국전자통신연구원
CPC classification number: G06F17/30657
Abstract: PURPOSE: A question type/domain recognition apparatus and a method thereof are provided to maximize the efficiency of character search for correct answer extraction by searching for an answer about user question in a local information restricting technique. CONSTITUTION: A question type recognition apparatus(100) recognizes the number of words and a question for information search. A question domain distributing unit(200) distributes a domain specified question/answer engine into a user question engine. A question/answer engine block(300) configures domain specified question/answer engines in response to distribution of the question domain distributing unit.
Abstract translation: 目的：提供一种问题类型/域识别装置及其方法，以通过在本地信息限制技术中搜索关于用户问题的答案来最大化正确答案提取的字符搜索的效率。 构成：问题类型识别装置（100）识别词的数量和信息搜索的问题。 问题域分发单元（200）将域指定的问题/答案引擎分发到用户问题引擎中。 问题/答案引擎块（300）根据问题域分发单元的分布配置域指定的问题/答案引擎。
-
公开(公告)号：KR1020120037841A
公开(公告)日：2012-04-20
申请号：KR1020100099546
申请日：2010-10-12
Applicant: 한국전자통신연구원
CPC classification number: G06F17/3087 , G06F17/30867 , G06Q30/0201 , H04L67/306
Abstract: PURPOSE: A personalized searching method of a mobile terminal and a mobile terminal thereof are provided to generate a preference analysis model based on usage record of a user and to apply the preference analysis model to an area search result. CONSTITUTION: A mobile terminal obtains location information of the mobile terminal and a query language from a user(210,220). The mobile terminal generates an area search result(230,240). The mobile terminal displays the area search result(250). The mobile terminal stores the usage record of the user(260). The mobile terminal generates a preference analysis model of the user.
Abstract translation: 目的：提供移动终端及其移动终端的个性化搜索方法，以基于用户的使用记录生成偏好分析模型，并将偏好分析模型应用于区域搜索结果。 构成：移动终端从用户获取移动终端的位置信息和查询语言（210,220）。 移动终端生成区域搜索结果（230,240）。 移动终端显示区域搜索结果（250）。 移动终端存储用户的使用记录（260）。 移动终端生成用户的偏好分析模型。
-
Patent Agency Ranking
公开(公告)号：KR1020120036761A
公开(公告)日：2012-04-18
申请号：KR1020110101937
申请日：2011-10-06
Applicant: 한국전자통신연구원
Abstract: PURPOSE: A system and a method for providing a multimedia service in a communication system are provided to accurately search contract information about various multimedia services by transmitting and receiving search protocol and a message. CONSTITUTION: A generating unit(230) generates a query message which searches contract information in a data base and checks contract information from the query result message. The generating unit generates a response message including contract information. A transmitting unit(240) transmits the query message to a database and transmits the response message including the contract information to a terminal.
Abstract translation: 目的：提供一种在通信系统中提供多媒体业务的系统和方法，用于通过发送和接收搜索协议和消息来准确地搜索关于各种多媒体业务的合同信息。 构成：生成单元（230）生成搜索数据库中的合同信息并从查询结果消息检查合同信息的查询消息。 生成单元生成包括合同信息的响应消息。 发送单元（240）将查询消息发送到数据库，并将包括合同信息的响应消息发送到终端。
-
公开(公告)号：KR101095866B1
公开(公告)日：2011-12-21
申请号：KR1020080125436
申请日：2008-12-10
Applicant: 한국전자통신연구원
Abstract: 본 발명은 정보검색, 질의응답, 시맨틱 웹(semantic web) 등에서 튜플(tuple) 또는 트리플(triple) 형태의 정보를 인덱싱(indexing)하여 저장하고 검색하는 기술에 관한 것이다. 이를 위해 본 발명은, 인덱싱의 대상이 되는 문서로부터 형태소 분석, 명사구 청킹(chunking) 등을 수행하고, 언어분석 결과를 입력받아 사람, 장소, 조직 등의 개체명을 인식하며, 문장으로부터 주어와 목적어를 인식하고 주어와 목적어의 관계를 추출하고, 추출된 정보들을 트리플 인덱스에 인코딩하여 저장하며, 사용자의 키워드 또는 자연어질의를 분석하여 개체명 및 관계추출 정보를 분석하는 것을 특징으로 한다. 본 발명으로 인해, 시맨틱 웹 및 정보검색 분야에서 튜플 또는 트리플 형태의 정보를 효율적으로 저장하고 빠르게 검색할 수 있다.
정보검색, 질의응답, 시맨틱 웹, 온톨로지(Ontology), 튜플, 트리플, 색인저장 하부구조-
公开(公告)号：KR101092355B1
公开(公告)日：2011-12-09
申请号：KR1020080131368
申请日：2008-12-22
Applicant: 한국전자통신연구원
Abstract: 이와같이, 본발명은한국어대용어인식에있어서대명사및 명사대용어및 생략된대용어(영형대용어)를포함하여대용어를인식할수 있을뿐만아니라선행어후보생성시에명사구로이루어진선행어와개체명으로이루어진선행어를후보로생성함으로써, 대용어가실제가리키는의미를정확하게파악할수 있다.
-
公开(公告)号：KR1020110064833A
公开(公告)日：2011-06-15
申请号：KR1020090121578
申请日：2009-12-09
Applicant: 한국전자통신연구원
CPC classification number: G06F17/30958
Abstract: PURPOSE: A knowledge graph purifying apparatus and method are provided to eliminate a duplicate entity and simplify the structure of a knowledge graph. CONSTITUTION: A knowledge graph generating unit(110) creates a knowledge graph. In the knowledge graph, the same entities are expressed as one node. An entity similarity calculator(120) calculates meaningful similarity between all possible entity pairs of the same entity type. A same entity integrating unit(130) selects a representative entity in the entity pair. The same entity integrating unit merges the rest entities to the representative entity.
Abstract translation: 目的：提供一种知识图提纯装置和方法，以消除重复的实体并简化知识图的结构。 构成：知识图生成单元（110）创建知识图。 在知识图中，相同的实体表示为一个节点。 实体相似度计算器（120）计算相同实体类型的所有可能实体对之间的有意义的相似性。 相同的实体集成单元（130）选择实体对中的代表实体。 同一实体整合单元将其余实体合并到代表实体。
-
公开(公告)号：KR1020110061229A
公开(公告)日：2011-06-09
申请号：KR1020090117819
申请日：2009-12-01
Applicant: 한국전자통신연구원
CPC classification number: G06F17/2735 , G06F17/278 , G06F17/30731
Abstract: PURPOSE: A system and method for semi-automatic construction of acronym dictionary is provided to improve the language analysis performance by semi-automatically finding an abbreviation word and an original word of the abbreviation word. CONSTITUTION: Some parts of an electronic document are extracted(S210). A proper noun which is included in the electronic document is acknowledged(S220). An abbreviation word about a proper noun candidate is extracted(S230). The abbreviation word candidates are listed in priority order(S240). An abbreviation word is selected(S250). The selected abbreviation word is stored into an abbreviation dictionary database.
Abstract translation: 目的：提供半自动构建首字母缩写词典的系统和方法，通过半自动找到缩写词的缩写词和原始词来提高语言分析性能。 规定：提取电子文档的某些部分（S210）。 电子文档中包含的专有名词被确认（S220）。 提取关于专名候选人的缩写词（S230）。 缩写词候选者以优先顺序列出（S240）。 选择缩写词（S250）。 所选择的缩写词被存储在缩写字典数据库中。
-
-
-
-
-
-
-
-
-
Search Results
130
records
(took 0.025 seconds)
