-
公开(公告)号:KR100331031B1
公开(公告)日:2002-05-09
申请号:KR1019980052181
申请日:1998-12-01
Applicant: 한국전자통신연구원
IPC: G06F17/27
Abstract: 본 발명은 부분 파싱을 이용한 확률 기반 품사 태거의 오류를 방지하기 위한 장치 및 그 방법을 제공하는데 그 목적이 있다.
본 발명에 따르면, 자연 언어 처리 시스템의 확률 기반 품사 태거(Tagger)의 오류 방지 시스템에 있어서, 태깅을 하기 위하여 입력된 단어에 대하여 각각의 품사와 해당 확률 정보를 가지고, 확률적 태거의 입력 구조를 생성하기 위한 패턴 생성 수단과; 상기 입력된 단어가 문맥에 따라 가질 수 있는 가변적인 확률 정보가 이미 설정된 확률 보정 규칙을 이용하여, 문장의 문맥을 인식하기 위한 규칙을 생성하기 위한 규칙 컴파일러 수단과; 상기 규칙 컴파일러 수단에 의하여 생성된 규칙이 입력되는 문장을 구성하는 단어에 적용 가능한지를 판단하기 위하여 부분 파싱(Partial Parsing)을 수행하여, 적용이 가능하면, 기설정된 확률 보정 규칙에 따라 확률값을 보정하기 위한 확률 보정 수단과; 상기 확률 보정 수단에 의하여 보정된 확률값에 따라, 확률 기반 태거를 수행하여, 상기 입력된 문장을 구성하는 단어의 품사를 결정하는 태거 수단을 포함하여 이루어진 것을 특징으로 하는 확률 기반 품사 태거(Tagger)의 오류 방지 시스템이 제공된다.-
公开(公告)号:KR100311224B1
公开(公告)日:2001-10-12
申请号:KR1019990061885
申请日:1999-12-24
Applicant: 한국전자통신연구원
IPC: G06F17/28
Abstract: 본발명은영어용언구의연결정보결정장치및 그방법에관한것으로, 한국어문장의구문구조정보를이용하여모든용언구의조합들을추출하는제 1 단계; 상기추출된용언구의연결관계조합에대하여연결패턴리스트를추출하는제 2 단계; 상기추출된연결패턴이한국어연결패턴조건을만족하는가를판단하는제 3 단계; 상기제 3 단계의판단결과에따라, 상기연결패턴데이터베이스에서기본연결패턴을선택하는제 4 단계; 상기제 3 단계의판단결과에따라, 해당연결패턴을선택하는제 5 단계; 상기추출된모든연결관계조합에대하여적용할연결패턴이모두선택되었는지를판단하는제 6 단계; 상기제 6 단계의판단결과에따라, 연결패턴집합선택기가선택된연결패턴들의리스트중에서영어접속정보를모두제공할수 있는가장적합한연결패턴의집합을결정하는제 7 단계; 상기제 6 단계의판단결과에따라, 상기제 2 단계로넘어가는제 8 단계; 및영어생성정보반영기에서상기연결패턴집합의연결정보를이용하여영어용언구의접속정보및 생성정보를제공하는제 9 단계를포함하며, 한영자동번역시스템등에이용됨.
-
公开(公告)号:KR1020010063834A
公开(公告)日:2001-07-09
申请号:KR1019990061935
申请日:1999-12-24
Applicant: 한국전자통신연구원
IPC: G06F17/28
Abstract: PURPOSE: An automatically translating device and method using verbal-based door frame is provided to construct and use a large quantity of verbal phrase door frames and the opponent language bilingual door frame as an actual sentence-objected. CONSTITUTION: Morphemes of the original sentence(101) received through an input device(101) are analyzed with respect to each work through a morpheme analyzing unit(102). A part of speech is decided in a speech part deciding unit(103). A fixed compression recognition unit(104) bundles words and attaches the corresponding new part of speech. A verbal phrase discovery unit(105) discovers verbal phrases displayed in a sentence and makes the verbal phrases as a base of a door frame search. A door frame searching unit(106) searches all door frames being included in the verbal phrase using the discovered verbal phrase as a searching key. A sentence structure deciding unit(107) creates a combination of all door frames being included in the verbal phrase with all marks being indicated in an actual sentence and decides the optimum input sentence structure. A bilingual door frame combination unit(108) adapts the verbal phrase door frame bilingual door frame to a word order and the grammar of the opponent language. A bilingual language creating unit(109) creates one completed opponent language bilingual sentence by performing a grammar process.
Abstract translation: 目的:提供一种使用基于口头门框的自动翻译装置和方法,以构建和使用大量口头语句门框和对等语言双语门框作为实际句子对象。 构成:通过语素分析单元(102),通过输入装置(101)接收的原始句子(101)的语素相对于每个作品进行分析。 在语音部分决定单元(103)中决定一部分语音。 固定压缩识别单元(104)捆绑词并附加对应的新的词性。 语言短语发现单元(105)发现句子中显示的语言短语,并使口头短语成为门框搜索的基础。 门框搜索单元(106)使用所发现的口头短语作为搜索关键字搜索包含在语言短语中的所有门框。 句子结构决定单元(107)创建包括在语言短语中的所有门框的组合,其中所有标记在实际句子中指示,并且确定最佳输入句子结构。 双语门框组合单元(108)将口语短语门框双语门框调整为单词顺序和对手语言的语法。 双语语言创建单元(109)通过执行语法处理来创建一个完成的对手语言双语句子。
-
公开(公告)号:KR1020010027882A
公开(公告)日:2001-04-06
申请号:KR1019990039860
申请日:1999-09-16
Applicant: 한국전자통신연구원
IPC: G06F17/28
Abstract: PURPOSE: An apparatus for discriminating idiom of phrase unit based target sentence frame and method thereof is provided to generate the target sentence by discriminating and processing the idiom of phrase unit included in original sentence. CONSTITUTION: An apparatus for discriminating idiom of phrase unit based target sentence frame has a controller(110), a target sentence frame constructing section(120), an idiom of phrase unit discriminating section(130) and a target word outputting section(140). The controller(110) controls operations on the basis of kind of an input signal. The target sentence frame constructing section(120) processes candidates of the target sentence frame and constructs a database of the target sentence frame and idiom of phrase unit. The idiom of phrase unit discriminating section(130) selects the target sentence frame corresponding to inputted original sentence by searching the database. The target word outputting section(140) generates and outputs the target word for the selected target sentence frame.
Abstract translation: 目的:提供一种用于识别基于短语单位的目标句子及其方法的设备,以通过区分和处理原始句子中包含的短语单元的成语来产生目标句子。 构成:用于区分基于短语单位的目标句子帧的成语的装置具有控制器(110),目标句子构建部分(120),短语单元识别部分(130)和目标词输出部分(140)的成语, 。 控制器(110)基于输入信号的种类来控制操作。 目标句子构成部(120)处理目标句子帧的候补,构成目标句子的数据库和短语单位的成语。 短语单元识别部分(130)的成语通过搜索数据库来选择与输入的原始句子相对应的目标句子帧。 目标字输出部(140)生成并输出所选择的目标句子帧的目标字。
-
公开(公告)号:KR1020010003382A
公开(公告)日:2001-01-15
申请号:KR1019990023662
申请日:1999-06-23
Applicant: 한국전자통신연구원
IPC: G06F17/30
Abstract: PURPOSE: A web browser for real-time automatic translation is provided to minimize setting conditions of a user by automatically translating a web immediately after finishing transferring a homepage, to improve incompleteness by adopting a DDE(dynamic data exchange) and to shorten time required to receive and translate the homepage by using a clipboard method. CONSTITUTION: A web browser(10) is connected to an internet(13) through a network controller(12) having a network card(11). The web browser displays an input command on a monitor by processing the command inputted through an input/output controller(14) such as a keyboard or monitor. The network controller transfers data by interfacing with the network card. The input/output controller performs input/output with a user from the input/output device such as the monitor and keyboard. The web browser performs automatic translation in real time under a window OS(operating system). The web browser consists of an internet explorer controller(101), a browser manager(102), a web robot device(103) and an automatic translator(104).
Abstract translation: 目的:提供用于实时自动翻译的网络浏览器,以通过在完成转移主页之后立即自动翻译网页,通过采用DDE(动态数据交换)来提高不完整性,并缩短所需时间,从而最小化用户的设置条件 使用剪贴板方式接收并翻译主页。 构成:网络浏览器(10)通过具有网卡(11)的网络控制器(12)连接到互联网(13)。 网络浏览器通过处理通过诸如键盘或监视器的输入/输出控制器(14)输入的命令,在监视器上显示输入命令。 网络控制器通过与网卡的接口传输数据。 输入/输出控制器从输入/输出设备(如显示器和键盘)与用户进行输入/输出。 Web浏览器在窗口OS(操作系统)下实时自动翻译。 网络浏览器由互联网浏览器控制器(101),浏览器管理器(102),网页机器人装置(103)和自动翻译器(104)组成。
-
公开(公告)号:KR100377475B1
公开(公告)日:2003-03-26
申请号:KR1019990029947
申请日:1999-07-23
Applicant: 한국전자통신연구원
IPC: G09B5/00
Abstract: PURPOSE: An apparatus is provided to automatically extract an information about a Korean discourse form using a mood information of a verb, a sentence structure information, and an analysis information of a preceding sentence, thereby extracting an information for helping the comprehension of the discourse. CONSTITUTION: An apparatus for extracting a Korean discourse form using a mood information includes a context restoration device(2), a proposition extracting device(3) of a Korean writing, a discourse response extracting device(5), and a discourse form extracting device(6). The context restoration device discriminates an elliptical sentence, an inversion sentence, and a substitute sentence by receiving a Korean sentence and a meaning code to restore a lack information. The proposition extracting device of a Korean writing extracts a proposition in logical expression using the Korean sentence and the meaning code. The discourse response extracting device extracts the mood information represented in an ending of a verb, and extracts a discourse response using the Korean sentence, the meaning code, and the discourse code. The discourse form extracting device extracts the discourse form using the proposition and the discourse response.
Abstract translation: 目的:提供一种装置,用于使用动词的情绪信息,句子结构信息和前一句话的分析信息自动提取关于韩语语篇形式的信息,由此提取用于帮助理解话语的信息。 本发明提供了一种使用情绪信息提取韩语语篇形式的装置,包括:上下文恢复装置(2),韩文写作命题提取装置(3),话语反应提取装置(5)和话语形式提取装置 (6)。 上下文恢复装置通过接收朝鲜语句子和含义代码来恢复不足信息来区分椭圆句子,反转句子和替代句子。 韩语写作的命题提取装置使用韩语句子和含义代码以逻辑表达提取命题。 话语反应提取装置提取在动词末尾表示的情绪信息,并且使用韩语句子,含义代码和话语代码提取话语应答。 话语形式提取装置使用命题和话语响应来提取话语形式。
-
公开(公告)号:KR1020010057775A
公开(公告)日:2001-07-05
申请号:KR1019990061182
申请日:1999-12-23
Applicant: 한국전자통신연구원
IPC: G06F17/28
Abstract: PURPOSE: A system and method for automatically performing a translation based on a sentence frame is provided to naturally generate a translated sentence by removing a structural ambiguity on the basis of a sentence frame, which represents a frame of a sentence. CONSTITUTION: A source language morpheme analyzer(20) analyzes a morpheme of a source language sentence, and detects information on parts of speech of each word. A translation dictionary(21) stores a source vocabulary, a substitutive word of the vocabulary and quality information related to the word. A phrase unit syntax analyzer(22) represents an input sentence using the information detected from the source language morpheme analyzer(20). An original frame searcher(23) searches a constraint. An original frame database(24) stores a basic constraint of a relevant original frame. A sentence frame converter(25) converts the original into morpheme columns of an object language. A translation sentence frame database(26) stores constraint information of an original slot and designated information of a translation sentence slot. A slot translation database(27) stores constraint information of a part of speech of the original and designated information of a part of speech of the translation. An object language morpheme generator(28) analyzes the morpheme of the object language.
Abstract translation: 目的:提供一种用于基于句子帧自动执行翻译的系统和方法,以通过基于代表句子框架的句子框架去除结构歧义来自然地生成翻译句子。 构成:源语言语素分析器(20)分析源语句的语素,并检测每个词的词性信息。 翻译词典(21)存储源词汇,词汇的替代词和与该词相关的质量信息。 短语单元语法分析器(22)表示使用从源语言语素分析器(20)检测的信息的输入语句。 原始帧搜索器(23)搜索约束。 原始帧数据库(24)存储相关原始帧的基本约束。 句子转换器(25)将原件转换为对象语言的语素列。 翻译句框数据库存储原始时隙的约束信息和翻译句时隙的指定信息。 时隙转换数据库(27)存储翻译的一部分语音的原始和指定信息的部分语音的约束信息。 对象语言语素生成器(28)分析对象语言的语素。
-
公开(公告)号:KR1020010057763A
公开(公告)日:2001-07-05
申请号:KR1019990061164
申请日:1999-12-23
Applicant: 한국전자통신연구원
IPC: G06F17/28
Abstract: PURPOSE: An apparatus and method for generating a translated sentence based on a partial translation pattern database is provided to generate a high-quality translated sentence at high speed by processing a language phenomenon, which is variously translated on the basis of a declinable word using a partial translation pattern database. CONSTITUTION: A partial syntax as the unit of a declinable word is detected from a syntax analysis result of a resource sentence. A partial translation pattern database is searched according to a key value. A partial translation pattern is detected through a matching with each component of the resource sentence inputted using a meaning category database. The optimal translation pattern is selected according to a priority among a plurality of candidate partial translation patterns detected. A partial translation sentence is generated according to a matching form of the translation pattern selected. A position between each adverb and a position in an input sentence are determined using generation information of an adverb. The adverb is added to the partial translation sentence using an adverb translation dictionary database. The generation information is processed. The final object language translated sentence is generated.
Abstract translation: 目的:提供一种基于部分翻译模式数据库生成翻译句子的装置和方法,用于通过处理语言现象以高速度生成高质量的翻译句子,该语言现象基于使用 部分翻译模式数据库。 构成:从资源句子的语法分析结果中检测出作为可下降词的单位的部分语法。 根据键值搜索部分翻译模式数据库。 通过与使用意义类别数据库输入的资源语句的每个组件的匹配来检测部分翻译模式。 根据检测到的多个候选部分转换模式中的优先级来选择最佳转换模式。 根据所选择的翻译模式的匹配形式生成部分翻译句子。 使用副词的生成信息来确定每个副词和输入句中的位置之间的位置。 使用副词翻译字典数据库将副词添加到部分翻译句子中。 生成信息被处理。 生成最终的对象语言翻译句子。
-
公开(公告)号:KR1020010010841A
公开(公告)日:2001-02-15
申请号:KR1019990029947
申请日:1999-07-23
Applicant: 한국전자통신연구원
IPC: G09B5/00
CPC classification number: G10L15/1822 , G06F17/2785
Abstract: PURPOSE: An apparatus is provided to automatically extract an information about a Korean discourse form using a mood information of a verb, a sentence structure information, and an analysis information of a preceding sentence, thereby extracting an information for helping the comprehension of the discourse. CONSTITUTION: An apparatus for extracting a Korean discourse form using a mood information includes a context restoration device(2), a proposition extracting device(3) of a Korean writing, a discourse response extracting device(5), and a discourse form extracting device(6). The context restoration device discriminates an elliptical sentence, an inversion sentence, and a substitute sentence by receiving a Korean sentence and a meaning code to restore a lack information. The proposition extracting device of a Korean writing extracts a proposition in logical expression using the Korean sentence and the meaning code. The discourse response extracting device extracts the mood information represented in an ending of a verb, and extracts a discourse response using the Korean sentence, the meaning code, and the discourse code. The discourse form extracting device extracts the discourse form using the proposition and the discourse response.
Abstract translation: 目的:提供一种装置,使用动词的情绪信息,句子结构信息和前一句的分析信息来自动提取关于韩语话语形式的信息,从而提取用于帮助理解话语的信息。 构成:使用情绪信息提取韩语话语表格的装置包括上下文恢复装置(2),韩语写作的命题提取装置(3),话语响应提取装置(5)和话语表单提取装置 (6)。 上下文恢复装置通过接收韩语句子和意义码来识别缺少信息的椭圆句子,翻译句子和替代句子。 韩国文字的提法提取装置使用韩语句子和含义代码以逻辑表达方式提取命题。 话语响应提取装置提取动词结尾中表示的情绪信息,并使用韩语句子,含义代码和话语代码提取话语响应。 话语形式提取装置使用命题和话语响应提取话语形式。
-
公开(公告)号:KR1019990047858A
公开(公告)日:1999-07-05
申请号:KR1019970066399
申请日:1997-12-05
Applicant: 한국전자통신연구원
IPC: G06Q50/00
Abstract: 본 발명은 모니터에 나타난 문자열에 대한 언어지식을 서비스하는 전자사전 시스템에 관한 것으로서, 특히, 사용자가 지시하는 임의의 위치에 있는 문자열에 내포된 언어적 구성요소와 그 구성요소들로 조합될 수 있는 언어지식을 서비스하는 언어분석에 기반한 전자사전 서비스 방법 및 그 시스템에 관한 것이다.
본 발명은, 어절단위나 연속된 문자열속에서 임의의 위치에 있는 구성요소나 구성요소들의 결합로 조합된 언어지식 획득을 목적으로 요구하는 서비스 선택사항과, 서비스요구 위치를 인식하고 그 위치의 전후에 나열된 문자열을 자동추출하여 기억하거나, 문자열 입력에 의한 서비스를 요구하는 문자열을 기억하는 제 1과정과, 언어 특성상 분석 범위결정에 필요한 지식을 갖은 매체상에 기록된 '언어별분석범위결정지식'과 '언어분석용지식'정보를 이용하여 분석대상문자열의 범위를 연산하고 대상문자열을 추출하는 제 2과정과, 제 2과정에서 추출한 문자열을 그 언어의 분석기술로 분석하는 제 3과정과, 분석함으로서 얻어지는 문자열에 내포된 언어의 구성요소와 그 구성요소들의 결합으로 가능한 언어적 현상들을 검색하여 사용자의 서비스 선택� �항에 따라 출력하는 제 4과정을 포함하여, 언어 미숙자에게 언어정보 습득의 편의성과 보다 유용한 언어지식을 제공할 수 있게 한다.
-
-
-
-
-
-
-
-
-