-
公开(公告)号：KR101023209B1
公开(公告)日：2011-03-18
申请号：KR1020080099995
申请日：2008-10-13
Applicant: 한국전자통신연구원
IPC: G06F17/28
CPC classification number: G06F17/271 , G06F17/2872
Abstract: 본 발명은 문서 번역 기법에 관한 것으로, 문서 내 텍스트에 대한 형태소를 분석하여 그에 대응하는 태깅을 수행하며, 태깅된 결과에 대한 텍스트 분석을 수행하고 그 분석 결과를 태깅 정보에 반영하여 출력하면, 출력된 문서에 대한 태깅 결과에 대해 구조 분석을 수행하고, 구조 변환을 수행한 후에, 텍스트 분석 정보를 참조하여 대역어를 선택하고, 형태소 생성을 수행하여 한글 문서에 대응하는 각 형태소를 생성하여 그 번역된 문서를 출력함으로써, 영문 문서를 정확하게 번역하여 그 번역 문서를 출력할 수 있고, 번역 문서에 대한 가독성(readability)을 향상시킬 수 있는 것이다.
문서 번역, 형태소 분석, 태깅 정보, 대역어-
公开(公告)号：KR1020110024889A
公开(公告)日：2011-03-09
申请号：KR1020090083069
申请日：2009-09-03
Applicant: 한국전자통신연구원
CPC classification number: G06F17/289 , G06F17/2745
Abstract: PURPOSE: A device and a method for verifying translation knowledge which is specialized in a domain are provided to automatically classify translation knowledge which is most adaptive in a corresponding domain by using parallel corpus per domain. CONSTITUTION: A lemma indexer(106) disperse a plurality of lemma indexed parallel corpus to a plurality of cluster servers. Through constructed translation knowledge, a reference translator automatically translates the lemma indexed parallel corpus. Through translation knowledge of a verification target, an evaluation translator(121) automatically translates the lemma indexed parallel corpus. An auto evaluation unit(118) compares a translation performance difference of a reference translation result(116) and an evaluation translation result(123).
Abstract translation: 目的：提供一种用于验证专业于域的翻译知识的设备和方法，以通过使用每个域的平行语料库自动分类在相应域中最适合的翻译知识。 构成：引理索引器（106）将多个引理索引平行语料库分散到多个群集服务器。 通过构建的翻译知识，参考翻译器自动翻译引文索引的平行语料库。 通过验证目标的翻译知识，评估翻译器（121）自动翻译引文索引平行语料库。 自动评估单元（118）比较参考翻译结果（116）和评估翻译结果（123）的翻译性能差异。
-
43.发明公开
公开(公告)号：KR1020100073164A
公开(公告)日：2010-07-01
申请号：KR1020080131758
申请日：2008-12-22
Applicant: 한국전자통신연구원
IPC: G06F17/40
CPC classification number: G06F17/2765 , G06F17/2755 , G06F17/289 , G06F17/30737
Abstract: PURPOSE: A substitute language specializing device and a method thereof are provided to extract air vocabulary through primitive language corpus and a target language corpus belongings to a target domain and extract substitute language candidate and reflect representative substitute language to substitute dictionary, thereby reducing the cost by manual. CONSTITUTION: A vocabulary arrangement unit(106) extracts co-occurrence vocabulary through primitive language tagged corpus and target language tagged corpus. The vocabulary arrangement unit extracts substitute candidates with mapping of band relation of co-occurrence vocabularies about a substitute dictionary. An error filtering unit(108) determines representative substitute language with filtering of substitute relation error about the extracted substitute candidate. A substitute dictionary reflector(110) reflects the determined representative substitute to the substitute dictionary.
Abstract translation: 目的：提供替代语言专业设备及其方法，通过原语言语料库和目标语言语料库提取空中词汇，并提取替代语言候选词，并反映代表性替代语言来替代词典，从而降低成本 手册。 构成：词汇排列单元（106）通过原语语言标注语料库和目标语言标签语料库提取同现词汇。 词汇排列单元通过对替代词典的同现词汇的频带关系的映射来提取替代候选。 错误过滤单元（108）通过对提取的替代候选者的替代关系错误进行滤波来确定代表性替代语言。 替代字典反射器（110）将确定的代表性替代物反映到替代字典。
-
公开(公告)号：KR1020100073163A
公开(公告)日：2010-07-01
申请号：KR1020080131757
申请日：2008-12-22
Applicant: 한국전자통신연구원
CPC classification number: G06F17/277 , G06F17/2755 , G06F17/2809
Abstract: PURPOSE: A compound noun recognition device and a method thereof are provide to generate a unit noun candidate in an input compound noun of a compound noun recognition device and determine a unit noun, thereby improving recognition about the compound noun. CONSTITUTION: A unit noun extractor(102) extracts a unit noun candidate group from an input compound noun. An affix applying unit(104) performs affix processing about a unit noun of the unit noun candidate group. If a descriptive noun exists among the unit nouns, a meaningful analyzing unit(106) determines meaningful information. According to the affix processing and meaning standing information and meaning restriction conditions, a unit noun determination unit(108) determines a unit noun area by weight applying.
Abstract translation: 目的：提供复合名词识别装置及其方法，以在复合名词识别装置的输入复合名词中生成单位名词候选，并确定单位名词，从而改善对复合名词的认识。 构成：单位名词提取器（102）从输入的复合名词中提取单位名词候选组。 附加应用单元（104）对单位名词候选组的单位名词执行附加处理。 如果在单位名词中存在描述性名词，则有意义的分析单元（106）确定有意义的信息。 根据附加处理和意义站立信息和含义限制条件，单位名词确定单元（108）根据权重应用确定单位名称区域。
-
公开(公告)号：KR1020100069119A
公开(公告)日：2010-06-24
申请号：KR1020080127709
申请日：2008-12-16
Applicant: 한국전자통신연구원
CPC classification number: G06F17/271 , G06F17/2755
Abstract: PURPOSE: A syntax analyzing method and a device thereof are provided to prevent creation of a chart or adjust chart weight by being matched previously constructed constraint condition. CONSTITUTION: A part-of-speech tagger(10) performs part of speech tagging about an original. A syntax information generator(20) generates structure syntax information about the tagging result. A chart parser(30) uses a chart about the morpheme sentence structure information as a default inactive chart, repeats rule matching and active and inactive chart generation, and performs bottom-up chart parsing. A node syntactic information generator(40) generates node syntactic information according to parsing result. A constraint condition application unit(50) applies a constraint condition according to the chart parsing result.
Abstract translation: 目的：提供语法分析方法及其装置，以防止通过匹配先前构造的约束条件来创建图表或调整图表权重。 构成：词性标签（10）执行关于原件的词性标签。 语法信息生成器（20）生成关于标记结果的结构语法信息。 图表解析器（30）使用关于语素句子结构信息的图表作为默认非活动图表，重复规则匹配和主动和非活动图表生成，并执行自下而上的图表解析。 节点句法信息生成器（40）根据解析结果生成节点句法信息。 约束条件应用单元（50）根据图解析结果应用约束条件。
-
Patent Agency Ranking
公开(公告)号：KR1020100031800A
公开(公告)日：2010-03-25
申请号：KR1020080090595
申请日：2008-09-16
Applicant: 한국전자통신연구원
CPC classification number: G06F17/2755 , G06F17/2827 , G06F17/2854
Abstract: PURPOSE: A machine translation error detection method using the parallel corpus and an apparatus thereof are provided so that by providing an error statistics information classified as for each type the reading mistake generating in the machine translation system of the rule base is detected. The performance of the machine translation system is improved. CONSTITUTION: A part of analyzing morpheme(104) analyzes the morpheme of the purpose language sentence interpreted by machine with the purpose language sentence of the parallel corpus corresponding to the primitive language sentence. It classifies into the word by word. A word by word sorter(106) stands in line the group of the purpose language sentence interpreted by machine and primitive language sentence and the group of the purpose language sentence of the primitive language sentence and parallel corpus to the word by word. It decodes. An error detection part(108) detects error through the comparison of the word by word of the decoded sentences as described above from the purpose language sentence interpreted by machine. It classifies error and the error category and performance analysis part(110) inspect the error pattern ratio. Error data is produced.
Abstract translation: 目的：提供使用平行语料库的机器翻译错误检测方法及其装置，通过提供针对每种类型分类的错误统计信息，检测规则库的机器翻译系统中的读取错误。 改进了机器翻译系统的性能。 构成：分析语素的一部分（104）分析机器解释的目的语言句子与原语言句子对应的平行语料库目的语句的语素。 它逐字分类。 一个单词分拣机（106）与机器和原语言句子解释的目的语言句子的组和原语语句的目的语言句子组和平行语料库组合在一起。 它解码 错误检测部分（108）通过如上所述的解码语句逐字从机器解释的目的语句来检测错误。 它对错误进行分类，错误类别和性能分析部分（110）检查错误模式比率。 产生错误数据。
-
公开(公告)号：KR1020100025885A
公开(公告)日：2010-03-10
申请号：KR1020080084626
申请日：2008-08-28
Applicant: 한국전자통신연구원
IPC: G06F17/28
CPC classification number: G06F17/2827 , G06F17/2775
Abstract: PURPOSE: A translation device in which a multi-level verb pattern is applied and an application method and extraction method thereof are provided to improve translating performance by applying a matching method of the multi-level verb pattern. CONSTITUTION: A default verb pattern matching unit(305) advances the matching by applying a semantic code to the default verb pattern through a default verb pattern DB(313) when a basic verb pattern is non-matched. A default word order matching unit(307) advances the matching through a default word order verb pattern DB(315) when the default verb pattern is non-matched. A default investigation matching unit(309) outputs a bilingual document by applying the semantic code to a default investigation pattern through a default investigation DB(317) when the default word order verb pattern is non-matched.
Abstract translation: 目的：应用多级动词模式的匹配方法，提供应用多级动词模式的翻译装置及其应用方法及其提取方法，以提高翻译性能。 构成：当基本动词模式不匹配时，默认动词模式匹配单元（305）通过默认动词模式DB（313）将语义代码应用于默认动词模式来推进匹配。 当默认动词模式不匹配时，默认单词顺序匹配单元（307）通过默认单词顺序动词模式DB（315）推进匹配。 默认调查匹配单元（309）通过默认调查DB（317）将语义代码应用于默认调查模式时输出双语文档，当默认单词顺序动词模式不匹配时。
-
公开(公告)号：KR100911834B1
公开(公告)日：2009-08-13
申请号：KR1020070128041
申请日：2007-12-11
Applicant: 한국전자통신연구원
IPC: G06F17/28
Abstract: 본 발명은 번역 시스템에서 오류 보정 패턴을 이용한 번역 오류 수정 기술에 관한 것으로, 원시언어 문장에 대한 목적언어 형태의 번역 정답 문장들을 포함하는 병렬 코퍼스를 이용한 오류 보정 학습을 통하여 오류 보정 패턴을 구축하고, 자동번역기를 이용하여 특정 문장에 대한 번역을 수행하고, 번역된 문장에 대해 오류 보정 패턴의 조건부와 비교하여 번역 오류를 파악하며, 추정된 번역 오류를 오류 보정 패턴의 행위부에 따라 수정하고, 오류 보정된 번역 문장을 출력하는 것을 특징으로 한다. 본 발명에 의하면, 비문이나 자연스럽지 못한 표현 등과 같은 자동 번역 시스템의 번역 오류를 실시간으로 보정함으로써, 자동 번역 시스템의 번역 성능을 향상시킬 수 있다.
자동 번역 시스템, 자동 번역 후처리, 오류보정 패턴-
公开(公告)号：KR101626386B1
公开(公告)日：2016-06-02
申请号：KR1020090123136
申请日：2009-12-11
Applicant: 한국전자통신연구원
Abstract: 본발명은자동번역시스템에서번역한번역문장내에나타나는번역오류를후처리보정장치를이용하여자동으로보정하여번역문장내의번역오류를수정함으로써번역품질을향상시키는번역오류후처리보정장치에관한것으로서, 특히목적언어코퍼스로부터번역오류보정을위한요소화언어모델을구축하고, 이에기반하여번역오류를탐색및 수정하는오류보정방법에관한것이다. 본발명은번역오류에대해후처리보정을함에있어서오류추정및 오류보정을위한기본지식으로사용되는요소화언어모델을구축하고이에기반하여단어신뢰도값을계산하고, 올바른오류보정성능을높임으로써, 오류가있는자동번역시스템의번역오류를수정하여보다높은품질의번역을수행할수 있다.
-
50.发明公开중국어 자동 번역을 위한 생략 성분 복원 방법 및 장치, 이를 포함하는 중국어 자동 번역 방법 및 장치 无效用于中国机器翻译的中国机器翻译的方法和装置，用于包含它们的中文机翻译的方法和装置
公开(公告)号：KR1020130102926A
公开(公告)日：2013-09-23
申请号：KR1020120024128
申请日：2012-03-08
Applicant: 한국전자통신연구원
CPC classification number: G06F17/2863 , G06F17/2755
Abstract: PURPOSE: An omitted component restoration method and an apparatus for Chinese machine translation, and a Chinese machine translation method and an apparatus including the same are provided to classify an omitted expression included in a Chinese dialogue sentence, and to restore the omitted expression according an omission type, thereby providing an accurate auto-translation result. CONSTITUTION: An input Chinese dialogue sentence is divided into work units, and the property is extracted based on morpheme analysis information with respect to the division result and the word (S1100). An omission type of the input Chinese dialogue sentence is estimated based on the extracted property (S1200). The omitted portion of the input Chinese dialogue sentence is restored according to the omission type (S1300). The estimation probability of the omitted component is improved by using the morpheme analysis information with respect to the word, and the morpheme analysis information includes the part of speech of the word and definition information. [Reference numerals] (AA) Start; (BB) End; (S1000) Divide an input Chinese dialogue into word units; (S1100) Extract the properties to determine whether there is any element deleted from the input Chinese dialogue sentence and to estimate the omission type based on the result of dividing the sentence into word units and morpheme analysis information with respect to each word; (S1200) Estimate an omission type of the input Chinese dialogue sentence based on the extracted property; (S1300) Restore the deleted element based on the estimated omission type
Abstract translation: 目的：提供一种省略的部件恢复方法和中文机器翻译装置，以及一种中文机器翻译方法及其装置，用于对包含在中文对话句中的省略表达进行分类，并根据省略恢复省略的表达 类型，从而提供准确的自动翻译结果。 构成：输入中文对话句分为工作单位，根据分词结果和单词的语素分析信息提取属性（S1100）。 基于所提取的属性来估计输入的中文对话句的省略类型（S1200）。 输入中文对话句的省略部分根据省略类型进行恢复（S1300）。 通过使用关于该词的语素分析信息来提高省略的分量的估计概率，并且语素分析信息包括单词和定义信息的部分语音。 （附图标记）（AA）开始; （BB）结束; （S1000）将输入的中文对话划分为字单位; （S1100）提取属性，以确定是否从输入的中文对话句中删除任何元素，并根据将句子分为单词和词素分析信息的结果来估计遗漏类型; （S1200）基于提取的属性估计输入的中文对话句的省略类型; （S1300）根据估计的省略类型恢复已删除的元素
-
-
-
-
-
-
-
-
-
Search Results
65
records
(took 0.024 seconds)
