System for evaluating relevancy between persons
    1.
    发明专利
    System for evaluating relevancy between persons 审中-公开
    评估人员之间的相关性的系统

    公开(公告)号:JP2007128163A

    公开(公告)日:2007-05-24

    申请号:JP2005318431

    申请日:2005-11-01

    Abstract: PROBLEM TO BE SOLVED: To appropriately determine a relevancy present between persons based on description of a text. SOLUTION: The system for evaluating a relevancy between persons comprises a relevancy detection means detecting, based on a text expressing a certain person's evaluation or emotion for the other person, a relevancy to the other person concerned by the person concerned; a storage means storing the detected relevancy; an evaluation means evaluating, for a set of a first person and second person that is an object of evaluation, the relevancy between the first person and the second person based on a plurality of relevancies extending from the first person to the second person through at least one person, including the stored relevancy; and an output means outputting the evaluated relevancy. COPYRIGHT: (C)2007,JPO&INPIT

    Abstract translation: 要解决的问题:基于对文本的描述,适当地确定人之间存在的相关性。 解决方案:用于评估人员之间的相关性的系统包括相关性检测装置,基于对另一人表达某人的评价或情感的文本,检测与有关的人相关的另一个人的相关性; 存储检测到的相关性的存储装置; 评估装置,对于作为评估对象的第一人和第二人的一组,基于通过至少从第一人延伸到第二人的多个相关性来评估第一人与第二人之间的相关性 一人,包括存储的相关性; 以及输出装置输出评估的相关性。 版权所有(C)2007,JPO&INPIT

    Efficient system for retrieving keyword from text and its method
    3.
    发明专利
    Efficient system for retrieving keyword from text and its method 有权
    从文本及其方法检索关键字的有效系统

    公开(公告)号:JP2007156739A

    公开(公告)日:2007-06-21

    申请号:JP2005349717

    申请日:2005-12-02

    CPC classification number: G06F17/3061 Y10S707/99935

    Abstract: PROBLEM TO BE SOLVED: To more efficiently retrieve a keyword used with high frequency in a text satisfying predetermined conditions than a conventional manner. SOLUTION: This system is stored with a first index for designating the list of keywords included in a text from the identification information of each text, a second index for designating the list of texts including the keywords from the identification information of each keyword and the number of texts including each keyword. Then, when accepting the input of text retrieval conditions, the estimation of a retrieval time based on the first index and the estimation of a retrieval time based on the second index are calculated, and which of the first index and the second index should be used for achieving much quicker retrieval is determined. Then, the keywords appearing with high frequency are retrieved based on the index determined to be used for quicker retrieval in the text satisfying the text retrieval conditions. COPYRIGHT: (C)2007,JPO&INPIT

    Abstract translation: 要解决的问题:为了更有效地检索与常规方式相比满足预定条件的文本中具有高频率的关键字。 解决方案:该系统存储有第一索引,用于从每个文本的标识信息指定包括在文本中的关键字的列表,第二索引,用于从每个关键字的识别信息指定包括关键字的文本列表的第二索引 以及每个关键字的文本数量。 然后,当接受文本检索条件的输入时,计算出基于第一索引的检索时间的估计和基于第二索引的检索时间的估计,并且应当使用第一索引和第二索引中的哪一个 确定实现更快的检索。 然后,基于在满足文本检索条件的文本中确定为用于更快检索的索引来检索出现高频的关键字。 版权所有(C)2007,JPO&INPIT

    Method, apparatus and computer program for obtaining keyword appearance frequency ranking
    4.
    发明专利
    Method, apparatus and computer program for obtaining keyword appearance frequency ranking 审中-公开
    方法,设备和计算机程序,用于获取关键字外观频率排名

    公开(公告)号:JP2013077272A

    公开(公告)日:2013-04-25

    申请号:JP2011218226

    申请日:2011-09-30

    Abstract: PROBLEM TO BE SOLVED: To provide a method, apparatus and computer program for efficiently calculating keyword appearance frequency ranking in distributed processing with a plurality of computers.SOLUTION: Keywords which do not overlap each other are assigned to n computers, where n is a natural number. From the keywords assigned to each computer, keywords of the highest frequency ranks are acquired. From all the acquired keywords, k keywords of the highest overall frequency ranks are selected, where k is a natural number. In doing this, probability P(n, k, t) that the keywords of the highest frequency ranks in each computer are selected as the k highest overall frequency ranks is calculated, where t is a natural number satisfying t

    Abstract translation: 要解决的问题:提供一种用于在多个计算机中有效地计算分布式处理中的关键字出现频率排序的方法,装置和计算机程序。

    解决方案:不重叠的关键字被分配给n台计算机,其中n是自然数。 从分配给每台计算机的关键字获取最高频率排名的关键字。 从所有获取的关键字中,选出最高总体频率秩的k个关键字,其中k是自然数。 在这样做时,计算出每个计算机中排列最高频率的关键字的概率P(n,k,t)作为k个最高总频率行列,其中t是满足t

    Information retrieval system, method, and program
    5.
    发明专利
    Information retrieval system, method, and program 有权
    信息检索系统,方法和程序

    公开(公告)号:JP2009211263A

    公开(公告)日:2009-09-17

    申请号:JP2008051871

    申请日:2008-03-03

    CPC classification number: G06F17/30684 G06F17/30625

    Abstract: PROBLEM TO BE SOLVED: To provide a technology for retrieving documents matching a dependency pattern at a high speed from a large volume of text documents. SOLUTION: An index creation part creates an index for acquiring by sequential access the array of appearance information (document ID, position on tree) of a node even from each word appearing as the node of the tree of a syntax analytic result. A query input part receives a query from a user or an external application. The query is configured of a retrieval pattern, a pivot (node as the reference of retrieval pattern extension), the maximum depth difference in the case of retrieving an extended node form the pivot, the maximum number of extended nodes to be presented in the order of frequency and a flag designating whether to retrieve the high rank node. The index reading part obtains the appearance information array of the pivot at a place matched with the retrieval pattern. The retrieval is performed until it reaches any node connecting the route to the pivot. COPYRIGHT: (C)2009,JPO&INPIT

    Abstract translation: 要解决的问题:提供一种用于从大量文本文档中高速检索匹配依赖模式的文档的技术。 解决方案:索引创建部分创建一个索引,用于通过顺序访问节点数组(文档ID,树上的位置),即使从出现作为语法分析结果树的节点的每个单词获取索引。 查询输入部分接收来自用户或外部应用程序的查询。 该查询由检索模式,枢轴(节点作为检索模式扩展的参考)配置,在从扩展节点形成枢轴的情况下获取最大深度差,按顺序呈现的扩展节点的最大数量 的频率和指定是否检索高等级节点的标志。 索引读取部分在与检索图案匹配的位置处获得枢轴的外观信息数组。 执行检索,直到到达连接到枢轴的路线的任何节点。 版权所有(C)2009,JPO&INPIT

    Expression detection system, expression detection method and program
    6.
    发明专利
    Expression detection system, expression detection method and program 有权
    表达检测系统,表达式检测方法和程序

    公开(公告)号:JP2006146567A

    公开(公告)日:2006-06-08

    申请号:JP2004335906

    申请日:2004-11-19

    Abstract: PROBLEM TO BE SOLVED: To suitably detect liking expression indicating people's liking for a commodity or the like. SOLUTION: An expression detection system for detecting liking expression indicating evaluator's liking to a specific evaluation target from texts in which the evaluation of the specific evaluation target is described stores a plurality of texts in which the evaluation of the specific evaluation target is described corresponding to the attributes of respective texts, extracts the evaluation expression indicting the evaluation of the specific evaluation target from respective texts, judges whether the extracted evaluation expression is positive polarity indicating positive evaluation to the specific evaluation target or negative polarity indicating negative evaluation to the specific evaluation target, inputs the attribute of a text specified as an object for detecting liking expression, detects evaluation expression detected from the text having the inputted attribute as liking expression, and then outputs the liking expression corresponding to the frequency of judgement that the liking expression in the text having the attribute is positive polarity or negative polarity. COPYRIGHT: (C)2006,JPO&NCIPI

    Abstract translation: 要解决的问题:适当地检测表示人们喜欢商品的喜好表达等。 解决方案:用于检测表达评估者对于特定评估目标的评估的喜好表达的表达检测系统,其中描述了特定评估对象的评估的文本存储了描述特定评价对象的评价的多个文本 对应于各文本的属性,从各文本中提取表示具体评价对象的评价的评价表达,判断所提取的评价表达是否为表示对特定评价对象的肯定评价的正极性,或表示对特定评价对象的否定评价的负极性 评估对象,输入被指定为检测喜好表达的对象的文本的属性,检测从具有输入属性的文本中检测到的评价表达作为喜好表达,然后输出与判断的频率相对应的喜好表达, 具有该属性的文本中的表达式为正极性或负极性。 版权所有(C)2006,JPO&NCIPI

    Method, program and system for finding correspondence between terms
    7.
    发明专利
    Method, program and system for finding correspondence between terms 有权
    方法,程序和系统,用于发现条款之间的相关性

    公开(公告)号:JP2012190080A

    公开(公告)日:2012-10-04

    申请号:JP2011050830

    申请日:2011-03-08

    CPC classification number: G06F17/2809

    Abstract: PROBLEM TO BE SOLVED: To provide a technique that makes it possible to list a set of corresponding terms in a different language using a random walk technique with a reasonable calculation amount even when a large number of terms are targeted.SOLUTION: The method comprises the steps of: creating a technical term set of a first language (e.g., Japanese), a general term set of the first language, a technical term set of a second language (e.g., English), and a general term set of the second language; creating a bipartite graph for each of the first and second languages by connecting the technical term set and the general term set of the language with links based on corpus information; creating links between the general terms using a translation dictionary associating general terms between the two languages (e.g., an English-Japanese dictionary) to create a bipartite graph connecting the general term sets of the two languages; creating an association matrix M based on weighted information of links of the bipartite graphs within each language and between the two languages; and calculating a similarity matrix Q defined by Q=(1-c)(I-cM).

    Abstract translation: 要解决的问题:提供一种使得可以使用具有合理计算量的随机游走技术以不同语言列出一组相应术语的技术,即使当针对大量术语时也是如此。 解决方案:该方法包括以下步骤:创建第一语言(例如,日语)的技术术语集,第一语言的通用术语集,第二语言的技术术语集(例如英语), 和第二语言的一般术语集; 通过将技术术语集和语言的一般术语集合与基于语料库信息的链接相连接来为每个第一和第二语言创建二分图; 使用与两种语言之间的一般术语(例如,英语 - 日语词典)相关联的翻译字典在一般术语之间建立链接,以创建连接两种语言的通用术语集合的二分图; 基于每种语言和两种语言之间的二分图的链接的加权信息来创建关联矩阵M; 并且计算由Q =(1-c)(I-cM) 1定义的相似度矩阵Q。 版权所有(C)2013,JPO&INPIT

    Information processor, natural language analysis method, program and recording medium
    8.
    发明专利
    Information processor, natural language analysis method, program and recording medium 有权
    信息处理器,自然语言分析方法,程序和记录介质

    公开(公告)号:JP2012185561A

    公开(公告)日:2012-09-27

    申请号:JP2011046709

    申请日:2011-03-03

    CPC classification number: G06F17/2785 G06F17/30684

    Abstract: PROBLEM TO BE SOLVED: To calculate the matching score of a sentence relative to a query pattern having a modification structure.SOLUTION: An information processor 100 of the invention includes: an input section 110 for obtaining, as input, an analysis object sentence 150, a query pattern 160 and an index value 170 indicating modification easiness between language units in the sentence; and a score calculating section 120 for calculating a matching score indicating to what extent the sentence matches the query pattern by expressing it as a function including at least, as variables, each index value corresponding to each modification relation included in the query pattern 160. The score calculating section 120 tries to associate a partial structure of the query pattern and a range of the sentence while storing the result of partial calculation of the function in a storage area 130 for reuse, and performs calculation recursively in relation to the partial structure and within the range so as to calculate the score.

    Abstract translation: 要解决的问题:计算相对于具有修改结构的查询模式的句子的匹配分数。 解决方案:本发明的信息处理器100包括:输入部分110,用于获得分析对象语句150,查询模式160和表示句子中语言单元之间的修改容易度的索引值170; 以及分数计算部分120,用于计算指示句子与查询模式匹配到多大程度的匹配分数,该表达方式至少包括作为变量,包括与查询模式160中包括的每个修改关系相对应的每个索引值的变量。 分数计算部分120尝试将查询模式的部分结构与句子的范围相关联,同时将功能的部分计算结果存储在存储区域130中以供重用,并且相对于部分结构递归地执行计算,并且内部 范围,以便计算得分。 版权所有(C)2012,JPO&INPIT

    Information processing apparatus, information processing method, and program
    9.
    发明专利
    Information processing apparatus, information processing method, and program 有权
    信息处理设备,信息处理方法和程序

    公开(公告)号:JP2010026689A

    公开(公告)日:2010-02-04

    申请号:JP2008185780

    申请日:2008-07-17

    CPC classification number: G06F17/30

    Abstract: PROBLEM TO BE SOLVED: To provide an information processing apparatus, an information processing method and a program for achieving the pattern matching of text information by saving a memory with high computational efficiency. SOLUTION: The information processing apparatus 10 includes: a generation section 60 configured to generate an ordered tree structure by defining transition edges between nodes using, as transition conditions, respective constraints from one or more constraint patterns 150 each including a plurality of constraints; a search section 72 configured to search for a second substructure K similar to a first substructure P from a root node by determining a set relation between the transition conditions of respective transition edges; and a transition defining section 74 configured to define an additional transition link from a tail node sK n of the second substructure K to a slave node sP n+1 of the first substructure P, the additional transition link adding a constraint to be met by an indeterminant identified from the set relation, wherein a data structure 100 for pattern matching is defined. COPYRIGHT: (C)2010,JPO&INPIT

    Abstract translation: 要解决的问题:提供一种信息处理装置,信息处理方法和程序,用于通过节省具有高计算效率的存储器来实现文本信息的模式匹配。 解决方案:信息处理装置10包括:生成部分60,其被配置为通过使用来自一个或多个约束模式150的各自的约束作为过渡条件来定义节点之间的过渡边缘来生成有序树结构,每个约束模式包括多个约束 ; 搜索部分72,被配置为通过确定各个过渡边缘的转换条件之间的集合关系,从根节点搜索类似于第一子结构P的第二子结构K; 以及转换定义部分74,其被配置为将第二子结构K的尾节点sK n 的附加转变链路定义到第一子结构的从节点sP n + 1 P,附加转移链路添加由从集合关系识别的不确定性要满足的约束,其中定义用于模式匹配的数据结构100。 版权所有(C)2010,JPO&INPIT

    Erzeugungsvorrichtung, Erzeugungsverfahren und Programm

    公开(公告)号:DE102014113988A1

    公开(公告)日:2015-04-30

    申请号:DE102014113988

    申请日:2014-09-26

    Applicant: IBM

    Abstract: Problem Ein abgeändertes Modell zu erzeugen, in dem ein durch Daten, die spezifisch für Trainingsdaten sind, verursachter Vorhersagefehler aus einem aus den Trainingsdaten erzeugten Vorhersagemodell verringert wird. Mittel zum Lösen Es werden eine Erzeugungsvorrichtung, die auf der Grundlage von Trainingsdaten ein Vorhersagemodell erzeugt, ein Erzeugungsverfahren und ein Programm bereitgestellt. Die Erzeugungsvorrichtung beinhaltet eine Modellerzeugungseinheit, die ein trainiertes Modell erzeugt, das durch Trainieren unter Verwenden der Trainingsdaten Ausgabemerkmalwerte aus Eingabemerkmalwerten vorhersagt, und eine Modellabänderungseinheit, die Grade von Auswirkungen einiger der Eingabemerkmalwerte in dem erzeugten trainierten Modell verringert und die ein abgeändertes Modell erzeugt, das tatsächlich für die Testdaten verwendet wird.

Patent Agency Ranking