首页> 外国专利> Paraphrase expression acquisition system, paraphrase expression acquisition method, and paraphrase expression acquisition program

Paraphrase expression acquisition system, paraphrase expression acquisition method, and paraphrase expression acquisition program

机译:复述表达获取系统,复述表达获取方法和复述表达获取程序

摘要

PPROBLEM TO BE SOLVED: To acquire paraphrasing expression speaking the same meaning content by different expression from a document group without needing syntax analysis and without giving some example in a specific relation in advance. PSOLUTION: A cooccurrence word pair context collection part 12 collects contexts containing a pair of optional cooccurrence words from the document group housed in a document group DB1 and stores individual contexts in a cooccurrence word pair context DB2 by each pair of the cooccurrence words. A context vector generation part 14 obtains the word frequency of words constituting the individual contexts corresponding to each pair of cooccurrence words, calculates a weight, and stores a context vector in a context vector DB4. A context vector similarity calculation part 15 obtains all the similarities between two context vectors. A cooccurrence word pair clustering part 16 clusters a pair of cooccurrence words of high similarity between the context vectors. A relation label acquiring part 17 acquires words expressing respective clusters, and an intra-cluster context selection part 18 selects a context containing the word as a paraphrased expression from the DB2. PCOPYRIGHT: (C)2006,JPO&NCIPI
机译:

要解决的问题:从文档组中以不同的表达获取具有相同含义内容的释义表达,而无需语法分析,也无需事先给出特定关系的示例。

解决方案:同现词对上下文收集部分12从文档组DB1中容纳的文档组中收集包含一对可选的同现词的上下文,并按每对同现词将各个上下文存储在同现词对上下文DB2中。上下文向量生成部14获得构成与每对同现词对相对应的各个上下文的词的词频,计算权重,并将上下文向量存储在上下文向量DB4中。上下文向量相似度计算部分15获得两个上下文向量之间的所有相似度。同现词对聚类部分16将上下文向量之间的高度相似的一对同现词聚类。关系标签获取部分17获取表达各个聚类的词,并且集群内上下文选择部分18从DB2中选择包含该词的上下文作为释义表达。

版权:(C)2006,JPO&NCIPI

著录项

  • 公开/公告号JP4252038B2

    专利类型

  • 公开/公告日2009-04-08

    原文格式PDF

  • 申请/专利权人 日本電信電話株式会社;

    申请/专利号JP20050002366

  • 发明设计人 長谷川 隆明;

    申请日2005-01-07

  • 分类号G06F17/28;

  • 国家 JP

  • 入库时间 2022-08-21 19:37:29

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号