首页> 外国专利> Dialogue destruction feature extraction device, dialogue destruction feature extraction method, program

Dialogue destruction feature extraction device, dialogue destruction feature extraction method, program

机译:对话破坏特征提取装置,对话破坏特征提取方法,程序

摘要

PROBLEM TO BE SOLVED: To provide a technique for extracting a dialogue destruction feature quantity that can be used for estimating a dialogue destructive force indicating an extent to which an utterance causes a dialogue to fail.SOLUTION: A dialog destruction feature quantity extraction apparatus includes a dialog destruction feature quantity extraction unit 110 configured to extract a dialog destruction feature quantity that is a combination of j-th type feature quantity (1≤j≤J), which is a feature quantity indicating the feature as to whether the dialog is destructive, from a dialog consisting of a series of utterances. The dialog destruction feature quantity extraction unit 110 includes a j-th type feature quantity calculation unit configured to calculate a j-th type feature quantity from the dialog for each j satisfying 1≤j≤J. The dialog destruction feature quantity is the combination including at least one of frequently occurring word string feature quantity, inter-utterance similarity feature quantity, sentence length feature quantity, turn number feature quantity, word vector feature quantity, question type feature quantity, question class feature quantity, and topic iteration number feature quantity.SELECTED DRAWING: Figure 1
机译:解决的问题:提供一种提取对话破坏特征量的技术,该技术可用于估计表示话语导致对话失败的程度的对话破坏力。对话破坏特征量提取单元110,被配置为提取对话破坏特征量,该对话破坏特征量是表示所述对话是否具有破坏性的特征的特征量的第j型特征量(1≤j≤J)的组合,包含一系列话语的对话框中对话破坏特征量提取单元110包括第j类型特征量计算单元,该第j类型特征量计算单元从对话中为满足1≤j≤J的每个j计算第j类型特征量。对话破坏特征量是包括以下各项中的至少一项的组合:频繁出现的字串特征量,话语间相似性特征量,句子长度特征量,转数特征量,词向量特征量,问题类型特征量,问题类别特征数量和主题迭代编号特征数量。选定的图纸:图1

著录项

  • 公开/公告号JP6630304B2

    专利类型

  • 公开/公告日2020-01-15

    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JP20170042710

  • 发明设计人 杉山 弘晃;

    申请日2017-03-07

  • 分类号G06F16/90;G06F40/30;G06F40/216;

  • 国家 JP

  • 入库时间 2022-08-21 11:33:43

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