首页> 外国专利> INTENTION RECOGNITION METHOD, APPARATUS, AND DEVICE, AND COMPUTER READABLE STORAGE MEDIUM

INTENTION RECOGNITION METHOD, APPARATUS, AND DEVICE, AND COMPUTER READABLE STORAGE MEDIUM

机译:意图识别方法,装置和设备,以及计算机可读存储介质

摘要

An intention recognition method, apparatus, and device, and a computer readable storage medium, relating to the technical field of artificial intelligence. The method comprises: performing word vector feature extraction on an unlabeled text by means of a first language model to obtain an unlabeled feature, and according to the unlabeled feature, annotating the unlabeled text to obtain an annotated training text (S10); constructing an attention neural network model on the basis of a second language model and the annotated training text (S20); obtaining a text to be recognized, and by means of the attention neural network model, performing feature extraction on the text to be recognized, so as to obtain a candidate feature set (S30); and calculating, according to the candidate feature set, the similarity of the text to be recognized, and according to the similarity, determining whether the text to be recognized corresponds to the same expression intention, so as to obtain an intention recognition result (S40). Feature extraction is performed in a neural network mode, and thus, the present invention can more comprehensively consider text features, and improve the accuracy of intention recognition.
机译:一种意图识别方法,装置和设备,以及与人工智能技术领域有关的计算机可读存储介质。该方法包括:通过第一语言模型对未标记的文本执行单词矢量特征提取,以获得未标记的特征,并且根据未标记的特征,注释未标记的文本以获得注释的培训文本(S10);基于第二语言模型和注释培训文本构建关注神经网络模型(S20);获取要识别的文本,并通过注意神经网络模型,在要识别的文本上执行特征提取,以便获得候选功能集(S30);根据候选特征集的计算,要识别的文本的相似性,并且根据相似性,确定要识别的文本是否对应于相同的表达意图,以便获得意图识别结果(S40) 。特征提取以神经网络模式执行,因此,本发明可以更全面地考虑文本特征,并提高意图识别的准确性。

著录项

  • 公开/公告号WO2021143018A1

    专利类型

  • 公开/公告日2021-07-22

    原文格式PDF

  • 申请/专利权人 PING AN TECHNOLOGY (SHENZHEN) CO. LTD.;

    申请/专利号WO2020CN93227

  • 发明设计人 ZENG BO;XU KAIHE;WANG SHAOJUN;

    申请日2020-05-29

  • 分类号G06F40/117;G06F40/194;G06F40/169;G06F40/284;G06F40/30;G06N3/08;

  • 国家 CN

  • 入库时间 2022-08-24 20:09:39

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