首页> 外国专利> System and Method for Korean Tagging Using the Concatenation of Jamo and Syllable Embedding

System and Method for Korean Tagging Using the Concatenation of Jamo and Syllable Embedding

机译:使用Jamo和音节嵌入的串联进行韩语标记的系统和方法

摘要

The present invention is an apparatus and method for analyzing Korean morphemes using a combination of letter and syllable embeddings that composes input in units of syllables and letters and uses the converted letter embedding to determine the correct part of speech even in frequently occurring typos. Related to, Jamo unit embedding unit for performing elementary, middle, and longitudinal embedding; syllable unit embedding unit for performing syllable unit embedding; combining the three alphabetic embeddings for elementary / middle / vertical, and additionally combining syllable embedding An input unit that expresses a syllable as a vector and provides it as an input of Bi-LSTM-CRF; a learning unit that performs learning using a back propagation algorithm after performing forward / backward steps of Bi-LSTM-CRF; optimal tag sequence Includes a Viterbi search algorithm to find, and an output unit that outputs a part-of-speech tag with symbols indicating the beginning, middle, and end of parts of speech. Is that.
机译:本发明是一种使用字母和音节嵌入的组合来分析韩语语素的装置和方法,该组合以音节和字母为单位构成输入,并使用转换后的字母嵌入来确定语音的正确部分,即使是在经常出现的拼写错误中。与此相关,Jamo单元嵌入单元用于执行基本,中间和纵向嵌入;音节单元嵌入单元,用于进行音节单元嵌入。组合用于基本/中间/垂直的三个字母嵌入,并另外组合音节嵌入一种将音节表示为向量并将其提供为Bi-LSTM-CRF的输入的输入单元;学习单元,在执行Bi-LSTM-CRF的向前/向后步骤之后,使用反向传播算法进行学习;最佳标签序列包括用于查找的维特比搜索算法,以及用于输出词性标签的输出单元,该词性标签带有指示词性的开始,中间和结尾的符号。就是它。

著录项

  • 公开/公告号KR102109858B1

    专利类型

  • 公开/公告日2020-05-12

    原文格式PDF

  • 申请/专利权人 동아대학교 산학협력단;

    申请/专利号KR20180119102

  • 发明设计人 고영중;김혜민;

    申请日2018-10-05

  • 分类号G06F40/20;G06N3/08;

  • 国家 KR

  • 入库时间 2022-08-21 11:04:44

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