首页> 外国专利> Korean morphological analysis having linear time complexity using limited neural network and recording medium thereof

Korean morphological analysis having linear time complexity using limited neural network and recording medium thereof

机译:使用有限神经网络的线性时间复杂度的韩国形态分析及其记录介质

摘要

In the Korean morpheme analysis method in the Korean morpheme analysis system based on the sequence-to-sequence model of the present invention, an input sequence is vertically applied by applying an attention mechanism to a sentence in Korean. When the information concentration is stored in the column, the output column is located in the horizontal column, and the attention weight is indicated in the diagonal direction, and the sentence is entered, the position where the attention weight is the largest in the information concentration on the input sentence is stored. Adding a step action to the output string to move the phosphorus focused position to a position increased by one, using a neural network, as much as a window having a predetermined size centered on the concentrated position from the first word of the input sequence in the information concentration degree of the input sentence; Performing a stemming operation only on a range of When the last word operation is performed and a step to end the operation. According to the present invention, even in the morphological analysis of a relatively long sentence, it is effective to reduce the computation time and to implement a fast processing speed.
机译:在基于本发明的序列到序列模型的朝鲜语词素分析系统中的朝鲜语词素分析方法中,通过对朝鲜语的句子施加关注机制来垂直地应用输入序列。当信息集中度存储在该列中时,输出列位于水平列中,并且在对角线方向上显示注意权重,并输入句子,注意权重在信息集中度最大的位置在输入句子上的存储。使用神经网络向输出字符串添加阶跃动作以将磷聚焦位置移动到增加一个的位置,该距离与具有预定大小的窗口(从输入序列中第一个单词的集中位置开始)的集中位置相同输入句子的信息集中度;仅在执行最后一个字操作的时间范围和结束操作的步骤上执行词干操作。根据本发明,即使在相对长的句子的形态分析中,有效的是减少计算时间并实现快速的处理速度。

著录项

  • 公开/公告号KR102049517B1

    专利类型

  • 公开/公告日2019-11-27

    原文格式PDF

  • 申请/专利权人

    申请/专利号KR20180049776

  • 申请日2018-04-30

  • 分类号G06F17/27;G06N3/04;

  • 国家 KR

  • 入库时间 2022-08-21 11:08:27

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