投稿入口

基于双向GRU与注意力机制的调度操作票审核方法

A method of examine dispatching operation ticket based on bidirectional GRU and attention mechanism

  • 摘要: 现有操作票审核技术主要依靠分词后形成规则库,从而依靠规则库对操作票进行审核,忽略了句子信息。本文通过研究一种操作票审核方法,基于双向GRU与注意力机制,首先收集并扩展数据集形成输入集,其次通过词向量生成及训练生成项目审核模型,再次通过句向量生成及训练生成任务审核模型,最后将待审核操作票放入模型中进行处理。通过对词向量、句向量信息的结合引入,能同时覆盖调度操作票词语、句子、句间的审核,扩大了审核范围,提升了审核准确率。

     

    Abstract: The prior art mainly relies on the mode of forming a rule base after word segmentation to examine the operation ticket, ignoring the sentence information. By studying an operation ticket examine method, based on the bidirectional GRU and attention mechanism. Firstly, collect and expand the data set to form the input set, Secondly, generate the project audit model through word vector generation and training, Thirdly, generate the task audit model through sentence vector generation and training, Finally, put the operation ticket to be audited into the model for processing. Through the combination and introduction of word vector and sentence vector information, it can cover the words, sentences and sentences in the dispatching operation ticket in the same time, expand the scope and improve the accuracy.

     

/

返回文章
返回