支持汽车故障数据增值的词汇增强实体识别  

Vocabulary-Enhanced Entity Recognition that Supports the Value-Added of Automobile Fault Data

作　　者：杨雯迪 任春华 孙洁香 Yang Wendi;Ren Chunhua;Sun Jiexiang(Department of Information Science and Technology,Southwest Jiaotong University,Chengdu 611756;Beijing Machinery Industry Automation Research Institute Co.,Ltd.,Beijing 100120)

机构地区：[1]西南交通大学信息科学与技术学院,成都611756 [2]北京机械工业自动化研究所有限公司,北京100120

出　　处：《现代计算机》2021年第26期8-14,共7页Modern Computer

基　　金：国家重点研发计划:区域综合科技服务工程应用集成技术研究(2017YFB1401400)。

摘　　要：在汽车故障文本知识提取中,由于语料不规范和领域的专业性,缺乏特定领域的命名实体识别模型。为弥补基于字符的命名实体识别模型没有考虑专业领域词汇的重要性,提出了一种支持汽车故障数据增值的词汇增强实体识别方法。首先利用基于词典的双层标注序列方式弥补了专业领域数据标注难问题,接着利用加权的方式将词汇信息融入到字符向量中,最终实现了词汇增强的命名实体识别模型。In the extraction of automobile fault text knowledge,due to the non-standard corpus and the professionalism of the field,there is a lack of named entity recognition models in specific fields.To make up for the character-based named entity recognition model that does not consider the importance of vocabulary in the professional field,a vocabulary-enhanced entity recognition method that supports the value-added of automobile fault data is proposed.This method first uses a dictionary-based double-layer tagging se⁃quence method to make up for the problem of data tagging in the professional field,then uses a weighted method to integrate vocabulary information into a character-based vector,and finally implements a vocabulary-enhanced named entity recognition model.

关 键 词：汽车故障数据 命名实体识别 领域词典 预训练语言模型 BiLSTM_CRF模型 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术] U472[自动化与计算机技术—计算机科学与技术]