基于空洞卷积块架构的命名实体识别模型  

Named entity recognition model based on dilated convolutional block architecture

作　　者：袁钺 王艳丽 刘勘[2] YUAN Yue;WANG Yanli;LIU Kan(Department of Information Management,Peking University,Beijing 100871,China;School of Information and Safety Engineering,Zhongnan University of Economics and Law,Wuhan 430073,Hubei,China)

机构地区：[1]北京大学信息管理系,北京100871 [2]中南财经政法大学信息与安全工程学院,湖北武汉430073

出　　处：《山东大学学报（工学版）》2022年第6期105-114,共10页Journal of Shandong University（Engineering Science）

基　　金：中央高校基本科研业务费交叉学科创新研究项目(2722021EK016)。

摘　　要：受到空洞卷积的启发提出面向二维文本嵌入的列式空洞卷积,设计空洞卷积块架构,基于此架构提出命名实体识别模型并开展进一步试验。在命名实体识别试验中,提出的模型的精密度、召回率和F_(1)超越了其他基线模型,分别达到了0.9187、0.8794和0.8986,表明空洞卷积块架构能够获取包含更多上下文信息的文本特征,从而支持模型对上下文长距离依赖特征的捕获和处理。感受野试验表明需要适当调整空洞率以减轻空洞卷积给模型带来的“网格效应”。提出的基于空洞卷积块架构能有效执行命名实体识别任务。Inspired by the dilated convolution,a column-wise dilated convolution towards two dimensional text embedding was proposed and a dilated convolutional block architecture was designed.A named entity recognition model based on the architecture was built for further experiments.In the named entity recognition experiment,the model surpassed other baseline models in the metrics of precision,recall,and F_(1)value,respectively reaching 0.9187,0.8794,and 0.8986,indicating that the dilated convolutional block architecture obtained features from context information,thereby supporting the extraction of the long-term dependency.The receptive field experiment showed that it was necessary to jointly adjust the dilation rate and the convolution kernel size to reduce the“gridding effect”.The dilated convolutional block architecture proposed could effectively perform the task of named entity recognition.

关 键 词：命名实体识别 空洞卷积块架构 感受野 神经网络 深度学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]