网络评论文本数据监管处理的优化研究--以茶产品为例  

Research on Optimization of Supervision and Processing of Web Comment Text Data--Taking Tea Products as an Example

作　　者：王书博 程贞敏 苏渝 Wang Shubo;Cheng Zhenmin;Su Yu(School of Mathematics and Statistics,Guizhou University,Guiyang 550025;School of Management,Guizhou University,Guiyang 550025)

机构地区：[1]贵州大学数学与统计学院,贵阳550025 [2]贵州大学管理学院,贵阳550025

出　　处：《情报杂志》2022年第5期118-123,189,共7页Journal of Intelligence

基　　金：贵州省社会科学基金项目“新时代贵州培育良好营商环境研究”(编号:19GZYB71)研究成果之一;贵州省教育厅基金项目“新时代下贵州省企业家精神培育的研究”(编号:2018GH02)研究成果之一。

摘　　要：[研究目的]通过拓展和强化文本,并提出基于Bert的改进模型,以期在对网络评论文本数据进行监督处理时获得更好的精度和效度。[研究方法]该文以茶产品的的网络评论文本为例,运用Word2Vec进行文本的深度学习,将当前经常使用的监管方法设为对照组,将提出的Bert+Transformer模型和Bert+XGB模型设为实验组,进行实证研究。[研究结论]结果表明,Bert+Transformer模型和Bert+XGB模型比当前已经应用的监管手段更加有效,能够在更高的效度和精度上对文本数据进行处理和分析。[Research purpose]By expanding and strengthening the text,an improved model based on Bert is proposed in order to obtain better accuracy and validity in the supervision and processing of network comment text data.[Research method]Taking the online review text of tea products as an example,this paper uses word2vec for in-depth learning of the text,sets the currently used regulatory methods as the control group,and sets the proposed Bert+Ttransformer model and Bert+XGB model as the experimental group for empirical research.[Research conclusion]The results show that the Bert+Transformer model and Bert+XGB model proposed in this paper are more effective than the currently applied regulatory means,and can process and analyze text data with higher validity and accuracy.

关 键 词：文本挖掘 文本数据 网络凭证 数据处理模型 细粒度情感分析 数据监管 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] G353[自动化与计算机技术—计算机科学与技术]