基于PCA-CNN的动态短文本分析研究  

An Analytical Research on Dynamic Short Texts Based on PCA-CNN

作　　者：林寒冰 金秀玲[1] 王婷 林云霞 LIN Hanbing;JIN Xiuling;WANG Ting

机构地区：[1]闽江学院数学与数据科学学院(软件学院),福建福州350108

出　　处：《科技创新与应用》2022年第11期44-48,52,共6页Technology Innovation and Application

基　　金：2020年国家级创新训练项目(202010395012)。

摘　　要：动态短文本分析研究旨在反馈用户关注内容发展的最新状况,使用户在海量碎片化文本信息中快速获取目标主流信息,提升用户生活效率,打造阅读新体验。利用网络爬虫技术获取大量微博实时数据,构建文本特征,融合PCA算法的高维度特征矩阵降维及卷积神经网络短文本语义及情感分析,提出PCA-CNN模型。该PCA-CNN模型运行速度快,小巧灵活,且模型准确度达到86.85%,提升对于短文本关键信息的提取效率,弥补前后文之间情感联系较弱的缺点,改进卷积神经网络在中文短文本中的应用情况,为获取信息提供更便捷的方法。The purpose of dynamic short text analysis research is to feedback users’attention to the latest situation of content development,so that users can quickly obtain target mainstream information in massive fragmented text information,improve user life efficiency and create a new reading experience.Using Web crawler technology to obtain a large number of Weibo real-time data to construct text features,combining the dimensionality reduction of high-dimensional feature matrix of PCA algorithm and semantic and emotional analysis of convolution neural network short texts,a PCA-CNN model is proposed.The PCA-CNN model is fast,compact and flexible,and the accuracy of the model is up to 86.85%,which improves the efficiency of extracting key information from short texts,makes up for the weak emotional connection between the two texts,and improves the application of convolution neural network in Chinese short texts,thus providing a more convenient method for obtaining information.

关 键 词：自然语言处理 网络爬虫技术 PCA降维 卷积神经网络 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]