摘要
针对传统方法存在的语义标注准确度不高,语义分类查全率较低以及语义特征提取能力不佳的问题,研究基于Python技术的半监督文本语义分类方法。利用Python编程技术通过调用接口和扩展库建立自动标注下的半监督文本语义分类模型,通过自动标注手段将初始文本标注后划分为训练文本和测试文本。经过文本预处理后,利用改进的CHI算法展开文本语义特征提取和归一化处理并输入到监督分类器内,使用支持向量机算法完成文本语义分类。实验结果表明,上述方法文本语义标注准确度高于95%,标注精度高;词频曲线与实际词频曲线重合度较高,特征提取能力强;拟合误差低,受非线性问题影响小,且平均查全率为97.21%,说明所提方法的文本语义分类能力较好。
Due to the low accuracy of semantic annotation,low recall of semantic classification and unsatisfactory ability of semantic feature extraction in traditional methods,a method of semi-supervised text semantic classification based on Python technology was studied.First of all,Python technology,call interface and extension library were used to construct a model of semi-supervised text semantic classification under automatic annotation.After annotating automatically,the initial text was divided into training text and test text.After preprocessing text,the improved CHI algorithm was adopted to expand the text semantic feature extraction and normalization and then input them into a su-pervised classifier.Finally,the support vector machine algorithm was used to complete the text semantic classification.Following conclusions can be drawn from the experimental results.By using above method,the accuracy of text semantic annotation is higher than 95%;The word frequency curve is highly overlapped with the ac-tual curve,and the feature extraction ability is strong.In addition,this method has low fitting error,so it is less af-fected by nonlinear problems.In the meanwhile,its 97.21%average recall indicates that the method has better ability of text semantic classification.
作者
孙川钘
朱镕申
张凌云
SUN Chuan-xing;ZHU Rong-shen;ZHANG Ling-yun(Chengdu College,University of Electronic and Technology of China,Sichuan Chengdu 611731,China)
出处
《计算机仿真》
北大核心
2023年第7期496-500,共5页
Computer Simulation
基金
分布式数据库隐私信息增量式更新方法仿真(2017ZY0725)。
关键词
半监督
文本语义
支持向量机
自动标注
特征提取
Python technology
Semi-supervised
Text semantic
Support vector machine
Automatic annotation
Feature extraction
