基于深度神经网络的医药专利文本聚类模型研究  

Research on Medical Patent Text Clustering Model Based on Deep Neural Network

作　　者：王思源 何先波[1] WANG Siyuan;HE Xianbo(School of Computer,China West Normal Univerdity,Nanchong 637002,China)

机构地区：[1]西华师范大学计算机学院,四川南充637002

出　　处：《太原师范学院学报（自然科学版）》2021年第3期23-27,共5页Journal of Taiyuan Normal University:Natural Science Edition

基　　金：国家自然科学基金项目(61871330);西华师范大学英才科研基金项目(17YC149).

摘　　要：传统的基于机器学习方法进行特征提取的文本聚类模型,得到的文本特征是高维、稀疏的,且不能很好地挖掘复杂专利文本的潜在语义信息;鉴于此,文章设计了一种基于深度神经网络的医药专利文本聚类模型.首先对获取到的医药专利文本进行文本预处理,然后进行词向量训练、使用设计的CBL深度特征提取网络对医药专利文本进行深度特征提取,最后将提取到的特征作为优化K-Means聚类算法的输入,得到专利文本聚类结果.实验结果表明,提出的医药专利文本模型聚类质量在四个指标上均达到94%.The traditional text clustering model based on machine learning method for feature extraction is high-dimensional and sparse,and can not well mine the potential semantic information of complex patent text.In view of this,this paper designs a Pharmaceutical patent text clustering model based on deep neural network.Firstly,the obtained medical patent text is preprocessed,then word vector training is carried out,and the CBL deep feature extraction network designed is used to extract the deep feature of the medical patent text.Finally,the extracted features are used as the input of the optimized K-means clustering algorithm to obtain the patent text clustering results.Experiments show that the clustering quality of the proposed medical patent text model reaches 94%on the four indexes.

关 键 词：医药专利 深度神经网络 文本聚类 深度特征提取 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]