基于多视图融合的论文自动分类方法研究

Research on paper automatic classification method based on multi-view fusion

为科研工作者精准推荐所需的学术论文,从而节约检索时间和精力,提高科研效率,并进一步提升论文自动分类的准确度。该文在传统单视图论文分类基础上,提出了一种基于多视图融合的论文自动分类方法,考虑论文标题、关键词、摘要三个视图的互补性和协调性,实现对海量论文的自动分类。文中抓取了中国知网9个主题的1 710篇论文作为实验语料,并构建决策树、K最近邻、随机森林、支持向量机、朴素贝叶斯分类器进行实验。结果表明,基于多视图融合的论文分类方法在准确率、召回率和F值上都有所提升,优于单视图的论文分类方法,且可以为论文自动分类、推荐系统、文本挖掘提供有效支撑,具有一定的应用前景和实用价值。

On the basis of the traditional single-view paper classification,an automatic classification method based on multiview fusion is proposed to accurately recommend the required academic papers for scientific research workers,so as to save the retrieval time and energy,improve the scientific research efficiency,and further increase the accuracy of paper automatic classification. In the method,the complementarity and coordination of the three views of the title,keyword and abstract in the paper are considered to realize the automatic classification of massive papers. The 1710 papers on nine topics on CNKI were grabbed as the experimental corpus,and the decision tree,K nearest neighbor,random forest,support vector machine and naive Bayes classifier were constructed for the experiments. The results show that the paper classification method based on the multi-view fusion can improve the precision,recall rate and F value,which is better than the single-view paper classification method. The algorithm can provide effective support for automatic classification,recommendation system and text mining,which has certain application prospect and practical value.