基于半监督支持向量机的期刊收稿系统自动分类方法
2018-12-14耿晓军
耿晓军
关键词: 期刊收稿系统; 自动分类; 专家审稿; 半监督支持向量机; 工作效率; 特征向量
中图分类号: TN957.52+3?34; TP393 文献标识码: A 文章编号: 1004?373X(2018)24?0174?04
An automatic classification method based on semi?supervised support vector
machine for periodical manuscript acceptance system
GENG Xiaojun
(Editorial Department of Modern Electronics Technique, Shaanxi Electronics Magazine Publishing Company, Xian 710032, China)
Abstract: In the current periodical manuscript acceptance system, the manual mode is mainly adopted to distribute the submitted e?mail manuscripts to the review experts in relevant professional fields, so as to complete manuscript review. However, the manual distribution mode is less efficient when facing with a large quantity of manuscripts. In order to solve the above problems and realize automatic distribution of submitted manuscripts, an automatic classification method based on the semi?supervised support vector machine is proposed. A vector space model based on TF/IDF feature weights is put forward to realize eigenvector representation of manuscripts. The semi?supervised support vector machine is used to classify datasets of manuscripts. The validity of the method was verified by analyzing manuscript acceptance instances of a certain journal. The experimental results show that the average F1 of the proposed automatic classification method based on the semi?supervised support vector machine for the periodical manuscript acceptance system is about 68%, which can improve the work efficiency of the periodical manuscript acceptance system while satisfying a certain accuracy condition.
Keywords: periodical manuscript acceptance system; automatic classification; expert review; semi?supervised support vector machine; work efficiency; feature vector
0 引 言
在信息技术时代,人们的生活和工作已经越来越依赖互联网,论文投稿系统网络化已经成为了现在的主流趋势[1?2]。但是,随着期刊对论文质量的要求越来越高,所需评审专家的专业领域细化和深化程度也随之提高。期刊编辑需要将作者的投稿分配给最合适的论文评审专家,以便逐渐提升期刊的稿件质量和专业层次。但是当面对大量的稿件时,人工分配方式的效率较低。因此,需要为投稿论文所属学科领域进行计算机自动分类,以便分配到相关专业领域的评审专家手中[3]。
传统的有监督学习方法通过大量的信息标注来提高自然语言处理的性能,取得了较好的效果[4]。但是训练数据的标注需要花费大量的时间和人工成本,在应对大数据任务处理时呈现出了不少问题。例如,没有充足标注语料或者标记资源质量较差的情况。因此,基于半监督学习的自然语言处理技术逐渐成为了研究的热点。半监督学习已经成为近期机器学习领域新兴的重要方向。与有监督学习方法的不同之处在于[5]:半监督学习同时使用标记数据和无标记数据。此外,相比仅利用标记数据的有监督学习方法,半监督学习具有更好的性能。
因此,为了实现投稿论文的自动分配,建立一种基于半监督支持向量机的论文自动分类方法。……
