基于二分网络分析方法的学术文献关键词自动抽取方法研究

The Automatic Extraction of Keywords from Academic Literature based on Bipartite Network Analysis Method

学术文献关键词的自动抽取在各种数字图书馆工作中有着广泛的应用,可以极大减少人工标注的成本和改善文献标注的准确度。本文在候选关键词的选择基础上,提出了一种利用文献结构中标题关键词和摘要关键词共现特征构造二分网络的方法,并详细说明了网络节点和边的权值设定依据和方法,据此实现了基于二分网络权值扩散方法的关键词自动抽取方法。最后,文章利用图情类文献数据集合的实验分析,对相关方法和评价结果都做了详细说明。

The automatic extraction of keywords from academic literature has been wildly used in all kinds of applications in digital libraries, which can distinctly reduce the cost of manual indexing and improve the accuracy of indexing. Based on the choosing of candidate keywords, this paper introduces how to compose the bipartite network with the co-occurrence of keywords in titles and abstracts of academic literature. The gauge and methods of weight assignment for nodes and edges are also proposed. Then an automatic extraction method of keywords based on the iteration algorithm of bipartite network is discussed in detail. Finally, we report the experiments results of algorithm and evaluation with the analysis in the collection of academic literature of Library and Information Science.