Abstract:

[Purposes] The classification of interdisciplinary journals from the perspective of knowledge elements in journal papers not only provides a new research perspective for content-based journal classification, but also offers valuable reference for researchers in selecting journals for submission. [Methods] Taking 20 core journals (CSSCI journals) in information resource management—characterized by notable interdisciplinarity and close knowledge exchange—as examples, the knowledge elements in these journals were extracted using the SciAIEngine platform. These knowledge elements were then vectorized using word embedding techniques. After dimensionality reduction with PCA, the knowledge elements were clustered. The centroid vectors of each cluster were obtained using vector synthesis methods, and similarities between centroid vectors of different journals were calculated. Journal similarity metrics were computed to construct a journal similarity matrix. Finally, clustering algorithms were applied to achieve journal categorization. [Findings] The classification results varied under different knowledge elements. Based on problem knowledge elements, the 20 journals were categorized into 3 groups of Library Science and 1 group of Information Science. When using method knowledge elements, the journals were divided into 2 groups of Library Science and 1 group of Information Science. [Conclusions] Compared to existing journal classification methods, the knowledge element-based journal classification method allows for “fine-grained” analysis of journal paper contents. It provides more objective, flexible, reliable, and dynamically adaptable classification results.

Key words: Knowledge element mining, Interdisciplinary, Academic journal, Journal classification

摘要：

【目的】从期刊论文知识元角度对交叉学科学术期刊进行分类,不仅为基于内容的期刊分类提供新的研究视角,也为科研人员投稿时的期刊选择提供有益参考。【方法】以学科交叉性明显的20种信息资源管理学科CSSCI期刊为例,在SciAIEngine平台抽取期刊中的知识元,利用词嵌入技术将知识元向量化,通过PCA降维处理后对知识元进行聚类。利用向量合成法获取每个知识元类团的质心向量,进而计算两两期刊之间不同质心向量的相似度。根据期刊相似度指标计算期刊间的相似度并构建期刊相似度矩阵,利用聚类算法实现期刊类别划分。【结果】不同知识元下的期刊分类结果有所不同,基于问题知识元将20种期刊分为3类图书馆学、1类情报学;基于方法知识元将20种期刊分为2类图书馆学、1类情报学。【结论】相比已有的期刊分类方法,基于知识元的期刊分类能够“细粒度”分析期刊论文的内容,分类结果更加客观、灵活、可靠且具有动态适应性。

关键词: 知识元, 交叉学科, 学术期刊, 期刊分类