基于知识元抽取的不同学科领域研究方法交流态势分析——以情报学与计算机科学学为例

Exchange Situation Analysis of Research Methods in Different Disciplines Based on Knowledge Element Extraction:Taking Information Science and Computer Science as Examples

[研究目的]为细粒度发现相近学科间的知识关联,推动学科间知识交流,通过抽取方法知识元,探析不同学科领域研究方法的交流态势。[研究方法]以情报学和计算机科学学为例,首先,编写方法知识元抽取规则,基于规则抽取两学科领域大样本文献的方法知识元,构建学科领域方法知识元库;再利用LDA模型抽取两学科领域代表性期刊的方法知识元;以此为基础,对比分析两学科领域的高频研究方法、已交流的研究方法、存在交流潜力的研究方法。[研究结论]研究发现,两学科已交流的研究方法有:聚类分析、LDA模型、BP神经网络、支持向量机、主成分分析、协同过滤、文献计量等;具有交流潜力的研究方法有:提高准确率的分类算法(AdaBoost算法、XGBoost算法、SMOTE算法)、图卷积神经网络、自然语言处理模型GloVe模型等。该研究不仅能为情报学研究者选择计算机科学学领域的研究方法提供参考,而且也为学科交叉研究提供了一种新思路。

[Research purpose]In order to find the knowledge association between similar disciplines in fine granularity and promote the knowledge exchange among disciplines,this paper analyzes the exchange situation of research methods in different disciplines by extracting method knowledge elements.[Research method]Taking information science and computer science as examples,firstly,the method knowledge elements extraction rules are written,and the method knowledge elements of large sample literature of the two disciplines is extracted based on the rules,and the method knowledge elements database of the discipline domain is constructed.Then LDA model is used to extract the method knowledge elements of the representative journals of the two disciplines.On this basis,the high-frequency research methods,the research methods that have been communicated,and the research methods that have the potential of communication are compared and analyzed.[Research conclusion]It is found that the two disciplines have shared the following research methods:cluster analysis,LDA model,BP neural network,support vector machine,principal component analysis,collaborative filtering,bibliometrics,etc.The research methods have communication potential include:classification algorithm(AdaBoost algorithm,XGBoost algorithm,SMOTE algorithm),graph convolution neural network,natural language processing model GloVe model,etc.This study can not only provide reference for information science researchers to choose research methods in computer science field,but also provide a new idea for interdisciplinary research.