摘要
【目的/意义】为探析研究前沿与发展趋向,突破现有跨学科科研协作研究在主题识别预测中的不足,本文提出跨学科科研协作新兴主题识别及预测研究框架,实现新兴主题识别及未来发展趋势预测。【方法/过程】以时间切片的形式对SciTS会议文本进行主题抽取,提出新兴主题测量指标,探测领域内新兴主题并构建新兴主题时间序列;而后分别采取BP神经网络和SVR两种经典机器学习算法对新兴主题未来三年发展趋势进行预测分析。【结果/结论】根据历史数据对跨学科科研协作新兴主题进行识别及预测有较好的效果,在识别出的五个新兴主题中,跨学科交流与对话、跨学科协作社区搭建、跨学科教育与培训等主题未来发展状态将趋热。【创新/局限】选取美国SciTS会议文本为典型案例展开探索性分析,丰富当前跨学科科研协作研究内容的层次性和多样性。
【Purpose/significance】To discover research fronts and development trends and overcome the limitation of research in therecognition and prediction currently,this paper proposed a research framework for revealing the emerging topics in interdisciplinaryscientific research collaboration and predicting their development trends.【Method/process】Firstly,this paper extracted the topics ofSci TS conference text under different time series,proposed measurement indicators of emerging topics and constructed their intensitytime series.Then,we used Back Propagation Neural Network and Support Vector Regression algorithms to make a prediction about thedevelopment trend of emerging topics in three years.【Result/conclusion】This paper has a good performance in the recognition and pre-diction of emerging topics.Meanwhile,five emerging topics have been discovered.Interdisciplinary communication skills,interdisciplin-ary collaboration community building,interdisciplinary education and traning will be the research fronts of the follow-up Sci TS confer-ence.【Innovation/limitation】This paper made some exploration based on the Sci TS conference text,which is the typical case in interdis-ciplinary scientific research collaboration,and to some extent enriched the width and diversity of interdisciplinary research.
作者
叶光辉
王灿灿
李松烨
YE Guang-hui;WANG Can-can;LI Song-ye(School of Information Management,Central China Normal University,Wuhan 430079,China)
出处
《情报科学》
CSSCI
北大核心
2022年第7期126-135,共10页
Information Science
基金
国家社会科学基金重大项目“新时代我国文献信息资源保障体系重构研究”(19ZDA345)
中央高校基本科研业务费项目“基于SciTS会议文本的跨学科科研协作主题识别及预测”(2020CXZZ123)。
关键词
跨学科
科研协作
新兴主题识别
趋势预测
文本挖掘
interdisciplinary
scientific research collaboration
emerging topic detection
trend prediction
text mining
