摘要
[目的/意义]深入分析AI4Science中的实验方案,通过分析科技文献中的实验方案,揭示AI技术在科研方法、工具和手段中的应用,为科研工作者提供新的研究视角。[方法/过程]首先,利用本体建模技术,实现不同学科领域实验方法与实验原理的统一知识建模,在本体建模的基础上构建有机太阳能电池领域知识图谱。然后,在领域知识图谱中挖掘实体之间关系,实现实验方案智能化推荐。[结果/结论]结合图嵌入表征技术Graph2vec和大模型语义嵌入表征GPT embedding,提出一种全新的知识图谱语义融合的实验方案推荐算法—GraphGPT Net,在Recall@20推荐核心指标上表现最为出色,达到了0.0299,能够证明知识图谱在实验方案推荐领域的有效性以及GraphGPT Net在推荐实验方案方面的显著能力。
[Purpose/Significance]Currently,the paradigm of data-intensive scientific discovery in scientific research is evolving towards intelligence.AI-driven scientific research(AI4Science)is becoming the engine of technological innovation and a new paradigm for scientific research.This study will delve into the experimental schemes within AI4Science,revealing the application of AI technology in research methods,tools,and means through the analysis of experimental schemes in scientific literature,providing new research perspectives for scientific researchers.[Method/Process]①Knowledge extraction and modeling of the experimental scheme.Ontology modeling technology was used to realize the unified knowledge modeling of experimental methods and experimental principles in different subject areas.The domain knowledge graph of organic solar cells was constructed on the basis of ontology modeling.②Research on intelligent recommendation of the experimental scheme based on the knowledge graph.The relationship between entities was mined in the domain knowledge graph to realize the intelligent recommendation of the experimental scheme.[Result/Conclusion]On the basis of Graph2vec representation technology and GPT embedding representation,GraphGPT Net is proposed as a new algorithm for knowledge graph semantic integration of the experimental scheme.The best performance is achieved on Recall@20 with a score of 0.0299,which proves its remarkable ability to recommend experimental schemes.
作者
张凯
石栖
Zhang Kai;Shi Qi(National Science Library,Chinese Academy of Sciences,Beijing 100190;Department of Information Resources Management,School of Economics and Management,University of Chinese Academy of Sciences,Beijing 100190;National Science Library(Chengdu),Chinese Academy of Sciences,Chengdu 610299)
出处
《知识管理论坛》
2024年第5期448-459,共12页
Knowledge Management Forum
基金
国家社会科学基金项目“支撑AI4Science的科技图书馆知识服务内容研究”(项目编号:22BTQ019)
中国科学院文献情报能力建设专项项目“‘智慧数据+AI’支撑科学创新实验方法的推理发现研究”(项目编号:E329090905)研究成果之一
关键词
知识图谱
有机太阳能电池
实验方案
推荐系统
knowledge graph
organic solar cells
experimental scheme
recommendation system
作者简介
张凯,硕士研究生,E-mail:zhangkai@mail.las.ac.cn;石栖,硕士研究生。
