摘要
[目的/意义]基于文献研究,揭示人工智能领域具有研究和发展潜力的新兴研究主题,为捕捉领域的发展动态和未来趋势提供参考。[方法/过程]构建人工智能领域国际顶级会议文献数据集,运用BERTopic主题模型实现主题发现,利用新颖性、增长性、主题强度,计算各研究主题的综合新兴潜力,遴选和评价领域新兴主题。[结果/结论]人工智能领域整体发展活力强,综合计算认为“目标检测”“联邦持续学习”和“梯度提升”是该领域当前最具潜力的新兴主题,“AI隐私保护”“3D点云”“音频与视觉交叉”是未来比较有潜力的新兴主题。文章提出的研究框架能够发现更具时效性、准确性和可解释性的研究结果,具备实际可行性。[局限]未对不同神经网络语言模型进行对比研究和综合分析,未来可综合利用不同模型,揭示更具系统性、针对性、方向性的研究结论。
[Purpose/significance]Revealing emerging research topics with research and development potential in the field of artificial intelligence from conference papers can provide a path to capture the dynamics and future trends.[Method/process]Es-tablish a dataset for the literature of top international conferences in the field of artificial intelligence,apply the BERTopic model to detect topics,and utilize novelty,growth,and topic strength to comprehensively calculate the emerging potential of each research topic,and evaluate the emerging topics in the field.[Result/conclusion]The overall development of the field of artificial intelli-gence is highly dynamic,and the comprehensive calculation considers“Object Detection”,“Federated Continual Learning”,and“Gradient Boosting”to be the most potential emerging topics at present.“3D point cloud”,“audio-vision crossover”,and“AI privacy protection”are promising emerging topics in the future.The research framework proposed in this paper is capable of discove-ring more current,accurate and interpretable findings with practical feasibility.[Limitations]we do not carry out comparative re-search and comprehensive analysis of different neural network language models,and does not have sufficient refinement of data with different focuses within the dataset.In the future,a combination of different models can be utilized to reveal more systematic,targe-ted,and directional findings.
作者
朱相丽
张敬
李伟伟
刘小平
耿国桐
Zhu Xiangli;Zhang Jing;Li Weiwei;Liu Xiaoping;Geng Guotong(National Science Library,Chinese Academy of Sciences,Beijing 100190;School of Economics and Management,University of Chinese Academy of Sciences,Beijing 100190;Geoscience Documentation Center of China Geological Survey,Beijing 100083;National Innovation Institute of Defense Technology,Academy of Military Sciences,Beijing 100071;Center for Information,Academy of Military Sciences,Beijing 100042)
出处
《情报理论与实践》
CSSCI
北大核心
2024年第9期147-155,共9页
Information Studies:Theory & Application
基金
国家自然科学基金项目“战略研究类:人工智能驱动的材料化学研究的战略研究”的成果之一,项目编号:22342011。
