|
|
|
| Fine-Grained Clustering of Scientific Fund Topics Based on Multi-task Heterogeneous Graph Representation Learning |
| Guan Zhengyi1, Xie Jing2,3, Liu Jianhua4,5 |
1.School of Information Management, Sun Yat-sen University, Guangzhou 510006
2.National Science Library, Chinese Academy of Sciences, Beijing 100190
3.Department of Information Resources Management, School of Economics and Management, University of Chinese Academy of Sciences, Beijing 100190
4.Beijing Wanfang Data Co., Ltd., Beijing 100036
5.Key Laboratory of New Publishing and Knowledge Services for Scholarly Journals, National Press and Publication Administration, Beijing 100190 |
|
|
|
|
Abstract The fine-grained topic clustering of scientific funding projects is of considerable significance for analyzing scientific funding layouts and optimizing resource allocation. To address the challenges of sparse features, high terminological density, and the lack of contextual information in funding texts, this paper proposes a fine-grained topic clustering method for scientific funds based on multi-task heterogeneous graph representation learning. The method first integrated a keyword co-occurrence network, literature citation topology network, and disciplinary classification hierarchy to construct a three-dimensional heterogeneous semantic graph. Subsequently, by combining the local feature extraction capability of graph convolutions with the global semantic association advantage of multi-head attention, a multi-task collaborative training mechanism was designed to learn the deep semantic representations of texts. Considering funded projects in the millimeter-wave domain as an example, the experimental results demonstrated that the proposed method significantly outperforms baseline models across multiple evaluation metrics. The constructed semantic representation space exhibits high intraclass compactness and interclass separation, effectively enabling fine-grained topic identification. Furthermore, the clustering results are highly consistent with the policy orientation of the fund guidelines. This method demonstrates a strong application value in the multi-dimensional text analysis of scientific funds, providing an effective tool for fund text mining and fine-grained project analysis.
|
|
Received: 15 September 2025
|
|
|
|
1 肖人毅. 基于网络分析法的国家自然科学基金资助项目后评价研究[J]. 中国科学基金, 2008, 22(5): 312-316.
2 吴锦池, 余维杰. 融合知识库语义的文本聚类研究[J]. 情报杂志, 2021, 40(5): 156-164.
3 王伟, 梁继文, 杨建林. 基于语义加权网络的重点领域科学基金资助态势识别研究[J]. 现代情报, 2025, 45(1): 46-59, 111.
4 张东鑫, 张敏. 图情领域LDA主题模型应用研究进展述评[J]. 图书情报知识, 2022, 39(6): 143-157.
5 Williams L, Anthi E, Arman L, et al. Topic modelling: going beyond token outputs[J]. Big Data and Cognitive Computing, 2024, 8(5): 44.
6 Hankar M, Kasri M, Beni-Hssane A. A comprehensive overview of topic modeling: techniques, applications and challenges[J]. Neurocomputing, 2025, 628: 129638.
7 龙艺璇, 安源, 王东晋, 等. 基于改进LDA模型的铁路领域主题发现研究[J]. 数字图书馆论坛, 2022(2): 26-32.
8 张振青, 孙巍. 基于特征测度和PhraseLDA模型的领域学科交叉主题识别研究——以纳米技术的农业环境应用领域为例[J]. 数据分析与知识发现, 2023, 7(7): 32-45.
9 伊惠芳, 刘细文. 一种专利技术主题分析的IPC语境增强Context-LDA模型研究[J]. 数据分析与知识发现, 2021, 5(4): 25-36.
10 许元杰, 李建平, 吴登生. 基于期刊论文主题耦合的学科结构识别方法研究[J]. 图书情报工作, 2024, 68(21): 3-15.
11 Bascur J P, Verberne S, Van Eck N J, et al. Which topics are best represented by science maps? An analysis of clustering effectiveness for citation and text similarity networks[J]. Scientometrics, 2025, 130(2): 1181-1199.
12 张金柱, 于文倩. 基于短语表示学习的主题识别及其表征词抽取方法研究[J]. 数据分析与知识发现, 2021, 5(2): 50-60.
13 王志强, 李宜展, 李云龙, 等. 基于BERTopic的大科学装置科学研究联合基金资助主题挖掘[J]. 图书情报工作, 2024, 68(24): 104-113.
14 高凡, 徐思佳. 基于BERTopic模型的数字政府治理领域的主题识别与内容分析[J]. 情报理论与实践, 2024, 47(8): 95-106.
15 Yadav A K, Gupta T, Kumar M, et al. A hybrid model integrating LDA, BERT, and clustering for enhanced topic modeling[J]. Quality Quantity, 2025, 59(3): 2381-2408.
16 Ravenda F, Ali Bahrainian S, Raballo A, et al. A self-supervised seed-driven approach to topic modelling and clustering[J]. Journal of Intelligent Information Systems, 2025, 63(1): 333-353.
17 谢士尧, 王小梅. 基于深度文本聚类的论文与专利数据融合方法研究[J]. 数据分析与知识发现, 2024, 8(4): 112-124.
18 Chen R H, Chen H G, Lu Y Y, et al. Supervised neural topic modeling with label alignment[J]. Transactions of the Association for Computational Linguistics, 2025, 13: 249-263.
19 冯小东, 惠康欣. 基于异构图神经网络的社交媒体文本主题聚类[J]. 数据分析与知识发现, 2022, 6(10): 9-19.
20 Soleymanian M, Mashayekhi H, Rahimi M. An incremental clustering algorithm based on semantic concepts[J]. Knowledge and Information Systems, 2024, 66(6): 3303-3335.
21 Song J, Lu X L, Hong J Y, et al. External information enhancing topic model based on graph neural network[J]. Expert Systems with Applications, 2025, 263: 125709.
22 Beltagy I, Lo K, Cohan A. SciBERT: a pretrained language model for scientific text[PP/OL]. V3. arXiv (2019-09-10). https://arxiv.org/pdf/1903.10676.
23 余朝晖. 科技论文中图分类号科学标引的实践与探讨[J]. 编辑之友, 2010(5): 100-101.
24 Cui Z C, Chen W L, Chen Y X. Multi-scale convolutional neural networks for time series classification[PP/OL]. V4. arXiv (2016-05-11). https://arxiv.org/pdf/1603.06995.
25 Yun S, Jeong M, Kim R, et al. Graph transformer networks[C]// Proceedings of the 33rd International Conference on Neural Information Processing Systems. Red Hook: Curran Associates, 2019: Article No.1073.
26 Vandenhende S, Georgoulis S, Van Gansbeke W, et al. Multi-task learning for dense prediction tasks: a survey[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(7): 3614-3633.
27 洪伟, 余超, 陈继新, 等. 毫米波与太赫兹技术[J]. 中国科学(信息科学), 2016, 46(8): 1086-1107.
28 Joulin A, Grave E, Bojanowski P, et al. Bag of tricks for efficient text classification[PP/OL]. V3. arXiv (2016-08-09). https://arxiv.org/pdf/1607.01759.
29 Jahanian M, Karimi A, Osati Eraghi N, et al. Introducing the cosine clustering index (CCI): a balanced approach to evaluating deep clustering[J]. SN Computer Science, 2024, 5(6): Article No.687.
30 Thulasidas M. A quality metric for K-means clustering based on centroid locations[C]// Proceedings of the 18th International Conference on Advanced Data Mining and Applications. Cham: Springer, 2022: 208-222.
31 Hassani M, Seidl T. Using internal evaluation measures to validate the quality of diverse stream clustering algorithms[J]. Vietnam Journal of Computer Science, 2017, 4(3): 171-183.
32 Halkidi M, Batistakis Y, Vazirgiannis M. On clustering validation techniques[J]. Journal of Intelligent Information Systems, 2001, 17(2/3): 107-145.
33 IMT-2020(5G)推进组. 5G愿景与需求(2014年)[R/OL]. (2014-05-15) [2025-09-15]. http://www.caict.ac.cn/kxyj/qwfb/bps/201804/ t20180426_158197.htm.
34 胡云. 毫米波多波束阵列天线研究[D]. 南京: 东南大学, 2019.
35 束锋, 杨淑萍, 许正文, 等. 毫米波无线通信系统混合波束成形综述[J]. 数据采集与处理, 2017, 32(3): 454-462.
36 Li J, Niu Y, Wu H, et al. Mobility support for millimeter wave communications: opportunities and challenges[J]. IEEE Communications Surveys Tutorials, 2022, 24(3): 1816-1842.
37 Hemadeh I A, Satyanarayana K, El-Hajjar M, et al. Millimeter-wave communications: physical channel models, design considerations, antenna constructions, and link-budget[J]. IEEE Communications Surveys Tutorials, 2018, 20(2): 870-913.
38 陈山枝, 时岩, 胡金玲. 蜂窝车联网(C-V2X)综述[J]. 中国科学基金, 2020, 34(2): 179-185.
39 关于发布2024年度国家自然科学基金区域创新发展联合基金项目指南(第二批)的通告[EB/OL]. (2024-03-15) [2025-04-29]. https://www.nsfc.gov.cn/p1/3381/2824/66906.html.
40 关于发布2024年度国家自然科学基金企业创新发展联合基金项目指南(第二批)的通告[EB/OL]. (2024-03-15) [2025-04-29]. https://www.nsfc.gov.cn/p1/2931/2932/77905.html.
41 国家自然科学基金委员会2024年度专项项目指南——月球与深空探测科学研究(二)[EB/OL]. (2024-09-14) [2025-04-29]. https://www.nsfc.gov.cn/p1/3381/2824/66824.html.
42 “未来集成电路新理论与技术基础研究”专项项目申请指南[EB/OL]. (2023-11-07) [2025-04-29]. https://www.nsfc.gov.cn/p1/ 2856/3156/73825.html.
43 国家自然科学基金“十四五”发展规划[EB/OL]. (2022-11-16) [2025-04-29]. https://www.nsfc.gov.cn/p1/2871/2875/2951/gjzrkxjjsswfz.html.
44 国家自然科学基金“十二五”发展规划[EB/OL]. (2011-07-15) [2025-04-29]. https://www.nsfc.gov.cn/bzgh_125/ml.html.
45 国家自然科学基金“十三五”发展规划[EB/OL]. (2016-06-14) [2025-04-29]. https://www.nsfc.gov.cn/bzgh_135/index.html. |
|
|
|
