图书情报工作

图书情报工作 >

2024 , Vol. 68 >Issue 13: 122 - 131

DOI: https://doi.org/10.13266/j.issn.0252-3116.2024.13.011

情报研究

基于动态多任务学习的科技文献推荐模型构建及实证研究

  • 李洁 ,
  • 张国标 ,
  • 周毅 ,
  • 郗玉娟 ,
  • 杨金庆
展开
  • 1苏州大学智能社会与数据治理研究院 苏州 215000;
    2 山东大学图书馆 济南 250100;
    3 苏州大学传媒学院 苏州 215000;
    4 苏州大学社会学院 苏州 215000;
    5 山东大学国际创新转化学院 青岛 266100;
    6 华中师范大学信息管理学院 武汉 430079
李洁，博士；张国标，讲师，博士，通信作者，E-mail：zgb0538@163.com；周毅，教授，博士，博士生导师；郗玉娟，助理研究员，博士；杨金庆，特任副教授，博士。

收稿日期: 2023-11-10

  修回日期: 2024-02-02

  网络出版日期: 2024-07-09

基金资助

本文系江苏省社会科学青年基金项目“疫情常态化背景下图书馆数字资源认知推荐研究”（项目编号：21TQC001）、国家社会科学基金青年项目“模糊认知视角下智慧图书馆资源推荐服务模式及实证研究”（项目编号：22CTQ009）和湖北省自然科学基金面上项目“基于大语言模型深度语义理解的领域技术谱系生成研究”（项目编号：2024AFB1018）研究成果之一。

收起

Construction and Empirical Study of Scientific and Technological Literature Recommendation Model Based on Dynamic Multi-task Learning

  • Li Jie ,
  • Zhang Guobiao ,
  • Zhou Yi ,
  • Xi Yujuan ,
  • Yang Jinqing
Expand
  • 1 Institute of Intelligent Society and Data Governance, Soochow University, Suzhou 215000;
    2 Shandong University Library, Jinan 250100;
    3 School of Communication, Soochow University, Suzhou 215000;
    4 School of Sociology, Soochow University, Suzhou 215000;
    5 School of Innovation and Entrepreneurship, Shandong University, Qingdao 266100;
    6 School of Information Management, Central China Normal University, Wuhan 430079

Received date: 2023-11-10

  Revised date: 2024-02-02

  Online published: 2024-07-09

Supported by

This work is supported by Youth Program of Social Science Foundation of Jiangsu Province project “Research on cognitive recommendation of library digital resources under the background of normalized epidemic prevention and control” (Grant No. 21TQC001), Youth Program of National Social Science Foundation of China project “Smart library resource recommendation service model and empirical research from the perspective of fuzzy cognition” (Grant No. 22CTQ009), Gernal Program of Natural Science Foundation of Hubei Province Project “Research on Domain Technology Genealogy Generation Based on Deep Semantic Understanding of Large Language Models” (Grant No. 2024AFB1018).

Fold

摘要

[目的/意义] 为实现科技文献推荐场景要素的交互增强，将各要素交互特性捕捉问题转化为多任务共同优化学习问题，构建基于动态多任务学习的科技文献推荐模型，以进一步提升科技文献推荐性能。 [方法/过程] 采用多任务学习方法，针对科技文献推荐要素可采集的关键特征进行子任务解构，借助多头注意力机制，进行子任务交互关系的动态学习，在动态学习各任务交互关系的基础上设计科技文献推荐模型。 [结果/结论] 根据CiteULike 数据实验结果，所构建的 DMRSTL 模型在 3 个评价指标上均显著优于对比模型，最高差值为 AUC指标提升 15.51%， MRR 指标提升 11.90%， nDCG@5 指标提升 16.45%，且通过任务组合对比实验进一步表明，借助推荐要素的交互增强，可以有效提升科技文献的推荐性能。

关键词： 科技文献推荐; 多任务学习; 多头注意力机制; 任务交互

本文引用格式

李洁 , 张国标 , 周毅 , 郗玉娟 , 杨金庆 . 基于动态多任务学习的科技文献推荐模型构建及实证研究[J]. 图书情报工作, 2024 , 68(13) : 122 -131 . DOI: 10.13266/j.issn.0252-3116.2024.13.011

Abstract

[Purpose/Significance] In order to enhance the interaction of scientific and technological literature recommendation scene elements and transform the problem of capturing the interactive characteristics of each element into a multi-task joint optimization learning problem, this paper constructs a scientific and technological literature recommendation model based on dynamic multi-task learning to further improve the performance of scientific and technological literature recommendation. [Method/Process] Based on multi-task learning method, the sub-tasks are deconstructed according to the key features collected from scientific and technological literature recommendation elements, and the multi-head attention mechanism was used to dynamically learn the interactive relationships of sub-tasks. A scientific and technological literature recommendation model was designed through dynamic learning of the interaction of each task. [Result/Conclusion] According to the experimental results of CiteULike data, the DMRSTL model constructed in this article is significantly better than the comparison model in three evaluation indicators. The highest difference is the increase of AUC indicator by 15.51%, MRR by 11.90%, and nDCG@5 indicator by 16.45%. The task combination comparative experiments further show that the interactive enhancement of recommendation elements can effectively improve the recommendation performance of scientific and technological literature.

Key words： scientific and technological literature recommendation; multi-task learning; multi-head attention mechanism; task interaction

参考文献

[1] World Development Indicators. Scientific and technical journal articles[EB/OL]. [2024-04-07]. https://databank.worldbank.org/source/world-development-indicators/Series/IP.JRN.ARTC.SC.
[2] 耿立校, 晋高杰, 李亚函, 等. 基于改进内容过滤算法的高校图书馆文献资源个性化推荐研究[J]. 图书情报工作, 2018, 62(21): 112-117. (GENG L X, JIN G J, LI Y H, et al. Research on personalized recommendation of university library literature resources based on improved content-based filtering algorithm[J]. Library and information service, 2018, 62(21): 112-117.)
[3] 马慧芳, 胡东林, 刘宇航, 等. 融合作者与文献影响力的科技论文推荐方法[J]. 西北师范大学学报(自然科学版), 2021, 57(2): 55-62. (MA H F, HU D L, LIU Y H, et al. Ecommendation method for scientific papers integrating influence of author and literature[J]. Journal of Northwest Normal University (natural science), 2021, 57(2): 55-62.)
[4] 熊回香, 唐明月, 叶佳鑫, 等. 融合加权异质网络与网络表示学习的学术信息推荐研究[J]. 现代情报, 2023, 43(5): 23-34. (XIONG H X, TANG M Y, YE J X, et al. Research on academic information recommendation combining weighted heterogeneous network and network representation learning[J]. Journal of modern information, 2023, 43(5): 23-34.)
[5] 杨艳, 李建中. 数字图书馆中基于文献拓扑的个性化推荐技术 [J]. 计算机研究与发展, 2004, 41: 484-489. (YANG Y, LI J Z. Topology-based paper recommendation in digital library[J]. Journal of computer research and development, 2004, 41: 484-489.)
[6] 席俊红. 基于案例推理的科技文献推荐系统研究[D]. 上海: 华东师范大学, 2005. (XI J H. Study on recommendation system of technological literatures using case-based rensoning[D]. Shanghai: East China Normal University, 2005.)
[7] 曾子明, 金鹏. 基于用户兴趣变化的数字图书馆知识推荐服务研究[J]. 图书馆论坛, 2016, 36(1): 94-99. (ZENG Z M, JIN P. Research on knowledge recommendation service of digital library based on users’ Interest drift[J]. Library tribune, 2016, 36(1): 94-99.)
[8] 熊回香, 杨雪萍, 高连花. 基于用户兴趣主题模型的个性化推荐研究[J]. 情报学报, 2017, 36(9): 916-929. (XIONG H X, YANG X P, GAO L H. Personalized recommendation research based on user interest topic model[J]. Journal of the China Society for Scientific and Technical Information, 2017, 36(9): 916-929.)
[9] 李宇佳, 王益成. 基于用户动态画像的学术新媒体信息精准推荐模型研究[J]. 情报科学, 2022, 40(1): 88-93, 101. (LI Y J, WANG Y C. Accurate recommendation model of academic new media information based on user dynamic profile[J]. Information science, 2022, 40(1): 88-93, 101.)
[10] KWAPONG B A, ANARFI R, FLETCHER K K. Personalized service recommendation based on user dynamic preferences[C]// Services computing–scc 2019. Berlin: Springer, 2019: 77-91.
[11] 孙雨生, 祝博. 基于知识图谱的信息推荐架构体系研究[J]. 情报理论与实践, 2021, 44(11): 116-123. (SUN Y S, ZHU B. Research on information recommendation architecture system based on knowledge graph[J]. Information studies: theory & application, 2021, 44(11): 116-123.)
[12] 李锴君, 牛振东, 时恺泽, 等. 基于学术知识图谱及主题特征嵌入的论文推荐方法[J]. 数据分析与知识发现, 2023, 7(5): 48-59. (LI K J, NIU Z D, SHI K Z, et al. Paper recommendation based on academic knowledge graph and subject feature embedding[J]. Data analysis and knowledge discovery, 2023, 7(5): 48-59.)
[13] ZHU Y, LIN Q, LU H, et al. Recommending scientific paper via heterogeneous knowledge embedding based attentive recurrent neural networks[J]. Knowledge-based systems, 2021, 215: 106744.
[14] 李贺, 刘嘉宇, 沈旺, 等. 基于模糊认知图的在线健康社区知识推荐研究[J]. 数据分析与知识发现, 2020, 4(12): 55-67. (LI H, LIU J Y, SHEN W, et al. Recommending knowledge for online health community users based on fuzzy cognitive map[J]. Data analysis and knowledge discovery, 2020, 4(12): 55-67.)
[15] ZHANG J, GU F, JI Y, et al. Personalized scientific and technological literature resources recommendation based on deep learning[J]. Journal of intelligent & fuzzy systems, 2021, 41(2): 2981-2996.
[16] 杨怀珍, 张静, 李雷. 融合类别偏好与项目时效因素的混合推荐[J]. 数据分析与知识发现, 2023, 7(7): 136-145. (YANG H Z, ZHANG J, LI L. Hybrid recommendation with category preferences and item timeliness factor[J]. Data analysis and knowledge discovery, 2023, 7(7): 136-145.)
[17] ZHANG Y, YANG Q. A survey on multi-task learning[J]. IEEE transactions on knowledge and data engineering, 2022, 34(12): 5586-5609
[18] CHEN S, ZHANG Y, YANG Q. Multi-task learning in natural language processing: an overview[EB/OL]. [2024-04-07]. https://arxiv.org/abs/2109.09138.
[19] DENG Y, ZHANG W, XU W, et al. A unified multi-task learning framework for multi-goal conversational recommender systems[J]. ACM transactions on information systems, 2023(3): 41.
[20] MISRA I, SHRIVASTAVA A, GUPTA A, et al. Cross-stitch networks for multi-task learning[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Las Vegas: IEEE, 2016: 3994-4003.
[21] MA J, ZHAO Z, YI X, et al. Modeling task relationships in multitask learning with multi-gate mixture-of-experts[C]//Proceedings of the 24th ACM sigkdd international conference on knowledge discovery & data mining. New York: Association for Computing Machinery, 2018: 1930-1939.
[22] LIU J, LI X, AN B, et al. Multi-faceted hierarchical multi-task learning for recommender systems[C]//Proceedings of the 31st ACM international conference on information & knowledge management. New York: Association for Computing Machinery, 2022: 3332-3341.
[23] XU Z, ZHAO M, LIU L, et al. Mixture of virtual-kernel experts for multi-objective user profile modeling[C]//Proceedings of the 28th ACM sigkdd conference on knowledge discovery and data mining. New York: Association for Computing Machinery, 2022: 4257-4267.
[24] LIU H Y, OTT M, GOYAL N, et al. RoBERTa: a robustly optimized Bert pretraining approach[EB/OL]. [2024-04-07]. https://arxiv.org/abs/1907.11692.
[25] KUMAR V, KHATTAR D, GUPTA S, et al. Deep neural architecture for news recommendation[C]//Conference and Labs of the Evaluation Forum. New York: Association for Computing Machinery, 2017: 85-104.
[26] ZHAO J J, DU B, SUN L L, et al. Deep multi-task learning with relational attention for business success prediction[J]. Pattern recognition, 2021, 2021(110): 50-64.
[27] 曾子明, 张瑜. 基于数据增强和多任务学习的突发公共卫生事件谣言识别研究[J]. 数据分析与知识发现, 2023, 7(11): 56-57. (ZENG Z M, ZHANG Y. Rumor detection of public health emergencies based on data augmentation and multi-task learning[J]. Data analysis and knowledge discovery, 2023, 7(11): 56-67.)
[28] WANG H, CHEN B, LI W J. Collaborative topic regression with social regularization for tag recommendation[C]//Twentythird international joint conference on artificial intelligence. New York: Association for Computing Machinery, 2013: 2719-2725.
[29] 新智元. 吴恩达的二八定律[EB/OL]. [2023-04-13]. https:// www.51cto.com/article/657024.html. (XIN Z Y. Andrew Ng’s the 80/20 Rule[EB/OL]. [2023-04-13]. https://www.51cto.com/article/657024.html.)
[30] SARWAR B, KARYPIS G, KONSTAN J, et al. Item-based collaborative filtering recommendation algorithms[C]// Proceedings of the 10th international conference on World Wide Web. New York: Association for Computing Machinery, 2001: 285-295.
[31] 孙金杨, 刘柏嵩, 任豪, 等. TAGAN：一种融合细粒度语义特征的学术论文对抗推荐算法[J]. 电信科学, 2021, 37(8): 57-65. (SUN J Y, LIU B S, REN H, et al. TAGAN: an academic paper adversarial recommendation algorithm incorporating finegrained semantic features[J]. Telecommunications science, 2021, 37(8): 57-65.)
Options
文章导航

/

〈 〉