Volume 43 Issue 3
Jun.  2025
Turn off MathJax
Article Contents
GAO Li, YANG Nuohan, LI Qing, WANG Yongheng, YAN Han, ZHAO Ruhao, MA Xiaoping. Deep Mining and Association Recommendation Method for Railway Safety Knowledge Based on Multimodal Information Fusion[J]. Journal of Transport Information and Safety, 2025, 43(3): 33-43. doi: 10.3963/j.jssn.1674-4861.2025.03.004
Citation: GAO Li, YANG Nuohan, LI Qing, WANG Yongheng, YAN Han, ZHAO Ruhao, MA Xiaoping. Deep Mining and Association Recommendation Method for Railway Safety Knowledge Based on Multimodal Information Fusion[J]. Journal of Transport Information and Safety, 2025, 43(3): 33-43. doi: 10.3963/j.jssn.1674-4861.2025.03.004

Deep Mining and Association Recommendation Method for Railway Safety Knowledge Based on Multimodal Information Fusion

doi: 10.3963/j.jssn.1674-4861.2025.03.004
  • Received Date: 2025-03-04
    Available Online: 2025-10-11
  • The rapid digital and intelligent transformation of railway information systems has created an urgent demand for fine-grained, explainable safety knowledge recommendations. To address the fragmentation of cross-modal associations and insufficient alignment with operational rules exhibited by traditional approaches, a framework integrating multimodal feature fusion with generative reasoning is investigated. A hierarchical railway safety knowledge graph is constructed, and topological features under business constraints are extracted via the Node2Vec algorithm. Simultaneously, a lightweight Transformer encoder (GTE) captured deep semantic embeddings of individual safety clauses. To balance contributions from graph and text features, a tunable weighting strategy is introduced, dynamically controlling the fusion ratio of text vectors and graph embeddings and applying a dual-constraint mechanism based on cosine similarity and predefined rules to generate candidate recommendations. A three-stage progressive retrieval architecture is designed to achieve precise multimodal alignment and suppress noise. Finally, the DeepSeek-R1 large language model served as the reasoning engine, with domain-specific prompting converting retrieved candidates into executable decision plans, thereby enhancing coherence and interpretability. Experiments on 27 safety documents from a railway operator, using a similarity threshold of 0.85 and a maximum of 10 recommendations per query, demonstrated a recommendation accuracy of 95% (an 8-percentage-point improvement over traditional methods) along with significant gains in contextual relevance and explainability. This investigation confirms the synergistic benefits of multimodal retrieval and generative reasoning, providing a robust technical foundation for evolving railway safety knowledge services from precise recommendation to intelligent decision support.

     

  • loading
  • [1]
    王朋辉, 李晓宇, 罗扬帆. 铁路安全治理过程管控体系构建及实施路径研究[J]. 中国铁路, 2024(2): 145-151.

    WANG P H, LI X Y, LUO Y F. Construction and implementation path of railway safety governance process control system[J]. China Railway, 2024(2): 145-151. (in Chinese)
    [2]
    毛骞, 谢维成, 乔逸天, 等. 推荐系统冷启动问题解决方法研究综述[J]. 计算机科学与探索, 2024, 18(5): 1197-1210.

    MAO Q, XIE W C, QIAO Y T, et al. Research on solutions to the cold-start problem in recommender systems: a review[J]. Computer Science and Exploration, 2024, 18(5): 1197-1210. (in Chinese)
    [3]
    AGGARWAL V, GARIMELLA A, SRINIVASAN B V, et al. ClauseRec: a clause recommendation framework for AI-aided contract authoring[C]. the 2021 Conference on Empirical Methods in Natural Language Processing, Online and Punta Cana, Dominican Republic: Association for Computational Linguistics, 2021.
    [4]
    党铮. 企业制度建设重要性及方法探究——从合规管理角度[J]. 现代管理, 2023, 13(11): 1405-1409.

    DANG Z. Research on the importance and methods of corporate system construction: from the perspective of compliance management[J]. Modern Management, 2023, 13(11): 1405-1409. (in Chinese)
    [5]
    严李强, 田博, 梁炜恒, 等. 藏文文本相似度计算方法研究[J]. 高原科学研究, 2021, 5(3): 70-77, 114.

    YAN LQ, TIAN B, LIANG WH, et al. Research on Tibetan text similarity calculation methods[J]. Plateau Science Research, 2021, 5(3): 70-77, 114. (in Chinese)
    [6]
    MIKOLOV T, CHEN K, CORRADO G, et al. Efficient estimation of word representations in vector space[OL]. (2013-01-14)[2025-04-12]. https://arxiv.org/abs/1301.3781.
    [7]
    PENNINGTON J, SOCHER R, MANNING C D. GloVe: global vectors for word representation[C]. The 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), Doha, Qatar: Association for Computational Linguistics, 2014.
    [8]
    张昱, 冯亚寒, 丁千惠. 融合Word2Vec词嵌入的多核卷积神经网络音乐歌词多情感分类方法[J]. 科学技术与工程, 2024, 24(20): 8598-8605.

    ZHANG Y, FENG Y H, DING Q H. Multi-emotion classification of music lyrics based on multi-kernel convolutional neural network with Word2Vec embedding[J]. Science Technology and Engineering, 2024, 24(20): 8598-8605. (in Chinese)
    [9]
    周燕. 基于GloVe模型和注意力机制Bi-LSTM的文本分类方法[J]. 电子测量技术, 2024, 45(7): 42-47.

    ZHOU Y. Text classification method based on GloVe model and attention-mechanism Bi-LSTM[J]. Electronic Measurement Technology, 2024, 45(7): 42-47. (in Chinese)
    [10]
    GROVER A, LESKOVEC J. Node2vec: scalable feature learning for networks[C]. The 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, USA: ACM, 2016.
    [11]
    HAMILTON W L, YING Z, LESKOVEC J. Inductive representation learning on large graphs[J]. Advances in Neural Information Processing Systems, 2017, 30: 1024-1034.
    [12]
    曹琨, 吴新年, 靳军宝, 等. 基于共词和Node2Vec表示学习的新兴技术识别方法[J]. 数据分析与知识发现, 2023, 7 (9): 89-99.

    CAO K, WU X N, JIN J B, et al. Emerging technology identification method based on co-word and Node2Vec representation learning[J]. Data Analysis and Knowledge Discovery, 2023, 7(9): 89-99. (in Chinese)
    [13]
    敬容, 杨逸民, 万福成, 等. 藏文短文本分类研究[J]. 中文信息学报, 2024, 38(9): 58-65.

    JING R, YANG Y M, WAN F C, et al. Tibetan short text classification research[J]. Journal of Chinese Information Processing, 2024, 38(9): 58-65. (in Chinese)
    [14]
    张茜茜. 基于文本挖掘的企业技术创新指标体系构建方法研究[D]. 北京: 北京交通大学, 2021.

    ZHANG Q X. Research on construction method of enterprise technology innovation indicator system based on text mining[D]. Beijing: Beijing Jiaotong University, 2021. (in Chinese)
    [15]
    何瑾琳, 刘学军, 徐新艳, 等. 融合Node2Vec和深度神经网络的隐式反馈推荐模型[J]. 计算机科学, 2019, 46(6): 41-48.

    HE J L, LIU X J, XU X Y, et al. Implicit feedback recommendation model integrating Node2Vec and deep neural networks[J]. Computer Science, 2019, 46(6): 41-48. (in Chinese)
    [16]
    孙明龙, 欧阳纯萍, 刘永彬, 等. 基于分层融合策略和上下文信息嵌入的多模态情绪识别[J]. 北京大学学报(自然科学版), 2024, 60(3): 393-402.

    SUN M L, OYANG C P, LIU Y B, et al. Multimodal emotion recognition based on hierarchical fusion strategy and contextual information embedding[J]. Journal of Peking University (Natural Science), 2024, 60(3): 393-402. (in Chinese)
    [17]
    王曙燕, 巩婧怡. 融合长短时序与文本分类的新闻推荐模型[J]. 西安邮电大学学报, 2023, 28(3): 82-87.

    WANG S Y, GONG J Y. News recommendation model integrating long and short-term sequences and text classification[J]. Journal of Xi'an University of Posts and Telecommunications, 2023, 28(3): 82-87. (in Chinese)
    [18]
    赵悦阳, 崔雷. 文本嵌入技术的研究与应用进展[J]. 数据与计算发展前沿, 2023, 5(3): 92-110.

    ZHAO Y Y, CUI L. Research and application advances of text embedding technologies[J]. Frontiers of Data and Computing Development, 2023, 5(3): 92-110. (in Chinese)
    [19]
    姚思雨, 赵天哲, 王瑞杰, 等. 规则引导的知识图谱联合嵌入方法[J]. 计算机研究与发展, 2020, 57(12): 2514-2522. DOI: 10.7544/issn1000-1239.2020.20200741.

    YAO S Y, ZHAO T Z, WANG R J, et al. Rule-guided knowledge graph joint embedding method[J]. Journal of Computer Research and Development, 2020, 57(12): 2514-2522. doi: 10.7544/issn1000-1239.2020.20200741
    [20]
    杨晨, 刘敬辉, 李晓宇, 等. 铁路安全生产标准化管理优化对策[J]. 中国铁路, 2023(11): 131-136.

    YANG C, LIU J H, LI X Y, et al. Optimization strategies for standardized management of railway safety production[J]. China Railway, 2023(11): 131-136. (in Chinese)
  • 加载中

Catalog

    通讯作者: 陈斌, bchen63@163.com
    • 1. 

      沈阳化工大学材料科学与工程学院 沈阳 110142

    1. 本站搜索
    2. 百度学术搜索
    3. 万方数据库搜索
    4. CNKI搜索

    Figures(7)  / Tables(4)

    Article Metrics

    Article views (34) PDF downloads(3) Cited by()
    Proportional views
    Related

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return