[目的/意义]消减时滞带来的数据冲突，增强多源数据的语义互补性与准确性，并提升知识融合的协同效应。[方法/过程]从主题时滞出发，通过在结构主题模型（structural topic model, STM）中设置来源协变量，利用主题内容、变化趋势和相似度衡量时滞方向和时滞期。结合时滞和语义相似度，将知识元语义分离为非时滞语义、时滞相似语义和时滞不相似语义3类，并提出相应的语义分离与融合策略。进而，构建多源数据集成型知识融合模式的实现方案，选用National Science Foundation和Web of Science中的人工智能领域数据进行实现应用与结果分析。[结果/结论]提出的局部时滞处理有助于在保持数据源特征的同时减弱其对数据融合负面影响的思路，是主题时滞测度和知识元相似度融合策略设计的依据。通过测算发现，两个数据的主题时滞现象既有提前也有滞后的情况，并且在不同主题上表现出不同的时滞方向和时长。知识元语义融合结果能够显著减少重复和杂乱语义，实现多源数据中偏移语义的对齐与去重，同时保证语义信息的完整性。表明考虑主题时滞所设计的“语义时滞—语义分离—语义融合”的路径方案，可有效地支撑多源数据集成型知识融合模式的实现。
[Purpose/Significance] This study aims to mitigate data conflicts caused by time lags, enhance semantic complementarity and accuracy of multi-source data, and improve the synergistic effect of knowledge fusion. [Method/Process] Focusing on topic time lag, this study incorporates the data source as a covariate in the structural topic model (STM). The direction and duration of the lag are measured by analyzing topic content, evolutionary trends, and similarity. By combining time lag and semantic similarity, the knowledge element semantics are separated into three categories: non-lagging semantics, lagging-similar semantics, and lagging-dissimilar semantics. And semantic separation and fusion strategies are proposed. Furthermore, an implementation framework for a multi-source data integration knowledge fusion model is constructed. This framework is applied and validated using artificial intelligence domain data from NSF and WoS. [Result/Conclusion] The proposed approach of local time lag processing mitigates their negative impact on data fusion while maintaining the characteristics of the data source. It is the basis for designing topic time lag measurement and knowledge element similarity fusion strategies. The calculation reveals that topic time lags of NSF and WoS data manifest as both leads and lags, and exhibit different lag directions and durations on different topics. The fusion of knowledge element semantics significantly reduces semantic redundancy and noise, achieves alignment and deduplication of divergent semantics in multi-source data, while preserving the integrity of semantic information. The proposed workflow of “Time Lag Measurement→Semantic Separation→Semantic Fusion” can effectively support the implementation of multi-source data integration knowledge fusion model.
[1] SMIRNOV A, LEVASHOVA T. Knowledge fusion patterns: a survey[J]. Information fusion, 2019, 52: 31-40.
[2] SMIRNOV A, LEVASHOVA T, SHILOV N. Patterns for context-based knowledge fusion in decision support systems[J]. Information fusion, 2015, 21: 114-129.
[3] SUN L, WANG Y Z. A multi-attribute fusion approach extending Dempster-Shafer theory for combinatorial-type evidences[J]. Expert systems with applications, 2018, 96: 218-229.
[4] ZHANG Y, WANG Y. Understanding delays in publishing interdisciplinary research[J]. Information processing & management, 2024, 61(5): 103826.
[5] 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.
[6] 范春竹, 邓小昭, 韩毅. 社交媒体用户消极使用行为：多源数据融合视角[J]. 图书情报工作, 2022, 66(15): 76-85. (FAN C Z, DENG X Z, HAN Y. Researches on passive behavior for social media users: from the perspective of the fusion of multi?source data[J]. Library and information service, 2022, 66(15): 76-85.)
[7] SCHMIDT M, KIRCHEIS W, SIMONS A, et al. A diachronic perspective on citation latency in Wikipedia articles on CRISPR/Cas-9: an exploratory case study[J]. Scientometrics, 2023, 128(6): 3649-3673.
[8] MIURA T, ASATANI K, SAKATA I. Revisiting the uniformity and inconsistency of slow-cited papers in science[J]. Journal of informetrics, 2023, 17(1): 101378.
[9] DING J, DU D. A study of the correlation between publication delays and measurement indicators of journal articles in the social network environment: based on online data in PLOS[J]. Scientometrics, 2023, 128(3): 1711-1743.
[10] XIE Y, WU Q, WANG Y, et al. Does the handling time of scientific papers relate to their academic impact and social attention? Evidence from Nature, Science, and PNAS[J]. Journal of informetrics, 2024, 18(2): 101504.
[11] MAJUMDAR S. A nine-year (2012-2020) study of publication delay in an open access Library and Information Science journal[J]. Malaysian journal of library and information science, 2023, 28(1): 35-50.
[12] RILEY W T, BIBB K, HARGRAVE S, et al. Publication rates from biomedical and behavioral and social science R01s funded by the National Institutes of Health[J]. Plos one, 2020, 15(11): e0242271.
[13] DAIM T U, MONALISA M, DASH P, et al. Time lag assessment between research funding and output in emerging technologies[J]. Foresight, 2007, 9(4): 33-44.
[14] MORRIS Z S, WOODING S, GRANT J. The answer is 17 years, what is the question: understanding time lags in translational research[J]. Journal of the Royal Society of Medicine, 2011, 104(12): 510-520.
[15] VAHID A, CALDERBANK R. Two-user erasure interference channels with local delayed CSIT[J]. IEEE transactions on information theory, 2016, 62(9): 4910-4923.
[16] 杨金庆, 陆伟, 吴乐艳. 面向学科新兴主题探测的多源科技文献时滞计算及启示——以农业学科领域为例[J]. 情报学报, 2021, 40(1): 21-29. (YANG J Q, LU W, WU L Y. Time-lag calculation and enlightenment of multi-source science and technology literature fusion for the detection of emerging research topic: a case study in the field of agriculture[J]. Journal of the China Society for Scientific and Technical Information, 2021, 40(1): 21-29.)
[17] ROBERTS M E, STEWART B M, TINGLEY D. STM: an R package for structural topic models[J]. Journal of statistical software, 2019, 91(2): 1-40.
[18] 滕广青, 吕晶, 江瑶, 等. 基于STM的科研资助对研究主题影响研究[J]. 现代情报, 2022, 42(5): 58-68. (TENG G Q, LYU J, JIANG Y, et al. Impact of research funding on research topics based on STM[J]. Journal of modern information, 2022, 42(5): 58-68.)
