摘要
多维度网络存在“信息过载”问题,但信息源存在单调性,在交互过程中不可避免地会出现模糊的用户意图。单一特征提取器只能从一个维度提取跨域信息特征,增加了跨域信息检索响应的开销变异系数。为此,提出一种基于支持度匹配的多维度网络跨域信息语义检索方法。采用层次约简算法规范关键字格式并评估信息属性权重,通过逆文本词频分割法处理冗余信息,实现高质量的数据清洗。引入支持度模糊查询机制,构建模糊集合和词词关联矩阵,提高匹配容错性,并将支持度作为评估相关性的重要指标,对清洗后的跨域信息展开匹配。结合语义关联,运用两组独立的特征提取器创建异构网络,从多个角度挖掘跨域数据间的语义关联与模态一致性,并采用多种损失函数优化特征空间,实现高效的跨域信息检索与聚类,确保用户查询能够准确触及到相关信息。仿真结果表明,所提方法在处理不同规模的数据集时均表现出色,F1值更高,且响应开销变异系数最低,证明了其稳定性和泛化能力。
Multi-dimensional networks have the problem of"information overload",but the monotonicity of the information source makes the ambiguous user intent inevitable in the interaction process.A single feature extractor can only extract cross-domain information features from one dimension,which increases the overhead coefficient of variation of cross-domain information retrieval response.To this end.a semantic retrieval method for cross-domain information on a multi-dimensional web based on support matching is proposed.The hierarchical simplicity algorithm is used to standardize the keyword format and evaluate the weight of information attributes,and the inverse text word frequency segmentation method is used to deal with the redundant information to achieve high-quality data cleaning.The support fuzzy query mechanism is introduced to construct the fuzzy set and word-word association matrix to improve the fault tolerance ofmatching,and the support degree is used as an important index to evaluate the relevance of the cleaned cross-domain information to be matched.Combined with semantic association,two independent feature extractors are used to create a heterogeneous network to explore the semantic association and modal consistency between cross-domain data from multiple perspectives,and a variety ofloss functions are used to optimize the feature space to achieve efficient cross-domain information retrieval and clustering to ensure that the user query can accurately reach the relevant information.The simulation results show that the proposed method performs well in dealing with datasets of different sizes,with higher F1 values and the lowest coefficient of variation of response overhead,which proves its stability and generalization ability.
作者
黄晓燕
赵峰涛
谢国坤
HUANG Xiao-yan;ZHAO Feng-tao;XIE Guo-kun(Xi'an Traffic Engineering Institute,Xi'an Shaanxi 710300,China;Xi'an Peihua University,Xi'an Shaanxi 710125,China)
出处
《计算机仿真》
2025年第11期248-251,516,共5页
Computer Simulation
基金
陕西省“十四五”教育科学规划课题(SGH22Y1853)。
关键词
支持度匹配
多维度网络
语义关联
跨域检索
Support matching
Multidimensional networks
Semantic associations
Cross-domain retrieval
