1. Automatic semantic modeling of structured data sources with cross-modal retrieval.
- Author
-
Xu, Ruiqing, Mayer, Wolfgang, Chu, Hailong, Zhang, Yitao, Zhang, Hong-Yu, Wang, Yulong, Liu, Youfa, and Feng, Zaiwen
- Subjects
- *
REPRESENTATIONS of graphs , *DATA modeling , *MACHINE learning , *SEARCH algorithms , *SEMANTICS , *LATENT semantic analysis - Abstract
Analyzing and modeling the implicit semantic relationships in data sources is the key to achieving integration and sharing of heterogeneous data information. However, manual modeling of data semantics is a laborious and error-prone task that demands significant human effort and expertise. The paper proposes a novel explainable representation learning-based method that adopts an attention-based table-graph cross-modal retrieval model as a rating function during incremental search for automatic semantic modeling. Our supervised model utilizes the graph representation learning technique to extract latent semantics from data and aims to retrieve the most reliable semantic model for structured data sources. Experimental results demonstrate the effectiveness and robustness of our method. • A novel supervised table-graph cross-modal retrieval model for automatic semantic modeling. • An explainable representation learning method combining graph representation learning with an incremental search algorithm. • An attention mechanism and iterative matching method to learn fine-grained semantic matches between different modalities. • The experimental results using two different semantic labelers validate the effectiveness and robustness of our method. • The cross-modal retrieval-based method outperforms the previous best method in terms of precision, recall, and f1-score. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF