An Explainable Approach to Semantic Link Mining in Multi-sourced Dynamic Data

Zhe Wang, Yi Sun, Hong Wu, Kewen Wang

Abstract: It is challenging to mine semantic links among multi-sourced data. Knowledge graphs can capture the semantics of data to support implicit links (cross-data sources) to be inferred through reasoning, known as the link prediction task. However, existing link prediction approaches are limited in their adaptability to data changes and cannot provide explanations for the predictions. In this work, we introduce a framework for semantic link mining through knowledge graphs and rule-based link prediction. In particular, rules representing higher-order patterns in the data are automatically mined and updated according to the dynamics of the data. We present a practical use case and a system for the semantic link mining of aviation data from multiple online sources. Besides, we evaluate our system by comparing it with several link prediction models to demonstrate the effectiveness of our approach in both static and dynamic link prediction and explanation.