Based on Entities Behavior Patterns of Heterogeneous Data Semantic Conflict Detection

Abstract

As the sources of data become more and more diversified, the importance of data conflict detection is emerging. We are committed to research a new method, through the use of behavior pattern detection of heterogeneous data semantic conflict. We find that the structured data which can represents the behavior of an entity contradict from the reality behavior of the entity which can be got from unstructured text, which is often referred to as pattern conflict. So in this paper, we convert the structured data with semantic into data-converted event. Combine them with the text event extracted from unstructured text, according to the relation between entities, get a large event graph G. Find the common conflict pattern through frequent sub-graph discovery on graph G. Then use the common conflict patterns to detect conflict data. The experiment shows that our method can detect the conflict data effectively with a high recall.