Prof. Dr. Felix Naumann

Role Matching in Temporal Data

Currently submitted to VLDB 2022


We present role matchings, a novel, fine-grained integrity constraint on temporal fact data, i.e., ⟨subject, predicate, object, timestamp⟩-quadruples. A role is a combination of subject and predicate and can be associated with different objects as the real world evolves and the data changes over time. A role matching is a novel constraint that states that  the associated object of two or more different roles should always match at the same timestamps. Once discovered, role matchings can serve as integrity constraints that, if  violated, can alert editors and thus allow them to correct the error. We present compatibility-based role matching (CBRM), an algorithm to discover role matchings in large datasets, based on their change histories.

We evaluate our method on datasets from the Socrata open government data portal, as well as Wikipedia infoboxes, showing that our approach can process large datasets of up to  3.5 million roles containing up to 17 million changes. Our approach consistently outperforms baselines, achieving almost 30 percentage points more F-Measure on average.


The following datasets will be made available soon. The extracted roles are immediately available for use in implementation (see below). The raw data is the original data source that from which the prepared role-sets were extracted.

Code Repositories

The following code-repositories are made available: