We will follow an entire research cycle from problem inception and literature research to algorithm development and, finally, to evaluation and deployment of a publicly available ERdataset repository. Together, we will prepare a research article and submit it to an international conference. Our goal is to bring clarity to the embarrassingly diffuse landscape of ER-Benchmark datasets. We lay a special focus on the analysis of different dataset versions. We want to develop algorithms that can (i) cluster ER-datasets to identify different versions of the same data and (ii) characterize the differences of multiple dataset versions using frameworkssuch as Explain-Da-V [5]. We start the project with a literature search phase. Afterwards, we will collect ERdataset versions typically used to benchmark ER-algorithms. Here we are not starting from scratch, but already have a few dataset versions as a starting point. Then, we will design and develop our algorithms for clustering and characterizing those dataset versions. Thereafter, we will conduct a sensitivity analysis and check whether using different versions of the same dataset has an impact on the reported quality of several state-of-the-art ER-approaches, such as Ditto [6]. Finally, we will build our repository and develop a dashboard to help users navigate within this repository. In summary, our approach consists of the following tasks:

• Collect available ER benchmark datasets
• Provide datasets and ground truths in a unified format
• Store meta-information about the datasets
• Cluster datasets to identify different versions of the same data
• Characterize the differences of multiple dataset versions
• Measure impact of different dataset versions (sensitivity analysis)
• Create a dashboard that allows filtering and querying the dataset repository
• Rights management (are we allowed to share them?)