Benchmarking Scalability and Effectiveness of Blocking Approaches for Entity Matching: A Comparative Analysis

To leverage the potential of emerging data volumes, entity matching is applied to integrate information from different sources. However, performing pairwise comparisons between all records results in quadratic complexity, rendering it impractical for large datasets. To mitigate this issue, blocking is employed to group similar records into blocks, restricting comparisons to entries within the same block and thereby significantly reducing computational costs. As blocking is critical for enabling efficient data integration, different approaches have emerged, each with distinct strengths and weaknesses. Furthermore, various parameters must be carefully configured by developers to optimize results. To support an optimal selection of approach and parameter configuration, we conduct a comprehensive benchmark of blocking approaches, using both benchmark datasets and real-world data sources. Our analysis investigates eight open source blocking approaches. The results provide valuable insights into the scalability and effectiveness of each approach and the influence of parameter configurations. The findings highlight the importance of both selecting the appropriate blocking approach and parameter configuration to ensure efficient and accurate data integration.