Scaling Entity Resolution: Meta-blocking Methods Evaluated on 6 Heterogeneous Datasets

Entity resolution methods for linking records across large, heterogeneous data collections. The work proposes new meta-blocking algorithms that improve precision by up to an order of magnitude and reduce overhead time by more than an order of magnitude. The approaches were evaluated in an experimental study over 6 real-world datasets by author George Papadakis.

Use Cases

• Improving precision in entity resolution pipelines based on the described meta-blocking methods.
• Reducing computational overhead for blocking techniques in large-scale data matching tasks.
• Benchmarking new entity resolution algorithms against the state-of-the-art methods described.
• Applying enhanced meta-blocking to heterogeneous data collections for data cleaning and integration.

Strengths

• Methods evaluated on 6 real-world, heterogeneous datasets.
• Proposed algorithms improve precision by up to an order of magnitude.
• Efficiency techniques reduce meta-blocking overhead time by more than an order of magnitude.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and specific dataset characteristics are unknown, which may limit suitability assessment.

Provenance

Source

George Papadakis

Collection Method

Experimental study evaluating proposed meta-blocking methods.