AutoBlock

Entity matching seeks to identify data records over one or multiple data\nsources that refer to the same real-world entity. Virtually every entity\nmatching task on large datasets requires blocking, a step that reduces the\nnumber of record pairs to be matched. However, most of the traditional blocking\nmethods are learning-free and key-based, and their successes are largely built\non laborious human effort in cleaning data and designing blocking keys.\n In this paper, we propose AutoBlock, a novel hands-off blocking framework for\nentity matching, based on similarity-preserving representation learning and\nnearest neighbor search. Our contributions include: (a) Automation: AutoBlock\nfrees users from laborious data cleaning and blocking key tuning. (b)\nScalability: AutoBlock has a sub-quadratic total time complexity and can be\neasily deployed for millions of records. (c) Effectiveness: AutoBlock\noutperforms a wide range of competitive baselines on multiple large-scale,\nreal-world datasets, especially when datasets are dirty and/or unstructured.\n