# ShallowBlocker: Improving Set Similarity Joins for Blocking

arXiv: https://arxiv.org/abs/2312.15835 · HTML: https://mlsanity.com/paper/2312.15835/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 5 of its 26 references that are in the map.
- First posted on arXiv 26 Dec 2023.

## Builds on (most cited first, 5 of 5)

- [Billion-scale similarity search with GPUs](https://mlsanity.com/paper/1702.08734/) (2017)
- [DeepER -- Deep Entity Resolution](https://mlsanity.com/paper/1710.00597/) (2017)
- [Neural Networks for Entity Matching: A Survey](https://mlsanity.com/paper/2010.11075/) (2020)
- [Auto-FuzzyJoin: Auto-Program Fuzzy Similarity Joins Without Labeled Examples](https://mlsanity.com/paper/2103.04489/) (2021)
- [AutoBlock: A Hands-off Blocking Framework for Entity Matching](https://mlsanity.com/paper/1912.03417/) (2019)

## Abstract (arXiv)

Blocking is a crucial step in large-scale entity matching but often requires significant manual engineering from an expert for each new dataset. Recent work has show that deep learning is state-of-the-art and has great potential for achieving hands-off and accurate blocking compared to classical methods. However, in practice, such deep learning methods are often unstable, offers little interpretability, and require hyperparameter tuning and significant computational resources. In this paper, we propose a hands-off blocking method based on classical string similarity measures: ShallowBlocker. It uses a novel hybrid set similarity join combining absolute similarity, relative similarity, and local cardinality conditions with a new effective pre-candidate filter replacing size filter. We show that the method achieves state-of-the-art pair effectiveness on both unsupervised and supervised blocking in a scalable way.