# Interpretable and Low-Resource Entity Matching via Decoupling Feature Learning from Decision Making Source: [Interpretable and Low-Resource Entity Matching via Decoupling Feature Learning from Decision Making]() Collection index: [llms.txt]() Data updated: 2026-01-04T05:05:27.525Z Authors: Zijun Yao, Chengjiang Li, Tiansi Dong, Xin Lv, Jifan Yu, Lei Hou, Juanzi Li, Yichi Zhang, Zelin Dai Published: 2021-06-08T08:27:31.000Z ## Abstract Entity Matching (EM) aims at recognizing entity records that denote the same real-world object. Neural EM models learn vector representation of entity descriptions and match entities end-to-end. Though robust, these methods require many resources for training, and lack of interpretability. In this paper, we propose a novel EM framework that consists of Heterogeneous Information Fusion (HIF) and Key Attribute Tree (KAT) Induction to decouple feature representation from matching decision. Using self-supervised learning and mask mechanism in pre-trained language modeling, HIF learns the embeddings of noisy attribute values by inter-attribute attention with unlabeled data. Using a set of comparison features and a limited amount of annotated data, KAT Induction learns an efficient decision tree that can be interpreted by generating entity matching rules whose structure is advocated by domain experts. Experiments on 6 public datasets and 3 industrial datasets show that our method is highly efficient and outperforms SOTA EM models in most cases. Our codes and datasets can be obtained from https://github.com/THU-KEG/HIF-KAT. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Entity Matching on I-A1 10% labeled (test)](): Entity Matching; I-A1 10% labeled (test); F1 Score: 96. - [Entity Matching on D-A1 1% labeled train (test)](): Entity Matching; D-A1 1% labeled train (test); F1 Score: 96.6. - [Entity Matching on D-S1 1% labeled (test)](): Entity Matching; D-S1 1% labeled train (test); F1 Score: 88.2. - [Entity Matching on I-A2 10% labeled train (test)](): Entity Matching; I-A2 10% labeled train (test); F1 Score: 54.9. - [Entity Matching on D-S2 1% labeled train (test)](): Entity Matching; D-S2 1% labeled train data (test); F1 Score: 79.5. - [Entity Matching on Phone 10% labeled train (test)](): Entity Matching; Phone 10% labeled train (test); F1 Score: 0.949. - [Entity Matching on Skirt 1% labeled (test)](): Entity Matching; Skirt 1% labeled (test); F1 Score: 96.7. - [Entity Matching on D-A2 1% labeled train (test)](): Entity Matching; D-A2 1% labeled train (test); F1 Score: 80.3. - [Entity Matching on iTunes-Amazon Structured I-A1](): Entity Matching; iTunes-Amazon Structured I-A1; F1 Score: 95.9. - [Entity Matching on Skirt (Real)](): Entity Matching; Skirt (Real); F1 Score: 99.7. - [Entity Matching on Toner Real](): Entity Matching; Toner Real; F1 Score: 99.9. - [Entity Matching on Toner 1% labeled train data (test)](): Entity Matching; Toner 1% labeled train data (test); F1 Score: 97.2. - [Entity Matching on Phone Real](): Entity Matching; Phone Real; F1 Score: 97.5. - [Entity Matching on I-A2 iTunes-Amazon Dirty](): Entity Matching; I-A2 iTunes-Amazon Dirty; F1 Score: 63.3. - [Entity Matching on D-A1 Structured DBLP-ACM](): Entity Matching; D-A1 Structured DBLP-ACM; F1 Score: 98.1. - [Entity Matching on D-S1 Structured DBLP-Scholar](): Entity Matching; D-S1 Structured DBLP-Scholar; F1 Score: 0.917. - [Entity Matching on D-A2 DBLP-ACM Dirty](): Entity Matching; D-A2 DBLP-ACM Dirty; F1 Score: 89.7. - [Entity Matching on D-S2 Dirty DBLP-Scholar](): Entity Matching; D-S2 Dirty DBLP-Scholar; F1 Score: 80.5. ## Original sources and implementation [Original paper]() [PDF]() [Implementation]()