Unsupervised Entity Resolution With Blocking and Graph Algorithms

Zhang, Dongxiang; Li, Dongsheng; Guo, Long; Tan, Kian-Lee

doi:10.1109/tkde.2020.2991063

Cited by 12 publications

(11 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Others aim to improve the quality of human labelers [45,54]. Alternative research directions aim to surpass the need for humans in-the-loop by introducing unsupervised learning solutions [24,52,57] and transfer learning methods [22,47] or limiting the amount of labels for training using active learning mechanisms [21,23,30] and few-shot learning [23,55]. In this work we analyze and aim to improve the quality of labels provided in a crowdsourcing setting.…”

Section: Related Workmentioning

confidence: 99%

HumanAL: Calibrating Human Matching Beyond a Single Task

Shraga¹

2022

Preprint

View full text Add to dashboard Cite

This work offers a novel view on the use of human input as labels, acknowledging that humans may err. We build a behavioral profile for human annotators which is used as a feature representation of the provided input. We show that by utilizing black-box machine learning, we can take into account human behavior and calibrate their input to improve the labeling quality. To support our claims and provide a proof-of-concept, we experiment with three different matching tasks, namely, schema matching, entity matching and text matching. Our empirical evaluation suggests that the method can improve the quality of gathered labels in multiple settings including cross-domain (across different matching tasks).

show abstract

Section: Related Workmentioning

confidence: 99%

HumanAL: Calibrating Human Matching Beyond a Single Task

Shraga¹

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Unsupervised ER approaches [5], [46], [49] are designed to perform ER without labeling. ZeroER [46] learns the match and mismatch distributions based on Gaussian Mixture Models.…”

Section: Related Workmentioning

confidence: 99%

“…EMBDI [5] performs ER by learning a compact graph-based representation for each tuple. ITER+CliqueRank [49] first constructs a bipartite graph to model the relationship between tuple pairs, and then develops an iterative-based ranking algorithm to estimate the similarity of tuple pairs. Despite the benefit of zero label requirement, unsupervised approaches are highly errorsensitive and may suffer from poor ER results when errors are contained in datasets.…”

Section: Related Workmentioning

confidence: 99%

“…The annotating process is labor-intensive and hence restricts the scope of its applications in real-world ER scenarios. Although several ER approaches [5], [46], [49] have tried to perform ER in an unsupervised way without any annotation, their ER performance is far from satisfactory due to the error-sensitive nature. A large body of research [3], [15], [26] has indicated that unsupervised methods can be easily misled/fooled or attacked since they do not include any supervision signal.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

CollaborER: A Self-supervised Entity Resolution Framework Using Multi-features Collaboration

Ge¹,

Wang²,

Lü³

et al. 2021

Preprint

View full text Add to dashboard Cite

Entity Resolution (ER) aims to identify whether two tuples refer to the same real-world entity and is well-known to be labor-intensive. It is a prerequisite to anomaly detection, as comparing the attribute values of two matched tuples from two different datasets provides one effective way to detect anomalies. Existing ER approaches, due to insufficient feature discovery or error-prone inherent characteristics, are not able to achieve a stable performance. In this paper, we present CollaborER, a self-supervised entity resolution framework via multi-features collaboration. It is capable of (i) obtaining reliable ER results with zero human annotations and (ii) discovering adequate tuples' features in a fault-tolerant manner. CollaborER consists of two phases, i.e., automatic label generation (ALG) and collaborative ER training (CERT). In the first phase, ALG is proposed to generate a set of positive tuple pairs and a set of negative tuple pairs. ALG guarantees the high quality of the generated tuples, and hence ensures the training quality of the subsequent CERT. In the second phase, CERT is introduced to learn the matching signals by discovering graph features and sentence features of tuples collaboratively. Extensive experimental results over eight real-world ER benchmarks show that CollaborER outperforms all the existing unsupervised ER approaches and is comparable or even superior to the state-of-the-art supervised ER methods.

show abstract

“…Generally, unsupervised entity resolution systems resort to matching threshold setting and end the entity matching process at that stage, such as in [13]. Other systems further expand the pipeline to identify entity profiles generalizing the entity matching output to more than two entity clusters.…”

Section: Introductionmentioning

confidence: 99%

ModER: Graph-based Unsupervised Entity Resolution using Composite Modularity Optimization and Locality Sensitive Hashing

Ebeid¹,

Talburt²,

Hagan³

et al. 2022

IJACSA

View full text Add to dashboard Cite

Entity resolution describes techniques used to identify documents or records that might not be duplicated; nevertheless, they might refer to the same entity. Here we study the problem of unsupervised entity resolution. Current methods rely on human input by setting multiple thresholds prior to execution. Some methods also rely on computationally expensive similarity metrics and might not be practical for big data. Hence, we focus on providing a solution, namely ModER, capable of quickly identifying entity profiles in ambiguous datasets using a graphbased approach that does not require setting a matching threshold. Our framework exploits the transitivity property of approximate string matching across multiple documents or records. We build on our previous work in graph-based unsupervised entity resolution, namely the Data Washing Machine (DWM) and the Graphbased Data Washing Machine (GDWM). We provide an extensive evaluation of a synthetic data set. We also benchmark our proposed framework using state-of-the-art methods in unsupervised entity resolution. Furthermore, we discuss the implications of the results and how it contributes to the literature.

show abstract

Unsupervised Entity Resolution With Blocking and Graph Algorithms

Cited by 12 publications

References 31 publications

HumanAL: Calibrating Human Matching Beyond a Single Task

HumanAL: Calibrating Human Matching Beyond a Single Task

CollaborER: A Self-supervised Entity Resolution Framework Using Multi-features Collaboration

ModER: Graph-based Unsupervised Entity Resolution using Composite Modularity Optimization and Locality Sensitive Hashing

Contact Info

Product

Resources

About