Cost-effective crowdsourced join queries for entity resolution without prior knowledge

Yin，Bo1; Zeng，Weilong1; Wei，Xuetao2

2022-02-01

10.1016/j.future.2021.09.008

Future Generation Computer Systems-The International Journal of eScience   影响因子和分区

0167-739X

The join query, which finds matching pairs from two object sets, is a fundamental operation in computer systems and helps to solve many real problems, e.g., entity resolution. In this paper, we address the problem of join queries by leveraging crowdsourcing to obtain matching relationships. The goal is to minimize the monetary cost while maintaining high quality of query results. However, existing approaches focused on finding matching pairs from a single object set and assumed the existence of prior knowledge, which is not applicable in real applications. We propose a cost-effective crowdsourced join query framework that minimizes the overall monetary cost by reducing the monetary cost of labeling single pairs and the amount of comparison pairs. Specifically, we first propose a novel two-level confidence-based labeling model that minimizes the cost for labeling a single pair with confidence guarantee. This model crowdsources easy-judging pairs to ordinary workers, and asks for skilled workers who may charge more than ordinary workers to compare only hard-judging pairs. Statistical estimations are used to aggregate crowdsourcing results with 1−α confidence. Then, we propose a transitivity-based query scheme that minimizes the number of comparison pairs on the basis of transitive relations. Guided by the principle of eagerly identifying matching pairs, especially matching pairs from a single set, our scheme carefully designs the processing order of pairs in order to make full use of transitivities to infer new labels. The results of our extensive experiments demonstrate that the proposed framework can save much more monetary cost while assuring the accuracy of results.

Crowdsourcing Entity resolution Join queries Monetary cost Quality

SCI ; EI

英语

其他

WOS:000706478900004

20213910954793

Cost effectiveness ; Search engines

Computer Software, Data Handling and Applications:723 ; Industrial Economics:911.2

2-s2.0-85115749173

Scopus

1.School of Computer and Communication Engineering,ChangSha University of Science and Technology,Changsha,410114,China
2.Department of Computer Science and Engineering,Southern University of Science and Technology,Shenzhen,518055,China

Yin，Bo,Zeng，Weilong,Wei，Xuetao. Cost-effective crowdsourced join queries for entity resolution without prior knowledge[J]. Future Generation Computer Systems-The International Journal of eScience,2022,127:240-251.

Yin，Bo,Zeng，Weilong,&Wei，Xuetao.(2022).Cost-effective crowdsourced join queries for entity resolution without prior knowledge.Future Generation Computer Systems-The International Journal of eScience,127,240-251.

Yin，Bo,et al."Cost-effective crowdsourced join queries for entity resolution without prior knowledge".Future Generation Computer Systems-The International Journal of eScience 127(2022):240-251.