Crowdsourcing Entity Resolution: When is A=B?

Abstract

There are several computational tasks for which the help of people is useful. One such task is entity resolution. For this task, human experts can help to identify whether two customers are identical given their profile. Since crowdsourcing is expensive, the goal is to ask as few questions as possible. At the same time, high quality results can only be achieved if several experts are asked for their opinion and for confirmation. This paper shows how to address this cost / quality trade-off and how to tolerate and resolve errors from the crowd. Specifically, this paper shows how to exploit mathematical properties such as symmetry, transitivity, and anti-transitivity of the is-same-entity-as relation to improve both cost and quality. The results of extensive experiments provide surprising insights on how best to crowd-source for entity resolution and other classification problems.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 46 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.