Research Article
Multi-truth Discovery with Correlations of Candidates in Crowdsourcing Systems
@INPROCEEDINGS{10.1007/978-3-030-92638-0_2, author={Hongyu Huang and Guijun Fan and Yantao Li and Nankun Mu}, title={Multi-truth Discovery with Correlations of Candidates in Crowdsourcing Systems}, proceedings={Collaborative Computing: Networking, Applications and Worksharing. 17th EAI International Conference, CollaborateCom 2021, Virtual Event, October 16-18, 2021, Proceedings, Part II}, proceedings_a={COLLABORATECOM PART 2}, year={2022}, month={1}, keywords={Crowdsourcing systems Multi-truth discovery Candidate correlations Markov random field}, doi={10.1007/978-3-030-92638-0_2} }- Hongyu Huang
Guijun Fan
Yantao Li
Nankun Mu
Year: 2022
Multi-truth Discovery with Correlations of Candidates in Crowdsourcing Systems
COLLABORATECOM PART 2
Springer
DOI: 10.1007/978-3-030-92638-0_2
Abstract
In the past decade, crowdsourcing has emerged as a popular internet-based collaborative computing paradigm. In crowdsourcing systems, requesters can ask users (‘sources’) for true values (‘truths’) of objects or events. Generally, an object may have multiple truths and sources could provide inconsistent or even conflicting answers (‘candidates’) about the object. For this scenario, the multi-truth discovery is a promising technique to deal with various candidates provided by different sources. However, most of the existing multi-truth discovery methods ignore the correlation between candidates so that the inferred truth could be different from the ground truth. In order to solve this problem, we propose MTD-CC, aMulti-TruthDiscovery withCandidateCorrelations. Specifically, we first design a metric of potential function to measure the correlation between each pair of candidates based on sources’ votes and reliabilities. Then, we construct a Markov Random Field (MRF) to represent these correlations. Next, we transform the MRF into a directed graph and cut it based on the Min-cut theorem to infer which candidates are truths. Last, we evaluate the proposed method on both real and synthetic datasets and experimental results demonstrate that the accuracy of MTD-CC outperforms existing solutions.
