On the discovery of continuous truth: a semi-supervised approach with partial ground truths

In many applications, the information regarding to the same object can be collected from multiple sources. However, these multi-source data are not reported consistently. In the light of this challenge, truth discovery is emerged to identify truth for each object from multi-source data. Most existing truth discovery methods assume that ground truths are completely unknown, and they focus on the exploration of unsupervised approaches to jointly estimate object truths and source reliabilities. However, in many real world applications, a set of ground truths could be partially available. In this paper, we propose a semi-supervised truth discovery framework to estimate continuous object truths. With the help of ground truths, even a small amount, the accuracy of truth discovery can be improved. We formulate the semi-supervised truth discovery problem as an optimization task where object truths and source reliabilities are modeled as variables. The ground truths are modeled as a regularization term and its contribution to the source weight estimation can be controlled by a parameter. The experiments show that the proposed method is more accurate and efficient than the existing truth discovery methods.