Truth Finding from Multiple Data Sources by Source Confidence Estimation

The volume of data on the Web has been growing at a dramatic pace in recent years and people rely more and more on the Web to fulfill their information needs. Numerous different descriptions of the properties towards the same objects can be obtained from a variety of data sources. This will inevitably lead to data incompleteness, data conflicts and out-of-date information problems. These issues make truth discovery among multiple data sources non-trivial. However, most of previous works consider only one single property, or deal with different properties separately by ignoring several characteristics of the properties, which will often cause unexpected deviations. In this paper, we propose a modified method to find the most trustable source and identify the true information. Our goal is to minimize the distance between the true information and the overall observed descriptions through considering the accuracy and the coverage of all the data sources at the same time. The experiments on the real dataset demonstrate the efficacy of our method.