Truth discovery on multi-dimensional properties of data sources

In the era of information explosion, data fusion has captured increasing attention from researchers as it plays an important part in data application. However, resolving the inconsistency of information generated by various data sources, i.e., truth discovery, has posed great challenges to data fusion. Although existing truth discovery methods mainly focus on source quality, they conduct the truth derivation iteratively only based on the accuracy of sources rather than the recall, which leads to the bad precision. Considering the assumption, sources are independent, cannot satisfy the reality of diverse correlations between them any more. A Gaussian Truth Finder with Correlations (GTFC) algorithm has been proposed in this paper. GTFC iteratively derives the truth, accuracy and recall of sources. The empirical results demonstrate that GTFC can significantly outperform the state-of-the-art algorithms.