A probabilistic model for truth discovery with object correlations

In the era of big data, information can be collected from many sources. Unfortunately, the information provided by the multiple sources on the same object is usually conflicting. In light of this challenge, truth discovery has emerged and used in many applications. The advantage of truth discovery is that it incorporates source reliabilities to infer object truths. Many existing methods for truth discovery are proposed with many traits. However, most of them ignore the characteristic of object correlations in data and focus on static data only. Object correlations exist in many applications. In this work, we propose a probabilistic truth discovery model that considers not only source reliability but also object correlations. This is especially useful when objects only claimed by few sources, which is common for many real applications. Furthermore, an incremental truth discovery method that considers object correlations is also developed when data provided by multiple sources arrives sequentially. Truth can be inferred dynamically without revisiting historical data, and temporal correlation is considered for truth inference. The experiments on both real-world and synthetic datasets demonstrate that the proposed methods perform better than the existing truth discovery methods.