Dynamic source weight computation for truth inference over data streams

Yi Yang, Quan Bai, Qing Liu

Abstract: In the era of big data, it is easy to obtain data for the same objects from multiple agents, eg, websites, crowd workers and social sensors. Each agent, performed as a data source 1, can provide her observations, views or answers to the objects that we are interested in. Unfortunately, the multi-source data on the same object is usually conflicting. Generally, the multi-source data cannot be used as inputs by machine learning or analytic tasks because poor quality data could have a negative impact on the performance of the learning algorithms. Thus, truth inference has emerged and it aims at inferring the truth of each object from multi-source data.