Contents I. Introduction
In the past few decades, the amount of helpful information available on the Web has been proliferating, which has brought dramatic changes to human society [1]. People are more dependent on the Web to fulfill their information needs than ever [2]. However, a huge amount of disinformation, outdated data, and factual errors are filled on the Web [3]. It is difficult for users to distinguish the truth from various information [4], [5], [6], [7]. When searching for the birthplace of Adolf Hitler on the web, the answers include “Austria”, “Braunau”, “Germany”. It is difficult for users to choose the correct information among these conflicting answers. Since the collected information may be informed, incomplete, outdated, or existing factual errors, it is crucial to discover truths from various information which improves accuracy in information extraction. To solve the problem, truth discovery has attracted researchers’ attention recently [8], [9], [10] in many real-world application scenarios including web, social sensing, crowd sensing, privacy sensing, and deep neural network applications. For example, true facts are found from a large amount of conflicting information on many objects provided by various websites [11]. The sensory data collected from various mobile devices are usually unreliable in mobile cloud computing. Truthful information is extracted from unreliable sensory data in mobile crowd sensing [12]. It is necessary to aggregate noisy information on the objects, entities, or events collected from various sources.
