An Integrated Bayesian Approach for Effective Multi-Truth Discovery

Abstract

Truth-finding is the fundamental technique for corroborating reports from multiple sources in both data integration and collective intelligent applications. Traditional truthfinding methods assume a single true value for each data item and therefore cannot deal will multiple true values (i.e., the multi-truth-finding problem). So far, the existing approaches handle the multi-truth-finding problem in the same way as the single-truth-finding problems. Unfortunately, the multi-truth-finding problem has its unique features, such as the involvement of sets of values in claims, different implications of inter-value mutual exclusion, and larger source profiles. Considering these features could provide new opportunities for obtaining more accurate truthfinding results. Based on this insight, we propose an integrated Bayesian approach to the multi-truth-finding problem, by taking these features into account. To improve the truth-finding efficiency, we reformulate the multi-truthfinding problem model based on the mappings between sources and (sets of) values. New mutual exclusive relations are defined to reflect the possible co-existence of multiple true values. A finer-grained copy detection method is also proposed to deal with sources with large profiles. The experimental results on three real-world datasets show the effectiveness of our approach.

Truth Discovery via Exploiting Implications from Multi-Source Data

“…• MBM [16]: this method deals with the multi-truth problem by incorporating new mutual exclusion definition.…”

“…According to the mutual exclusion assumption, by claiming a value as the prospective truth (i.e., positive claims), a source is believed to implicitly disclaim all the other values on the same data item (i.e., negative claims). Several existing truth discovery methods [6,16] have incorporated this assumption. Specially, all probabilistic models [20,21] naturally conform to this assumption, as truth probability is exactly the measure that is capable of indicating the truthfulness of a value by itself.…”

“…In view of the significance of the positive and negative claims to truth discovery, we investigate the distribution of the two types of claims in various real-world datasets, e.g., the author dataset [17], the biography dataset [11], and the movie dataset [16]. We observe imbalanced distributions of positive and negative claims in most of the investigated datasets, due to the long-tail characteristic of the real-world datasets [8,15]:…”

“…In [20], Zhao et al propose a Bayesian model that is compatible with the multi-truth scenario, which shows a positive effect on the accuracy and allows for the existence of multiple true values. In [16], Wang et al define a new mutual exclusive relation among values for multi-truth discovery, and incorporate sources' confidence on their claims and a finer-grained copy detection technique into a Bayesian framework to address the problem. Both above methods avail from using the probabilistic methods, which can automatically detect the number of true values as a byproduct.…”

Empowering Truth Discovery with Multi-Truth Prediction

“…• Multi-truth discovery method, i.e., the Multi-truth Bayesian Model (MBM) [16], which comprehensively incorporates a new mutual exclusion definition for multitruth discovery from the reformatted claims.…”