[论文解读] Truth Discovery to Resolve Object Conflicts in Linked Data

Wenqiang Liu|arXiv (Cornell University)|Sep 1, 2015

Semantic Web and Ontologies参考文献 33被引用 3

一句话总结

本文提出 TruthDiscover，一种新颖的真相发现方法，用于解决在无标度链接数据中对象冲突的问题，其中大多数数据源仅提供少量冲突对象，而少数数据源则提供大量冲突对象。该方法利用源信念图的拓扑特性来估计源的先验可靠性，并使用隐马尔可夫随机场（HMRF）建模对象之间的相互依赖关系，从而在数据分布偏斜的情况下提升可信度估计的准确性。该方法在六个真实世界数据集上的表现显著优于现有方法。

ABSTRACT

In the community of Linked Data, anyone can publish their data as Linked Data on the web because of the openness of the Semantic Web. As such, RDF (Resource Description Framework) triples described the same real-world entity can be obtained from multiple sources; it inevitably results in conflicting objects for a certain predicate of a real-world entity. The objective of this study is to identify one truth from multiple conflicting objects for a certain predicate of a real-world entity. An intuitive principle based on common sense is that an object from a reliable source is trustworthy; thus, a source that provide trustworthy object is reliable. Many truth discovery methods based on this principle have been proposed to estimate source reliability and identify the truth. However, the effectiveness of existing truth discovery methods is significantly affected by the number of objects provided by each source. Therefore, these methods cannot be trivially extended to resolve conflicts in Linked Data with a scale-free property, i.e., most of the sources provide few conflicting objects, whereas only a few sources have many conflicting objects. To address this challenge, we propose a novel approach called TruthDiscover to identify the truth in Linked Data with a scale-free property. Two strategies are adopted in TruthDiscover to reduce the effect of the scale-free property on truth discovery. First, this approach leverages the topological properties of the Source Belief Graph to estimate the priori beliefs of sources, which are utilized to smooth the trustworthiness of sources. Second, this approach utilizes the Hidden Markov Random Field to model the interdependencies between objects to estimate the trust values of objects accurately. Experiments are conducted in the six datasets to evaluate TruthDiscover.

研究动机与目标

• 解决在源之间存在无标度冲突对象分布的链接数据中真相发现的挑战。
• 降低源贡献模式偏斜对真相发现中可信度估计的负面影响。
• 通过建模冲突对象之间的相互依赖关系，提升真相识别的准确性。
• 开发一种对源贡献不平衡具有鲁棒性的方法，这种不平衡在现实世界的链接数据中普遍存在。
• 提供一种能够适应现实世界数据图结构特性的可靠真相发现框架。

提出的方法

• 构建源信念图以表示源-对象关系，并利用其拓扑特性推导源的先验信念。
• 基于图的结构应用先验信念平滑机制，以改进初始可靠性估计。
• 使用隐马尔可夫随机场（HMRF）对冲突对象之间的相互依赖关系进行建模。
• HMRF 捕获对象之间的条件依赖关系，从而在个体源可靠性之外进一步优化可信度评分。
• 通过联合推理过程迭代更新源可靠性与对象可信度值。
• 通过在 HMRF 模型上的概率推理，联合估计源可信度与对象真实性。

实验结果

研究问题

• RQ1 如何使真相发现对链接数据中冲突对象的无标度分布具有鲁棒性？
• RQ2 源-对象图的拓扑特征在多大程度上能改善源的先验信念估计？
• RQ3 对冲突对象之间的相互依赖关系进行建模能否提升真相发现的准确性？
• RQ4 TruthDiscover 在真实世界数据集上与现有真相发现方法相比表现如何？
• RQ5 源贡献模式偏斜对传统真相发现方法有何影响？

主要发现

• 与基线方法相比，TruthDiscover 在六个真实世界数据集上显著提升了真相发现的准确性。
• 在源信念图中使用拓扑先验可减少稀疏贡献者带来的偏差，并增强可靠性估计。
• 基于 HMRF 的对象相互依赖关系建模可实现更准确的可信度评分传播。
• 即使少数源主导了冲突数据，该方法仍表现出稳健的性能。
• 实验表明，TruthDiscover 在源贡献不平衡较高的场景下，优于现有真相发现技术。
• 该方法有效缓解了无标度数据分布对真相发现的不利影响。