Maximal fusion of facts on the web with credibility guarantee

摘要: Abstract The Web became the central medium for valuable sources of information fusion applications. However, such user-generated resources are often plagued by inaccuracies and misinformation as a result inherent openness uncertainty Web. While finding objective data is non-trivial, assessing their credibility with high confidence even harder due to conflicts between sources. In this work, we consider novel setting fusing factual from guarantee maximal recall. ultimate goal that not only should be extracted much possible but also its must satisfy threshold requirement. To end, formulate problem instantiating set precision larger than pre-defined threshold. Our proposed approach learning process optimize parameters probabilistic model captures relationships sources, contents, underlying information. automatically searches best without pre-trained data. Upon convergence, used instantiate guarantee. evaluations real-world datasets show our outperforms baselines up 6 times.