Truth Discovery: CoTruth

. More and more people choose to obtain news online in today’s world. However, in recent times, there exists many fake claims on web. Rumor detection has been a popular topic. Recent search on truth discovery still doesn’t satisfy the realistic requirments now. As most of them focus on structured data, others proposed a new method to detect fake news with unstructured data. That is combine source reliability with stance classifier to determine the truth. But this method still have limitations with poor performance in stance classifier and neglection of source bias. . This paper overcomes the limitations in prior work by imporving stance classifier with non-linear training method. In addition, we also propose a new source reliability assessment method BiasR to solve the problem before.