Two Applications of Statistical Relational Learning: Fake News Detection and Congress Voting Patterns

Abstract

I applied statistical relational learning to the prediction of tweets and congress votes. The algorithm achieved high level of prediction accuracy in predicting whether the contents of new tweets were consistent with old ones that were shown to be fake or true. This way of prediction is content-aware and not based on heuristic classifiers. The algorithms predicted the voting behaviors of the congress members. It showed that the voting results of the congress was self-consistent over time and highly predictable.