Braid: Weaving Symbolic and Neural Knowledge into Coherent Logical Explanations

Traditional symbolic reasoning engines, while attractive for their precision and explicability, have a few major drawbacks: the use of brittle inference procedures that rely on exact matching (unification) logical terms, an inability to deal with uncertainty, need precompiled rule-base knowledge (the “knowledge acquisition” problem). To address these issues, we devise novel reasoner called Braid, supports probabilistic rules, uses notion custom unification functions dynamic rule generation overcome knowledge-gap problem prevalent in traditional reasoners. In this paper, describe algorithms used implementation distributed task-based framework builds proof/explanation graphs input query. We simple QA example from children’s story motivate Braid’s design explain how various components work together produce coherent explanation. Finally, evaluate Braid ROC Story Cloze test achieve close state-of-the-art results providing frame-based explanations.