We implement the FactReasoner 2.0 ideas, basically detecting logical inconsistencies in the response (see slides for examples), basically time alignment issues, logical contradictions between atoms or clusters of atoms, etc.
We capture all these in a probabilistic logic representation (see slides).
Inference will be done via a belief propagation type of algorithm (see slides).
We implement the FactReasoner 2.0 ideas, basically detecting logical inconsistencies in the response (see slides for examples), basically time alignment issues, logical contradictions between atoms or clusters of atoms, etc.
We capture all these in a probabilistic logic representation (see slides).
Inference will be done via a belief propagation type of algorithm (see slides).