Answer Validation Through Robust Logical Inference

The paper features MAVE, a knowledge-based system for answer validation through deep linguistic processing and logical inference. A relaxation loop is used to determine a robust indicator of logical entailment. The system not only validates answers directly, but also gathers evidence by proving the original question and comparing results with the answer candidate. This method boosts recall by up to 13%. Additional indicators signal false positives. The filter increases precision by up to 14.5% without compromising recall of the system.