Combining Natural Logic and Shallow Reasoning for Question Answering

Gabor Angeli, Neha Nayak, Chris Manning

Abstract: Broad domain question answering is of-ten difficult in the absence of structuredknowledge bases, and can benefit fromshallow lexical methods (broad coverage)and logical reasoning (high precision).We propose an approach for incorporatingboth of these signals in a unified frame-work based on natural logic. We extendthe breadth of inferences afforded by nat-ural logic to include relational entailment(e.g.,buy→own) and meronymy (e.g.,a person born in a city is born the city’s country). Furthermore, we train aneval-uation function– akin to gameplaying –to evaluate the expected truth of candidatepremises on the fly. We evaluate our approach on answering multiple choice sci-ence questions, achieving strong results onthe dataset.