The Learning-Knowledge-Reasoning Paradigm for Natural Language Understanding and Question Answering

Given a text, several questions can be asked. For some of these questions, the answer can be directly looked up from the text. However for several other questions, one might need to use additional knowledge and sophisticated reasoning to find the answer. Developing AI agents that can answer these kinds of questions and can also justify their answer is the focus of this research. Towards this goal, we use the language of Answer Set Programming as the knowledge representation and reasoning language for the agent. The question then arises, is how to obtain the additional knowledge? In this work we show that using existing Natural Language Processing parsers and a scalable Inductive Logic Programming algorithm it is possible to learn this additional knowledge (containing mostly commonsense knowledge) from question-answering datasets which then can be used for inference