A Framework to build Abductive-Deductive Chatbots, based on Natural Language Processing and First-Order Logic

This paper presents a framework based on natural language processing and first-order logic, which implicitly simulate the human brain features of selecting properly information related to a query from a knowledge base (abductive pre-stage), before to infer new knowledge from such a selection acting as deductive database. Such features are used with the aim of instantiating cognitive chatbots, able of human-like fashioned reasoning, supported by a module which automatically transforms polar and wh-questions into one or more likely assertions, in order to infer Boolean values or snippets with variable length as factoid answer from a conceptual knowledge base. The latter is splitted into two layers, representing both long- and short-term memory, and the transition of information between the two layers is achieved leveraging both a greedy algorithm and the engine's features of a NoSQL database, with promising timing performance than respect using one layer. Furthermore, such chatbots don't need any scripts updates or code refactory when new knowledge has to income, but just the knowledge itself in natural language.