Organizers: Franz Berto, Luca Incurvati, Julian Schlöder
Organizers: Jakub Szymanik, Willem Zuidema
Reasoning in natural language – including entailment, contradiction, implicatures – is studied in a variety of fields. In formal semantics, accounting for how we reason with various language fragments is a central challenge, usually approached by using some expressive variety of logic. In computational linguistics, recognizing textual entailment in particular has become a central topic, considered a key component of an integral natural language processing system. The dominating approach here is based on statistical inference (with probability distributions defined over relatively simple symbolic ‘backbones’), with more recently increasing interest in vectorial and neural models. In this workshop we will address the relation between these two worlds of symbolic and sub-symbolic approaches to NL reasoning, and discuss the relevance of various models for the cognitive science of human reasoning and natural language understanding. The workshop will highlight new interfaces between logic, computational semantics, psychology of reasoning, and computer implementations of natural reasoning systems.