A workshop held in conjunction with John J. Camilleri's PhD thesis defence.
Organised by Gerardo Schneider (main host) and John J. Camilleri (web page, program), with the support of the REMU project.
Slots are 50 mins each (30–40 mins talk + questions), but the schedule can flex a little as needed.
The paper adapts the syntax of the AceRules system, an existing controlled natural language (CNL), and pipes the resulting output rules to a defeasible inference engine. With this, we can represent and reason with knowledge bases (KB) with strict and defeasible rules. For strict rules, the consequent of a strict rule must hold if the premises hold; for a defeasible rule, the consequent of a defeasible rule ought to hold if the premises hold. Strict rules are always applied whenever applicable, while defeasible rules are maximally applied (in terms of rule subsets), subject to consistency. In the CNL, defeasible rules are expressed in terms of what is “usual”. The CNL expressions of rules are parsed and semantically represented so as to be used by an inference engine to accurately generate possible states of affairs. A verbaliser expresses the results in natural language, which facilitates understanding. The approach to defeasible rules is contrasted with available approaches, e.g. negation-as-failure, exceptions, and the use of abnormality predicates. The work is broadly set in the context of argumentation. The paper overviews the underlying formalism and the CNL; it provides an extended example.
Meng will briefly introduce and report on various threads of work in progress:
and outline possibilities for collaboration along the above threads.
Meng Wong is currently a CodeX Fellow at Stanford working on CompK.stanford.edu, and concurrently one of the co-founders of Legalese.com, which aims to translate recent academic research into contract and legal formalisation toward industrial application via an opensource startup.