Inspired by the 2019 Holy Grail Challenge, we present ZebraTutor, a mostly automated tool that solves a logic grid puzzle (also known as a Zebra puzzle) given the clues in natural language clues and a list of the entities in the puzzle. The challenge is to both handle the natural language input, and to produce a human-understandable explanation of how the solution is obtained. We achieve this by translating the natural language clues into logic using a typed version of the semantical framework of Blackburn and Bos. The logical representation is then used in a novel explanation-based reasoning procedure, on top of the IDP knowledge base system. A
novel aspect of the explanation is that it is ordered by mental
effort required to understand the reasoning step, which is estimated by the number of previously derived facts needed to derive new facts. The outcome is a stepwise visualisation of the clue(s) used and the resulting changes on the grid. This can be used both to solve a puzzle, or as a step-wise ‘help’
function for people stuck while solving a puzzle.
|Unpublished - 7 Nov 2019
|BNAIC 2019 - Brussels, Belgium
Duration: 7 Nov 2019 → 8 Nov 2019
|7/11/19 → 8/11/19