13:00
14:00

The first question when deploying a program synthesis approach is the design of the target programming language, called domain-specific language (DSL). This raises a number of challenges: the language must be rich and expressive, while using a simple and minimal syntax. We study language sharpening, which is the task of improving a DSL by removing redundant and useless programs. We propose a rule-based approach for language sharpening, and present two technical contributions. The first is an algorithm for automatically generating equations satisfied by all programs, which we use to significantly reduce the space of all programs. The second is a new compilation algorithm for our rules-based approach, based on tree automata models. We integrate both algorithms in a program synthesis pipeline to evaluate the benefits of language sharpening in two application domains.

salle 178, zoom: https://bordeaux-inp-fr.zoom.us/j/89508306159