A développement environment integrating algorithms, inferences and learning

Abstract

Most programming languages are based on context free grammars (CFGs). The purpose of grammatical inference is to infer a grammar, in our situation a CFG, from positive examples of sentences and possibly incorrect ones, for a given language. Based on these two fundamental definitions, we propose an environment followed by an implementation unifying different aspects of programming in machine learning settings. The central idea of this work is to use grammatical inference (GI) as a unifying framework for achieving this integration. Because any program can be considered as a string of characters, we show that the use of grammatical inference can not only unify different aspects of programming but also extend to wider areas of applications. The work sums up the following contributions: • State of the art of language theory and of grammatical inference; • Design and development of an environment integrating machine learning and first-order logic (FOL); • Design and development of a FOL system for parsing sentences independently or with a learning module; • Design and development of a heuristics-based polynomial-time complexity algorithm enhancing the learning process in grammatical inference. • Interaction between grammatical inference and control systems. The present work bears a promising line of research, contributing further to programming languages integration, aiming at the improvement of these languages with a machine learning layer. ACM Categories and Subject Descriptors D.3.1 [Formal definitions and theory], D.3.2 [Language classifications], Design languages, F.4.2 [Grammars and other rewriting systems], Parsing, F.4.3 [Formal Languages], I.2 [Artificial intelligence], I.2.3 [Deduction and theorem proving], Inference engine, I.2.6 [Learning], Language acquisition.