Genetic algorithm with modified reproduction operators for grammatical inference.

Abstract

A grammatical inference (GI) algorithm is proposed, that utilizes a Genetic Algorithm (GA), in conjunction with a pushdown automaton (PDA) and the principle of minimum description length (MDL). GI is a methodology to infer context-free grammars (CFGs) from training data. It has wide applicability across many different fields, including natural language processing, language design, and software engineering. GAs is a search methodology that has been used in many domains and we utilize GAs as our primary search algorithm. The proposed algorithm incorporates a Boolean operator-based crossover and mutation operator with a random mask. Here, Boolean operators (AND, OR, NOT, and XOR) are applied as a diversification strategy. A PDA simulator is implemented to validate the production rules of a CFG. The performance is evaluated against state-of-the-art algorithms. Statistical tests demonstrate the superiority of the proposed algorithm over the algorithms implemented in this paper.