Key points are not available for this paper at this time.
Regular and context-free languages form a central pillar of formal language theory. This is because a variety of formalisms are known that define these classes of languages. For example, we have that finite automata, monoids, algebraic recognizability, regular expressions, regular grammars, monadic-second order logic, etc. , can be used to represent regular word languages. However, the situation is less clear for formal languages over graphs, and open problems persist. This is because generalizing notions from words to graphs has been more successful for some of the cited formalisms than for the other ones. Bruno Courcelle has introduced hyper-edge replacement () algebras for generalizing the notion of context-free languages from words to graphs. At the same time, -algebras support the generalization of algebraic recognizability from words to graphs, a notion that has been proven to be equivalent to definability in (counting) monadic-second order logic () over graphs of bounded tree-width. In this paper, we deal with generalizing regular word grammars to graphs. We propose regular grammars for (unordered and unranked) trees, series-parallel graphs, and graphs of tree-width 2, where the qualifier regular is justified because these grammars define exactly the recognizable resp. -definable subsets of the respective graph classes.
Bozga et al. (Fri,) studied this question.