Grammatical inference of graph grammars for syntactic pattern recognition

This paper presents a new approach to the use of graph grammars for syntactic pattern recognition. First an introduction concerning the representation of complex two or three-dimensional patterns with labelled nodes and edges is given. Then follows the definition of a special class of graph grammars with rules producing chainand starstructured subgraphs. A constructive inference algorithm is developed which uses the knowledge contained in the actual grammar for the productions needed for the next sample pattern. For the same class of graph grammars, a nondeterministic syntax analysis algorithm is described which analyzes new patterns pursuing the same strategy as the inference algorithm. Some remarks about the possibilities of extending this approach to stochastic and/or programmed grammars are added. Finally some examples from an implementation of the graph grammar inference and analysis system are given.