• Summary
• Related Documents
  • Active Bibliography
  • Co-citation
• Version History

Abstract:

Abstract-- Sentences of visual languages may often be regarded as assemblies of pictorial objects like "circles", "arrows " or "strings " with spatial relations like "above " or "contains " between them, i.e. their underlying structure is a kind of directed graph. Therefore, graph grammars are a natural means for defining the syntax of visual languages. Their main drawback until now is the lack of general enough and efficiently working parsing algorithms. All published graph grammar or---more general--- visual language parsing algorithms are only able to deal with context-free graph grammars, where the left-hand side consists of a single nonterminal vertex only. This makes syntax definitions of visual languages hard to read, prohibits the use of complex pattern matching, and disallows graph-grammars which specify transformation processes. We have developed the first parsing algorithm for context-sensitive graph grammars which allows left- and right-sides of productions to be almost arbitrary graphs. The algorithm is divided into two phases, where the first one constructs bottom-up a set of all eventually useful production applications. The second one extracts top-down viable derivations from the computed set of production applications. This separation into two phases leads to more comprehensible algorithms. Furthermore, it allows for independent optimization efforts in the form of heuristics which reduce the algorithm's exponential time and space requirements dramatically for "real world " examples. 1

Citations