Parsing Hypergraphs using Context-Free Positional Grammars
By: Gennaro Costagliola, Federico Vastarini
Potential Business Impact:
Helps computers understand complex drawings and rules.
We present a novel work-in-progress approach to the parsing of hypergraphs generated by context-free hyperedge replacement grammars. This method is based on a new LR parsing technique for positional grammars, which is also under active development. Central to our approach is a reduction from hyperedge replacement to positional grammars with additional structural constraints, enabling the use of permutation-based operations to determine the correct ordering of hyperedges on the right-hand side of productions. Preliminary results also reveal a distinction between ambiguity in graph generation and ambiguity in graph recognition. While the exact class of hyperedge replacement languages parsable under this method remains under investigation, the approach provides a promising foundation for future generalisations to more expressive grammar formalisms. Graph parsing remains a broadly relevant problem across numerous domains, and our contribution aims to advance both the theoretical and practical understanding of this challenge.
Similar Papers
GramTrans: A Better Code Representation Approach in Code Generation
Software Engineering
Makes computers write better code by simplifying its rules.
Unraveling Syntax: How Language Models Learn Context-Free Grammars
Computation and Language
Helps computers learn language rules better.
A New Graph Grammar Formalism for Robust Syntactic Pattern Recognition
Formal Languages and Automata Theory
Finds hidden patterns in messy pictures.