Contextual Graph Grammars for Generative Modeling
Year: 2026
Type: preprint
Abstract: This paper proposes a novel approach to generative modeling using Contextual Graph Grammars (CGGs). Traditional graph grammars operate with static rules, limiting their ability to generate diverse and realistic graph structures. We introduce CGGs, which extend standard graph grammars by incorporating contextual embeddings – typically derived from language models – directly into the generation rules. This allows the grammar to dynamically adapt its output based on the surrounding context, resulting in richer and more contextually appropriate graph structures. The core claim is that integrating contextual information significantly improves the diversity and realism of graph generation. The mechanism involves defining rules within the graph grammar that utilize these contextual embeddings, effectively guiding the generation process. This represents a shift from static graph grammars to a dynamic system influenced by contextual factors. We explore the theoretical foundations of CGGs and outline a framework for their implementation, demonstrating the potential for generating complex graph structures for various applications.
Author: Zhang, Jincheng
Language: English
Topic: Graph Theory and Algorithms
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7207515924
