Gene entity normalization, the mapping of a gene mention in free text to a unique identifier, is one of the primary subtasks in the biomedical information extraction pipeline. Gene entity normalization provides many challenges, specifically with the high ambiguity of gene names and the many-to-many relationship between gene names and identifiers. Drawing inspiration from recent work in word sense disambiguation, this paper presents a gene entity normalization system based on entity relationship graphs. This system creates a concept graph from the possible entities and their relationships within a full-text document, and takes advantage of a node ranking algorithm to rank and score each potential candidate entity. This system is a prototype to represent a specific approach to gene normalization, and the results reflect this. However, this system demonstrates that the relationship graph-based approach, an approach grounded in a theoretical basis, can potentially be useful for gene normalization and possibly for the normalization of various biomedical entities.