Document representation as Generalized Latent Semantic Analysis (GLSA) vectors were claimed to give performance improvement on several task such as synonymy test, document classification, and clustering compared to traditional Latent Semantic Analysis (LSA) based systems, however GLSA performance has never been tested on automated essay grading system. This experiment propose an GLSA based automatic essay grading system design that will be used to examine the effect of GLSA implementation on automated essay grading system and to evaluate its performance compared to LSA based system.