The Interplay of Dataset Characteristics in Automated Grammar Generation: A Study with the AGGREGATION System