Bechtler et. al. [10] allocated 90% while Hosoz and Ertunc [7] reserved 70% of total data to train the ANN modal. In this study, it was decided to use 90% of total data for training and the remaining 10% for testing. The testing sets are randomly selected and are used to test the ability of the trained ANN model to generalize, which refers to how well the model can infer the behavior of the actual AAC system for conditions represented in the data not used for training [14]