TOWARDS MACHINE LEARNING APPROACHES TO IDENTIFY SHRIMP DISEASES BASED ON DESCRIPTION