Optimizing Text Data in Deep Learning: An Experimental Approach