Please no. I am all for neutrality, but the underlying cause is the training dataset. Change that if you want different results, but do not alter artificially.