The two-stage pipeline can produce poems that transfer poorly to Shakespearean style, and the style-transfer dataset poorly represents poetic vocabulary.