Imitations of Immortality (2020). Learning to generate poetry in the style of the poet can make models style experts, but humans who create imitative works take a more general approach that incorporates domain knowledge outside the poet’s particular style. Instead of learning from a large corpus of one poet’s works, can a machine imitate the deep structure of her style using only one example of her work? To explore generating poetic variations, I wrote 8 poems that imitated the structure and language of 8 diverse poets, and used them to fine tune a transformer model that has seen only one poem by each of the authors. The poetic art works presented show structures borrowing from the human imitation in addition to prompted content of the original work, suggesting the model has learned some aspects of how humans write variations on content by imitating style.
See the poems on this live webpage and the accompanying paper here