Plotting Poetry paper accepted
"Controlling Style in Machine-generated Poetry", a paper jointly authored by Dr. Steffen Eger, Jörg Wöckener, and Luise Borek (all of TU Darmstadt), Thomas Haider (Max Planck Institute for Empirical Aesthetics), and Dr. Tristan Miller (OFAI), has been accepted for presentation at the international conference Plotting Poetry (and Poetics) 3.
The paper focuses on style-conditioned poetry generation, where the user specifies stylistic and logological aspects, such as sentiment and alliteration, for better control over the generated form and content. We find that deep models can be successfully conditioned on binary sentiment, learning to generate words of positive or negative sentiment when conditioned on same.
The conference will be held from September 26 to 27, 2019 at ATILF in Nancy.