Synthetic Literature: Writing Science Fiction in a Co-Creative Process

TY - CHAP

T1 - Synthetic Literature

T2 - Writing Science Fiction in a Co-Creative Process

AU - Manjavacas, Enrique

AU - Karsdorp, F.B.

AU - Burtenshaw, Ben

AU - Kestemont, Mike

PY - 2017/9

Y1 - 2017/9

N2 - This paper describes a co-creative text generation system applied within a science fiction setting to be used by an established novelist. The project was initiated as part of The Dutch Book Week, and the generated text will be published within a volume of science fiction stories. We explore the ramifications of applying Natural Language Generation within a cocreative process, and examine where the cocreative setting challenges both writer and machine. We employ a character-level language model to generate text based on a large corpus of Dutch novels that exposes a number of tunable parameters to the user. The system is used through a custom graphical user interface, that helps the writer to elicit, modify and incorporate suggestions by the text generation system. Besides a literary work, the output of the present project also includes user-generated meta-data that is expected to contribute to the quantitative evaluation of the text-generation system and the co-creative process involved.

AB - This paper describes a co-creative text generation system applied within a science fiction setting to be used by an established novelist. The project was initiated as part of The Dutch Book Week, and the generated text will be published within a volume of science fiction stories. We explore the ramifications of applying Natural Language Generation within a cocreative process, and examine where the cocreative setting challenges both writer and machine. We employ a character-level language model to generate text based on a large corpus of Dutch novels that exposes a number of tunable parameters to the user. The system is used through a custom graphical user interface, that helps the writer to elicit, modify and incorporate suggestions by the text generation system. Besides a literary work, the output of the present project also includes user-generated meta-data that is expected to contribute to the quantitative evaluation of the text-generation system and the co-creative process involved.

M3 - Contribution to conference proceedings

SP - 29

BT - Proceedings of the Workshop on Computational Creativity in Natural Language Generation (CC-NLG 2017)

PB - Association for Computational Linguistics (ACL)

CY - Santiago de Compostela, Spain

ER -