Story Generation Using Intermediate Plot Representation
Open Access
- Author:
- Karanam, Kavya Laalasa
- Graduate Program:
- Computer Science and Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- June 16, 2021
- Committee Members:
- Chitaranjan Das, Program Head/Chair
C Lee Giles, Thesis Advisor/Co-Advisor
Rui Zhang, Committee Member
Ting-Hao (Kenneth) Huang, Thesis Advisor/Co-Advisor - Keywords:
- Story Summary Block Generation
Natural Language Processing
Deep Learning
Storyline Planning - Abstract:
- Writing the stories had always been a challenging task as it is always possible to get confused when writing huge drafts with limited constraints. Automated story telling is process where AI is used to structure and generate long stories when certain lines of a story is given as an input. Even for human beings it is really a struggle to craft or come up with good stories especially long connected sentences which are meaningful . There have been numerous datasets that were previously experimented or tested in this creative task. However, most of datasets comprises simple and short paragraphs which contains approximately 5 to 7 sentences that restricts complexity that the machines have to deal with to learn the stories making the story generation task straightforward. This work investigates that for a given set of story blocks, if story generation models would be able to construct practical follow-up stories for realistic human-written long stories. All the experiments are performed on Book Corpus dataset which proved that the model is better at generating short summaries than longer summaries for the generated follow-up stories. Human Evaluation results indicate that stories are better ranked for generated summaries than the human-written ground truth stories.