"Kurosawa": A Script Writer's Assistant: Related Work

Table of Links

• Abstract and Intro
• Motivation
• Related Work
• Dataset
• Experiments and Evaluation
• Results and Analysis
• Conclusion and Future Work
• Limitations and References
• A. Appendix

3. Related Work

3.1. Automatic Story Generation

Neural models have been able to produce stories by conditioning on different contents like visuals (Huang et al., 2016) and succinct text descriptions (Jain et al., 2017). Work on plot controllable, plandriven story generation abounds (Riedl and Young, 2010; Fan et al., 2019; Pérez and Sharples, 2001; Rashkin et al., 2020). A related kind of work is automatic poetry generation based on keywords or descriptions (Yan, 2016; Wang et al., 2016).

3.2. Plot Generation

Plot Machines (Rashkin et al., 2020) generate multiparagraph stories based on some outline phrases. Fan et al. (2018) introduce a hierarchical sequenceto-sequence fusion model to generate a premise and condition that in turn generate stories of up to 1000 words. This work- unlike ours- is non-neural and template-driven and is, therefore, much less creative and novel compared to what we generate.

3.3. Scene Generation

Automatic scene or script generation has received comparatively less attention. Dialogue generation (Li et al., 2016; Huang et al., 2018; Tang et al., 2019; Wu et al., 2019) with a semblance of scene generation has been done. There has recently been some work focusing on guiding dialogues with the help of a narrative (Zhu et al., 2020). We generate scenes in which the main elements come from a small prompt as input.