Storyline visualizations might be made use of to existing fictions, meeting information or application evolutions. On the other hand, it is tough to style and design this kind of visualizations and current application has only limited style and design alternatives and structure flexibility.
A recent review implies utilizing reinforcement learning to develop a software which facilitates the effortless generation of storyline visualizations. The agent is qualified to master how designers commonly make storyline structure very clear. Alternatively of pure automation, this software seamlessly integrates the get the job done of computational brokers and persons on a shared difficulty.
Through the interviews with style and design authorities, it was stated that the software centered on artificial intelligence balances the aesthetic goals and the narrative constraint much more effectively than techniques centered only on the optimization. Moreover, a researcher on visual analytics recognized that storylines produced with the novel software can arouse the emotion of viewers.
Storyline visualizations are an effective usually means to existing the evolution of plots and reveal the scenic interactions between figures. On the other hand, the style and design of storyline visualizations is a tough task as people will need to balance in between aesthetic goals and narrative constraints. Regardless of that the optimization-centered techniques have been enhanced significantly in terms of generating aesthetic and legible layouts, the existing (semi-) computerized techniques are nonetheless limited concerning one) productive exploration of the storyline style and design place and 2) flexible customization of storyline layouts. In this get the job done, we suggest a reinforcement learning framework to coach an AI agent that assists people in checking out the style and design place efficiently and creating well-optimized storylines. Based mostly on the framework, we introduce PlotThread, an authoring software that integrates a established of flexible interactions to guidance effortless customization of storyline visualizations. To seamlessly integrate the AI agent into the authoring process, we use a mixed-initiative approach exactly where the two the agent and designers get the job done on the same canvas to enhance the collaborative style and design of storylines. We appraise the reinforcement learning model by qualitative and quantitative experiments and exhibit the utilization of PlotThread utilizing a assortment of use conditions.
Hyperlink: https://arxiv.org/stomach muscles/2009.00249