PlotThread: Creating Expressive Storyline Visualizations using Reinforcement Learning

Storyline visualizations may well be utilised to existing fictions, assembly material or software program evolutions. Having said that, it is hard to design these types of visualizations and existing software program has only limited design possibilities and layout adaptability.

A recent review suggests applying reinforcement mastering to build a device which facilitates the straightforward development of storyline visualizations. The agent is experienced to master how designers typically make storyline layout clear. As an alternative of pure automation, this device seamlessly integrates the get the job done of computational agents and people today on a shared trouble.

Image credit history: Tan Tang et al. / arXiv:2009.00249

For the duration of the interviews with design industry experts, it was stated that the device primarily based on synthetic intelligence balances the aesthetic plans and the narrative constraint far more successfully than systems primarily based only on the optimization.  Moreover, a researcher on visual analytics found that storylines produced with the novel device can arouse the emotion of viewers.

Storyline visualizations are an efficient usually means to existing the evolution of plots and expose the scenic interactions amongst people. Having said that, the design of storyline visualizations is a hard undertaking as customers will need to stability involving aesthetic plans and narrative constraints. In spite of that the optimization-primarily based strategies have been enhanced substantially in phrases of producing aesthetic and legible layouts, the current (semi-) computerized strategies are nevertheless limited pertaining to one) efficient exploration of the storyline design house and two) flexible customization of storyline layouts. In this get the job done, we suggest a reinforcement mastering framework to teach an AI agent that assists customers in checking out the design house effectively and producing nicely-optimized storylines. Primarily based on the framework, we introduce PlotThread, an authoring device that integrates a set of flexible interactions to help straightforward customization of storyline visualizations. To seamlessly combine the AI agent into the authoring course of action, we hire a blended-initiative method where each the agent and designers get the job done on the exact same canvas to improve the collaborative design of storylines. We consider the reinforcement mastering model through qualitative and quantitative experiments and show the utilization of PlotThread applying a selection of use circumstances.

Backlink: https://arxiv.org/stomach muscles/2009.00249