Natural Language Control for for Visually Interactive Decision Support Tools in Supply Chain Management

Abstract

Supply chains are complex networks where changing one variable can have unforeseen effects on the entire chain. Interactive supply chain visualizations are useful for understanding these effects, and can lead to decreased cost. However, these interactive visualizations can require technical and domain expertise to operate and understand. A solution for this is natural language interfaces, allowing users to use natural language commands to control the visualization. Additionally, natural language interfaces can be difficult to implement, and require applications specific programming or training. This thesis proposes integrating a pre-trained large language model as the natural language interface. An example application is created using an existing supply chain network visualization application. Various large language models are then evaluated for usability, functionality, and accuracy. We find that a state of the art commercial model is able to practically fulfill the role of a natural language interface, but that open-source large language models are not currently capable of functioning in this way.