Type-directed synthesis of visualizations from natural language queries

We propose a new technique based on program synthesis for automatically generating visualizations from natural language queries. Our method parses the query into refinement type specification using intents-and-slots paradigm and leverages type-directed to generate set of visualization programs that are most likely meet user's intent. system captures useful hints present in allows algorithm reject violate well-established design guidelines input data set. have implemented our ideas tool called Graphy evaluated it NLVCorpus, which consists 3 popular datasets over 700 real-world experiments show significantly outperforms state-of-the-art tools, including transformer rule-based ones.