Date created: | Last Updated:

Category: Project

Description: Lodestar is an interactive visualization notebook that allows the user to perform visual data analysis by selecting from a list of recommendations. Selecting a recommendation adds the corresponding Python code to the notebook and executes it, thus generating new output; no programming expertise is required on behalf of the user. The recommendation mechanism is inspired by autocomplete mechanisms, where a partial query is used to show suggestions for how to finish the query. The recommendations are derived from a directed graph of analysis states: one manually curated from online data science tutorials, another by automatically analyzing the Python code from Jupyter notebooks on GitHub.