Visualizing a Tabular Data Repository to Facilitate Descriptive Tag Augmentation for New Tables

Jianhao Cao, Tamara Munzner, Rachel Pottinger

Abstract: Many online tabular datasets are maintained in centralized repositories and annotated with descriptive tags. These tags are helpful for data practitioners to search and understand tables. However, manually annotating descriptive tags for new tables added to a large repository is expensive and may be inconsistent. In this extended abstract, we propose tag inference methods and implement an interactive visual explainer prototype to visualize a table repository with respect to a new table and to help a human user examine whether a recommended tag is suitable for the new table.