Abstract:
Understanding the semantics of data wrangling scripts is a common requirement for data workers. However, due to the complexity and diversity of the types of data transformations and their parameters, understanding the semantics of wrangling scripts is time-consuming and laborious. We worked with four data workers and developed an interactive visualization system, called ChangeVis, to help data workers understand the changes of tables in the process of data wrangling. Based on the Focus+Context technique, ChangeVis scales well and illustrates the changes of table structure, the semantics of data transformations, the changes of column information, and the changes of cell data on a selected code snippet. We used two use cases and a semi-structured lab study to demonstrate the usability and effectiveness of ChangeVis.