Table Scraps: An Actionable Framework for Multi-Table Data Wrangling From An Artifact Study of Computational...

Area of Science:

• Computational Journalism
• Data Science
• Information Science

Background:

• Data wrangling is crucial for journalists using data in news reporting.
• Existing research on data wrangling primarily focuses on enterprise data analysis, leaving a gap in understanding journalistic practices.
• Journalists face unique challenges and pain points in their data wrangling workflows.

Purpose of the Study:

• To investigate the specific data wrangling operations and processes used by journalists.
• To identify the pain points encountered by journalists during data wrangling.
• To develop a framework for multi-table data wrangling tailored for computational journalism.

Main Methods:

• Conducted a technical observation study of 50 public repositories of data and analysis code.
• Analyzed code from 33 professional journalists across 26 news organizations.
• Developed two detailed taxonomies: one for data wrangling actions and one for processes.

Main Results:

• Identified extensive use of multiple tables in journalistic data wrangling, a factor often overlooked in prior analyses.
• Developed two comprehensive taxonomies detailing data wrangling actions and processes specific to computational journalism.
• Created a novel, actionable framework for general multi-table data wrangling, treating tables as first-class objects.

Conclusions:

• The developed framework addresses a gap in multi-table data wrangling, particularly for computational journalism.
• The new taxonomies and framework can inform the design of future interactive data wrangling tools.
• This work provides foundational insights into the data wrangling needs of journalists and supports the advancement of computational journalism practices.