Publikationen

Abstract

Visualization research has always stressed the need for visual tools for data exploration and sense making. Despite the fact that many visualization technologies are available nowadays, their application in modern data science workflows is limited. One of the manifold reasons behind this is the development of visual analytics tools as standalone applications, featuring the complete pipeline from data loading to visualization. Other tools are targeted towards specific use cases (e.g., data wrangling), but to not provide standardized interfaces for import and export. This does not reflect the approach of stitching together several tools as it is employed in data science workflows nowadays. In this paper we outline the differences between standalone tools and notebook-style workflows for a specific use case for time series analysis. The outcomes demonstrate the benefits of notebook- style interfaces for tracking the steps in a data analysis workflow in a narrative way, for reporting, and for collaboration. We therefore argue that not considering the current developments towards the increased application of notebook-style interfaces for data science will lead to a reduced application and acceptance of visualization techniques in these domains. We outline the barriers for the integration of visualization techniques in narrative workflows, and describe directions for future research.