Keshif: Rapid and Expressive Tabular Data Exploration for Novices
General purpose graphical interfaces for data exploration are typically based on manual visualization and interaction specifications. While designing manual specification can be very expressive, it demands high efforts to make effective decisions, therefore reducing exploratory speed. Instead, princ...
Gespeichert in:
Veröffentlicht in: | IEEE transactions on visualization and computer graphics 2018-08, Vol.24 (8), p.2339-2352 |
---|---|
Hauptverfasser: | , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | General purpose graphical interfaces for data exploration are typically based on manual visualization and interaction specifications. While designing manual specification can be very expressive, it demands high efforts to make effective decisions, therefore reducing exploratory speed. Instead, principled automated designs can increase exploratory speed, decrease learning efforts, help avoid ineffective decisions, and therefore better support data analytics novices. Towards these goals, we present Keshif, a new systematic design for tabular data exploration. To summarize a given dataset, Keshif aggregates records by value within attribute summaries, and visualizes aggregate characteristics using a consistent design based on data types. To reveal data distribution details, Keshif features three complementary linked selections: highlighting, filtering, and comparison. Keshif further increases expressiveness through aggregate metrics, absolute/part-of scale modes, calculated attributes, and saved selections, all working in synchrony. Its automated design approach also simplifies authoring of dashboards composed of summaries and individual records from raw data using fluid interaction. We show examples selected from 160+ datasets from diverse domains. Our study with novices shows that after exploring raw data for 15 minutes, our participants reached close to 30 data insights on average, comparable to other studies with skilled users using more complex tools. |
---|---|
ISSN: | 1077-2626 1941-0506 |
DOI: | 10.1109/TVCG.2017.2723393 |