The Eyes Have It: A Task by Data Type Taxonomy for Information Visualizations Ben Shneiderman, 1996
Background the growth of computing + graphic user interface 1987 scientific visualization 1989 information visualization more and more data types and tasks order terms information retrieval (in bibliographic systems) database management (in structured database systems) newer notions information gathering, seeking, visualization data mining, warehousing, filtering 0016046 Pei Shan Tsai 1 Introduction 2
Information Exploration Goals known item search find a set of items that satisfy a well-understood information need browse developing an understanding of unexpected patterns 0016046 Pei Shan Tsai 1 Introduction 3
Information Exploration Challenges the growth of data size data overload problem how to present data in orderly provide interaction with user 0016046 Pei Shan Tsai 1 Introduction 4
Data Presentation Design utilize humans perceptual abilities visual perceptual ability is greater than the other senses data visualization 0016046 Pei Shan Tsai 2 Visual Information Seeking Mantra 5
Data Visualization scientific visualization make 3d phenomena visible and comprehensible e.g. heat conduction, airflow information visualization reveal patterns, clusters, gaps, or outliers in abstract data e.g. stock market trades, document collection 0016046 Pei Shan Tsai 2 Visual Information Seeking Mantra 6
Visual Information Seeking Mantra overview first, zoom and filter, then details-on-demand basic visual design principle a starting point in trying to characterize information visualizations 0016046 Pei Shan Tsai 2 Visual Information Seeking Mantra 7
Task by Data Type Taxonomy 7 data types organized by the problems users are trying to solve 1d, 2d, 3d, temporal, multi-dimensional, tree, network data 7 tasks actions that users wish to perform overview, zoom, filter, details-on-demand, relate, history, extract 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 8
1d Data e.g. textual document, source code, alphabetical list features linear data items are organized in a sequential manner the problems users are trying to solve basic tasks count, search, select, etc. 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 9
1d Data Visualization Example bifocal display (Spence and Apperley, 1982) detailed information in the focus area less information in the surrounding area Figure 1: an information space Figure 2: wrap around two uprights Figure 3: view from an appropriate direction 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 10
1d Data Visualization Example SeeSoft (Eick et al., 1992) Figure 1: SeeSoft shows locations of characters within a text 0016046 Pei Shan Tsai 1 Introduction 11
2d Data e.g. geographic map, floorplan, newspaper layout features planar data items cover some part of the total area items may be rectangular or not the problems users are trying to solve deal with 2d space relationship between items up/down, left/right, inside/outside basic tasks 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 12
2d Data Visualization Example multilayer approach Figure 1: a geographic information systems as a layered cake 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 13
3d Data e.g. molecule, human body, building features real world objects items with volume items may have complex relationship between each other the problems users are trying to solve deal with 3d space relationship between items up/down, left/right, front/back, inside/outside basic tasks 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 14
3d Data Visualization Example visible human project (North et al., 1996) Figure 1: a poster of National Library of Medicine s Visible Human Project 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 15
Temporal Data e.g. medical history, project management, video features time lines items have a begin and end time items may overlap the problems users are trying to solve deal with time at, before, after some moment during some time period basic tasks 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 16
Temporal Data Visualization Example perspective wall (Robertson et al., 1993) Figure 1: perspective wall with the bifocal display 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 17
Temporal Data Visualization Example LifeLines (Plaisant et al., 1996) Figure 1: LifeLines shows personal medical history 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 18
Multi-dimensional Data e.g. most relational databases, statistical databases features items with n attributes is in a n-dimensional space the problems users are trying to solve deal with relationship between items patterns, clusters, gaps, outliers deal with relationship between attributes correlation, dependence basic tasks 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 19
Multi-dimensional Data Visualization Example parallel coordinates (Inselberg, 1985) Figure 1: parallel coordinates for iris dataset 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 20
Tree Data e.g. computer directory file management features hierarchy items have a link to one parent item, except the root links may have attributes the problems users are trying to solve deal with tree structural properties depth, degree, path, etc. basic tasks 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 21
Tree Data Visualization Example indented outlines (Egan et al., 1989) Figure 1: varied indented outline styles 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 22
Tree Data Visualization Example cone trees (Robertson et al., 1993) Figure 1: cone tree 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 23
Network Data e.g. World Wide Web features items have links to an arbitrary number of other items links may have attributes the problems users are trying to solve deal with network structural properties cycle, shortest path, etc. basic tasks 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 24
Network Data Visualization Example Netmap Figure 1: Netmap layout of nodes on a circle with links crisscrossing the central area 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 25
Task by Data Type Taxonomy 7 data types organized by the problems users are trying to solve 1d, 2d, 3d, temporal, multi-dimensional, tree, network data 7 tasks actions that users wish to perform overview, zoom, filter, details-on-demand, relate, history, extract 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 26
Overview gain an overview of the entire collection pan or scroll through the collection Figure 1: pan icons Figure 2: scroll icon 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 27
Zoom zoom in on items of interest control zoom focus and zoom factor Figure 1: zoom in icon Figure 2: zoom out icon 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 28
Overview and Zoom Strategy both change views of data overview plus detail view / context plus focus view field-of-view box Figure 1: field-of-view box with the bifocal display 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 29
Overview and Zoom Strategy both change views of data overview plus detail view / context plus focus view fisheye (Furnas, 1986) Figure 1: fisheye with the bifocal display 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 30
Filter filter out uninteresting items apply dynamic queries to items to control the contents of display rapid display update < 100 milliseconds 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 31
Details-on-demand select an item or group and get details when need click on to get a popup window with details 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 32
Relate view relationships among items highlight items that have similar attributes to the selected one 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 33
History keep a history of actions to support undo, replay, and progressive refinement since information exploration is rarely finished in a single step keeping the history of users actions for retracing Figure 1: undo icon Figure 2: redo icon 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 34
Extract allow extraction of sub-collections and of the query parameters extract the desired items for further use save, send, print, etc. 0016046 Pei Shan Tsai 3 Task by Data Type Taxonomy 35
Dynamic Query direct-manipulation query the use of rapid, incremental, and reversible actions sliders, buttons the immediate display of feedback Boolean expressions OR within an attribute AND of attributes across attributes (attr1.qry1 attr1.qry2) (attr2.qry1 attr2.qry2) 0016046 Pei Shan Tsai 4 Advanced Filtering 36
Boolean Expressions Visualization Venn diagrams (Michard, 1982), decision tables (Greene et al., 1990) become confusing as query complexity increases filter-flow model the metaphor of water flowing flow from left to right through a series of pipes and filters Figure 1: Venn diagrams Figure 2: decision table 0016046 Pei Shan Tsai 4 Advanced Filtering 37
Filter-flow Model AND a linear sequence of filters OR within an attribute, a single filter across multiple attributes, parallel filters NOT an inverter in filters Figure 1: (x > 7) (x > 9) Figure 2: (x < 7) (x > 9) Figure 3: (x > 9) (y > 9) Figure 4: (x > 7) 0016046 Pei Shan Tsai 4 Advanced Filtering 38
Summary while research prototypes typically deal with only one data type, commercial products will have to accommodate several and support full tasks although the computer contributes to the information explosion, it also provide powerful capabilities to deal with dynamic queries and data visualization 0016046 Pei Shan Tsai 5 Summary 39