What is visualisation ?What is visualisation ?

• Visualise: (vb) to form a mental image or vision of …

• Cognitive ability

• Allows us to internalise data– Gain insight and understanding

• Internal Map = Cognitive Model

What are data types ?What are data types ?

• Various different types of data

• Numerical

• Ordinal– Naturally order ( days of the week )

• Categorical– Not ordered ( animal names )

IndentationIndentation• Tree controlTree control

• FisheyeFisheye

ContainmentContainment• Treemaps

• Pad++

IndentationIndentation• Tree controlTree control

• FisheyeFisheye

ContainmentContainment• Treemaps

• Pad++

ClusteringClustering

• Galaxy of NewsGalaxy of News

• ThemeScapeThemeScape

• Hot SauceHot Sauce

GeographicGeographic

• Floor plansFloor plans

• Street mapsStreet maps

ClusteringClustering

• Galaxy of NewsGalaxy of News

• ThemeScapeThemeScape

• Hot SauceHot Sauce

GeographicGeographic

• Floor plansFloor plans

• Street mapsStreet maps

Node-link diagramsNode-link diagrams

• 2D diagrams 2D diagrams

• SemNetSemNet

• Cone TreeCone Tree

• Fisheye Cone TreeFisheye Cone Tree

• Hyperbolic viewerHyperbolic viewer

• FSNFSN

• XML3DXML3D

Node-link diagramsNode-link diagrams

• 2D diagrams 2D diagrams

• SemNetSemNet

• Cone TreeCone Tree

• Fisheye Cone TreeFisheye Cone Tree

• Hyperbolic viewerHyperbolic viewer

• FSNFSN

• XML3DXML3D

Basic Visualization ApproachesBasic Visualization Approaches

Examples of VisualisationExamples of Visualisation

• London Underground – Harry Beck

• Connectivity

• Deals with connections, not focused on geography

• Differs from other maps, as familiar geography was not overriding concern

London Underground Map 1927London Underground Map 1927

London Underground Map 1990sLondon Underground Map 1990s

1855 London Cholera Epidemic

Broad StreetPump

Dr. John Snow:Dr. John Snow:Statistical Map VisualizationStatistical Map Visualization

Visualising Tree Data 1Visualising Tree Data 1

• CS use of trees for data storage

Visualising Tree Data 2Visualising Tree Data 2

• Difficult to visualise large tree structures

• Take a company– CEO as the root node– People reporting to him at next level– So on until all the employees are included

Tree Maps 1 – SchneidermanTree Maps 1 – Schneiderman

Tree Maps 2 – SchneidermanTree Maps 2 – Schneiderman

•Johnson & Schneiderman, University of Maryland, Vis’91

Space filling~~30003000 objects objects

•MicroLogic’s DiskMapper

Hyperbolic Browsing - LampingHyperbolic Browsing - Lamping

H3 - 1997H3 - 1997

Munzner, Stanford Univ., InfoVis’97Projected onto sphere: 20,000 nodes20,000 nodes

Information Visualisation in Information Visualisation in Information RetrievalInformation Retrieval

• on-line information • diversity of users of such resources• potential overload• establish new formats for the presentation and

manipulation of electronic data• spatial ability is an important predictor of

effectiveness and efficiency when performing common information (i.e. textual) search tasks

Usefulness of Visualisation in IRUsefulness of Visualisation in IR

• Allows semantic relationships to be represented

• Use of Metaphors such as– spatial proximity – visual links

• Allows users to develop a conceptual map of the information space

Linking IR to real world tasksLinking IR to real world tasks

• Searching & Browsing of information can be related to real world navigation

• Complex Datasets can hide trends / information– A well design graph can express shopping

trends through the use of Store Card information

IR and HypermediaIR and Hypermedia

• WWW – another information space• Overview Maps & Zooming/Panning• Improve performance and satisfaction• Move ‘load’ from cognitive to perceptual

processes• visualising and directly interact with

conventional hypermedia and unstructured text

Combing IR and VR – new Combing IR and VR – new perceptions of dataperceptions of data

• Virtual Reality (VR) environments can further enhance visualisations

• Allows for– Real Time Interactivity– Viewing of relationships between object from

unlimited number of perspectives– Can allow for haptic or non-visual methods of

feedback to the user

Visualization Taxonomy - 1994Visualization Taxonomy - 1994

•ImplicitImplicit (use of perspective)•Continuous focus and contextContinuous focus and context•FilteredFiltered (removing items of low interest)•Discrete focus and contextDiscrete focus and context•DistortedDistorted (size, shape, position of

elements)•AdornedAdorned (color, texture)

Reference: Noik (Graphics Interface’94)

Approaches to IVApproaches to IV

• Core approaches - Colebourne et al. (1994)

1. 'Benediktine' cyberspace

2. statistical clustering and proximity

3. hyper-structures

4. human centred

• Categories are not mutually exclusive

'Benediktine' cyberspace'Benediktine' cyberspace

• Benedikt - 1991

• assigns object attributes (e.g. file size, age, key words) on to extrinsic (x,y,z) and intrinsic (e.g. shape) dimensions.

• Well suited to data that is explicitly structured

'Benediktine' cyberspace'Benediktine' cyberspace

Statistical Clustering and Statistical Clustering and ProximityProximity

• Applies statistical models to data prior to presenting the visualisation

• conveys spatially the underlying semantic structure.

• spatial proximity of documents -> reflect their semantic similarity

• Various techniques generate these semantic proximities (eg Vector Space Model)

Statistical Clustering and Statistical Clustering and ProximityProximity

Hyper-structuresHyper-structures

• extend the notion of hypertext directly

• use 3-D graph drawing algorithms to create the visualisation

• Works well where explicit links exist, eg in hypertext

• Various graph visualisation techniques available

Hyper-structure (Cone Tree 1)Hyper-structure (Cone Tree 1)

Robertson, Mackinlay & Card, Xerox PARC, CHI’91

Limits:10 levels10 levels1000 nodes1000 nodesUp to 10,000Up to 10,000

Robertson, Mackinlay & Card, Xerox PARC, CHI’91

Limits:10 levels10 levels1000 nodes1000 nodesUp to 10,000Up to 10,000

Hyper-structure (Cone Tree 2)Hyper-structure (Cone Tree 2)

Human centredHuman centred

• Two main areas

1. Exploit the user's real world experience, by representing information spaces using real world metaphors

2. Allow the user themselves to organise the information in a manner that they find intuitive

Human centred – Exploit user Human centred – Exploit user experienceexperience

Human centred – User Human centred – User themselves organise datathemselves organise data

Visual Information Seeking 1Visual Information Seeking 1

• Research by Ben Schneiderman

• Direct-manipulation interfaces

• Certain tasks a visual presentation is much easier to comprehend than text

• Mantra: Overview first, zoom and filter, then details on demand

Visual Information Seeking 2Visual Information Seeking 2

• Schneiderman – 7 Data Types

• 1-, 2-, 3-d data, temporal, multi-dimensional, tree and network data

• All items have attributes and simple search task is to find all items which a certain set of attributes

Visual Information Seeking 3Visual Information Seeking 3

• Overview: of a collection

• Zoom: on items of interest

• Filter: out uninteresting items

• Details-on-Demand: of a item or group of items

• Relate: relationship between items

• History & Extract

Combining Sound & Visual Combining Sound & Visual retrievalretrieval

• Aural presentation contains addition information not found in visual representations

• Omni directional information• Encoding of information, multiple streams• “Cocktail Party Effect” - Arons 1992• Recognition of sounds, is most often sufficient to

hear only 500 ms to 2 seconds of the characteristic or significant part of a sound (Warren 1999)

Further ReadingsFurther Readings

• Chen, C. (1999) Information Visualisation and Virtual Environments

• Card, S et al (1999) Readings in Information Visualization: Using Vision to Think

• Spence, R. (2001) Information Visualization

• http://www.cribbin.co.uk/infovis.htm