The Many Dimensions of SHAPE The image above is from www.scholastic.com/ispy. The “I Spy” books by Walter Wick and Jean Marzollo are filled with visual images and riddles requiring visual search, typically using shape, in order to solve. “I Spy” books are available from Amazon, Barnes and Noble, etc. © 2009 Richard Brath except as otherwise noted. Richard Brath: [email protected] or 01-416-203-3003 ext 242 Page 1 The Many Dimensions of Shape Outline • Motivation •Shape? •Background • 6 Experiments using shapes on different datasets •A model? * All conjecture – nothing proven! © 2009 Richard Brath except as otherwise noted. Page 2 Shapes There are Birth Rate vs. Death Rate 1997 for 190 countries – each country assigned to a shape. 4-10 basic 30 shapes in most charts, paint 25 packages, etc. 20 te a ath r de 15 10 5 Why limited to 10? 0 What if you 0 102030405060 need more birth rate than 10? Excel chart. 10 shapes, then repeats (with different color)! © 2009 Richard Brath except as otherwise noted. Page 3 Pictographs • There are limitless number of pictographs. •But… – Design required – Abstract nouns, adjectives difficult to encode/decode – Ambiguity – Perceptually effectiveness? www.aiga.org/content.cfm/symbol-signs © 2009 Richard Brath except as otherwise noted. Page 4 What is Shape? A. Simple geometry: • What’s in-between these two? • Can shapes convey more than one (or two or 5) attribute(s)? • Are some shapes more perceptually effective? B. Pictographs: 1 t⌧☯ ZH][FYEPt A#a28<:9= Limitless visual representation, typically of nouns. © 2009 Richard Brath except as otherwise noted. Page 5 Intermediate Shapes… Alphanumerics Symbols More Symbols •A B C D E • & ! ( ) ; : @ ? • a b c d e •€£¥₣ • 1 2 3 4 5 •+ -/ x • i ii iii iv v vi • √∞≠≤≥≈ But: • Are these shapes perceptually efficient? • Can they convey quantitative values (i.e. magnitudes), not just separate categories? • Can they convey more than one attribute? © 2009 Richard Brath except as otherwise noted. Page 6 Compound Shapes… Pictograph Hair on 1,2,3…n on Basic a Base bars, charts & Geometry Shape small multiples. Edward Tufte: Beautiful Evidence Icons from AIGA Anymails: Visualization of Email Inbox www.aiga.org/content.cfm/symbol-signs Design & Concept: Carolin Horn Code: Florian Jenett • Are there ways to convey multiple attributes other than just welding component shapes together? © 2009 Richard Brath except as otherwise noted. Page 7 What are the potentials of Shape? Where are the limits? What are the dimensions? Basic Geometry Compound Symbols •What’s in here? Shapes Alphabets • “welded objects” • small multiples Pictographs © 2009 Richard Brath except as otherwise noted. Page 8 Many domains utilize shape for differentiation, aesthetics, search, identification, functions, etc. I Spy books: Walter Wick and Jean Marzollo. Scholastic Sample flower shapes. From Aircraft silhouette identification –LIFE Magazine 1944, Publishing. www.scholastic.com theseedsite.co.uk/flowershapes.html Andreas Feininger. Via Google LIFE image search. Leaf Morphology from A few of examples of more than 14,000 unique Wassily Kandinsky - Composition Example shape attributes used in bird wikimedia.org LEGO parts catalogued by fans on: peeron.com VIII - 1923 wassilykandinsky.net identification, from A Field Guide to the Birds by Roger Tory Peterson © 2009 Richard Brath except as otherwise noted. Page 9 Many Shape Techniques in InfoVis and SciVis but is there a formal list of shape attributes? Star Coordinates Sparklines Literary Organism Superquadrics www.bonavistasystems.com/OnlineDemoReports.html Whisker plot, Star plot, Exvis Stick (Kandogan, 2001) (Posavec 2007) Time Wheel Genome Valence From Colin Ware’s book. (Barr 1981) Info- Kindlmann Lotus (Fry, 2000) (Zhang (Chua & et al, Malwarez Eick, 1997) 2005) Compound Glyph authors. Dragulescu espective r Iconic Techniques for Voreen © Feature Visualization Glyphs (Brath, 1997) Blobby Text Bodies and (Chernov) Faces this page on ages m © 2009 Richard Brath except as otherwise noted. Frank Post Page 10 (Stephen Rose, ???) Rohrer 1998 All i ology h c 9x) Visual Attributes ) y 6) x) 5 ( s 8 9 8 ( en ) r x n ( d as defined by visualization 7) ay ( Table of Visual Attributes h 0 l 6 in ac and suitability for encoding a ( ptual P elan in ( K E researchers and perceptual z data according to various ce ilkinso e researchers. Bert Clev Mac Mac W Maz P psychologists Attribute CQ na CQCQCQCQ P Color Hue Yx Y YYY~Y Yx Y Brightness Y Y Y Y ~ Y Y x Y Y Saturation Y Y ~ Y Y Transformation Length/Width Y Y Y x Y Y Size (Area) YY Y YYYYY x ~ Y Volume Y Y Y Orientation Y x Y Y Y ~ Y x Y Y Slope Y Y Y Form Shape x x Y x Y Y Y x Added Marks Y x Curvature x ~ Y Concavity/Convexity x ~ Closure Y Intersection Y Terminators Y Holes Y Spatial 2D position YY Y YYYY ~Y Y Grouping/Containment Y ~ x x x Y 3D Depth x x Y Sources: Connection Y Numerosity x Y Y Bertin 67: Semiology of Graphics Shadow Direction Y Cleveland 85: The Elements of Graphing Data Movement/Optics Flicker ~ x Y MacKinlay 86: Automating the Design of Graphical Presentations Motion x ~ Y MacEachren 95: How Maps Work Transparency ~ ~ Y Blur/Crispness x ~ Y Wilkinson 99: The Grammar of Graphics Shininess Y Mazza 09: Introduction to Information Visualization Texture Yx YXYYY* Perceptual Psychology: scholarpedia.org/article/Visual_search C = Categoricial/Nominal scale (Bertin's Selection) Q = Quantitative/Ordinal scale. © 2009 Richard Brath except as otherwise noted. P = Perceptually preattentive, i.e. can preattentively distinguish between two categories. Page 11 Note: Researchers are not consistent, even in terminology Rotation/Angle/Orientation/Slope: • Wilkinson: Rotation • Cleveland: Angle • Bertin: Orientation • MacKinlay: Slope/Angle • MacEachren: Orientation •Mazza: Orientation Vary in orientation only. My opinion – Angle and orientation are two separate dimensions of visual attributes Vary in angle only. © 2009 Richard Brath except as otherwise noted. Page 12 Gaps in our knowledge about shape: Attribute BBinaryinary CCategoryategory Order/QuantityOrder/Qty Consensus. •SSize*ize* YYYY YY But these are uniform •OrientationOrientation YYYY N - transformations on shape, not about shape ----------------------------------------------------------------------------- itself. •SShapehape ∆◊O+? YY N N •AAddedng leMarks ?Y ? N •CCurvatureurvature YY NN-- •AngleAdded Marks Y N •CClosurelosure YY Incomplete •IIntenrsteecrstionection YY No consensus •TTerminatorserminators YY … •HHolesoles YY 1 2 Therefore: do a series of small experiments to 3 explore these attributes more.. © 2009 Richard Brath except as otherwise noted. Page 13 Experiment #1: Categories of Curves GOAL: • Use curves to represent AttributeAttribute BBinaryinary CateCategorygory Order/QtyOrder/Qty •S•Sizieze** YYYY YY different categories. •• OOririentationentation YYYY - - -------------------------------------------------------------------- HOW: pe Y N •S•Shhaape Y N Y N •• AAddeddded Marks Marks Y N • Gas station survey. ature Y N - •C•Cuurvrvature Y N - le 1000 people surveyed: •A•Annggle ure Y •C•Clolossure Y – Which gas station section Y •• InterIntersection Y inators Y –Age •T•Teermrminators Y •Holes Y – Fuel grade purchased •Holes Y (regular, mid-grade, premium) – Payment method –Gender –Etc. © 2009 Richard Brath except as otherwise noted. Page 14 Gas Survey: Fuel Grade Curve Who sells the most premium (U) gasoline? Do the U’s stand out? I think so! Arco Chevron Mobil Other Shell Unocal 18-25 26-35 36-45 46-55 56-65 66-75 75+ Fuel Grade: Regular Mid-grade Premium © 2009 Richard Brath except as otherwise noted. Page 15 Experiment #1 • Seems like curves can effectively show 3 different categories. Experiment #2: More categories of curves • How about more curves? © 2009 Richard Brath except as otherwise noted. Page 16 More Curves using a morphological approach Gas station survey data set: • See both “fuel grade” and “payment method” using only curves: Credit Branded Cash Debit Card Card Note: there are different types of curves: e.g. conic sections vs. bezier. This bezier stands Regular out from the other half ellipses. Mid-Grade Premium 12 different curve shapes. © 2009 Richard Brath except as otherwise noted. Page 17 Gas Survey: Payment Method 2nd Curve How are they paying for that gasoline – cash or charge? Arco Chevron Mobil Other Shell Unocal 18-25 26-35 36-45 46-55 56-65 66-75 75+ Fuel Grade: – Regular u Mid-grade U Premium Payment: Cash Debit Credit Card Branded Card © 2009 Richard Brath except as otherwise noted. Page 18 Experiment #2 Findings Attribute Binary Category Order/Qty 1. Curves could work for more than •Shape Y N 3 categories • Added Marks Y N •Curvature Y y- 2. Multiple attributes of curves can •Angle be utilized to create different •Closure Y • Intersection Y perceptually distinguishable •Terminators Y curve shapes: •Holes Y -amplitude -skew Attribute Binary Category Order/Qty - bulginess •Shape Y N • Added Marks Y N •Curvature Y y- 3. And, we should repeat this •Angle •Closure Y experiment for angle, hole, • Intersection Y terminators, etc •Terminators Y •Holes Y © 2009 Richard Brath except as otherwise noted. Page 19 Experiment #3: Treemap of S&P500 GOAL: Replace a Treemap using quantitative shapes Attribute Binary Category Order/Qty Shape Y N 4 hours later… Added Marks Y N Curvature Y y- Angle Closure Y Intersection Y Terminators Y Holes Y © 2009 Richard Brath except as otherwise noted. Page 20 Inspiration: the use of angles in the Chappe Telegraph Image src: wikipedia.org www.ucalgary.ca/~bakardji/ElectricComm/Diag%201.gif © 2009 Richard Brath except as otherwise noted. Page 21 Chappe Telegraph as applied to S&P 500 data left arm = percent change; right arm = volume relative to 30 day avg; horizontal = 0. Up Above Average Stock Price Volume Percent change in No Change Average Today’s volume compared to stock price today 30 day average volume Down Below Average © 2009 Richard Brath except as otherwise noted. Page 22 Experiment #3: Chappe Telegraph Findings Attribute Binary Category Order/Qty •Shape Y N Looks like angles work.