CP SC 8810 Data Visualization. Joshua Levine

Transcription

1 CP SC 8810 Data Visualization Joshua Levine

2 Lecture 15 Text and Sets Oct. 14, 2014

3 Agenda Lab 02 Grades! Lab 03 due in 1 week

4 Lab 2 Summary

5 Preferences on x-axis label separation 10 could simply indicate laziness

6 Preferences on Rotated vs. Normal Text for the y-axis label Rotated Normal Both!

7 Choice for the Best Summary View (Note: These OFTEN did not agree with the answer to Task 2) Dots Connected Dots Lines Area Multiple User Selects

8 Task 2 Answers for Best Not Selected means argued for multiple or forgot to answer Dots Connected Dots Lines Area Not Specified

9 Grade Distribution: Mean 8.84 StdDev 0.93 Great Job Overall!

10 How Did They Look?

11 Good Colors Effective Legend

12 Good Colors Interactive Legend

13 Nice Legend Interactive Mouseover Different x-axis separation

14 Interactive Measuring Bar Less Saturated Colors

15 Effective Transparency Interactive Legend Informative Mouseover

16 Different Aspect Ratio Reinforced Visual Encodings

17 Task 1: Minimal Ink Histogram of y- values

18 Combined Line Plots with Area Plots Interactive Legend + Mode Selection Both Normal and Rotated Text for y-axis

19 Last Time: Trees and Graphs

20 Design Choices for Trees and Graphs

21 Node-Link Layouts Strengths: Understandable visual mapping Can show overall structure, clusters, paths Flexible, many variations Limitations: All but the most trivial algorithms are > O(N 2 ) Not good for dense graphs: Hairball problem! Small changes in the graph can cause dramatic changes to the layout See Frishman and Tal. Online Dynamic Graph Drawing. Proc EuroVis 2007 hs s

22 Matrix Layouts Instead of node-link diagrams, use the adjacency matrix to represent A A B C D E B C A B C D E D E

23 Matrix Representations Strengths: Great for dense graphs Visually scalable Can spot clusters Limitations Abstract visualization Hard to follow paths

24 Spotting Patterns in Matrices Henry 2006

25 Adjacency Matrices

26 Attribute-Driven Layout

27

28 Paul Butler

29 Attribute-Driven Layout Large node-link diagrams get messy! Are there additional structures we can exploit? Idea: use data attributes to perform layout e.g., scatterplot based on node values Dynamic queries and/or brushing can be used to enhance perception of connectivity

30 Cerebral Barsky 2008

31 CHI 2006 Proceedings Visualization 1 April 22-27, 2006 Montréal, Québec, Canada Visual Exploration of Multivariate Graphs Martin Wattenberg Visual Communication Lab, IBM Research 1 Rogers St., Cambridge MA mwatten@us.ibm.com Figure 1. A PivotGraph visualization of a large graph rolled up onto two categorical dimensions Author Keywords information visualization, graph drawing

32 Pivot Graph Task abstraction Show relationship between node attributes and connections in a multi-attribute graph Data abstraction Relational dataset Nodes (and edges) have multiple discrete attributes Rollup and selection transformations

33 Visual Encoding Line (1D) or grid (2D) layout Area subdivided by number of values for an attribute Number of nodes based on attribute count, not original graph node count Size of nodes and edges related to number of aggregated original nodes and edges Scalability through abstraction, not layout algorithm

34 Visual Encoding Line for 1D rollup, or grid for 2D case Wattenberg 2006

35 Interaction Changing rollup/selection choices Animated transitions between states

36 Pivot Graph In general, more compact than matrix representation 81

37 Critique: What Do You Think?

38 Tree and Graph Vis Summary Trees: Indentation: simple, effective for small trees Node-link and layered: look good but needs exponential space Enclosure (treemaps): great for size related tasks but suffer in structure related tasks Graphs Node-link: familiar, but problematic for dense graphs Adjacency matrices: abstract, hard to follow paths Attribute-driven: not always possible Takeaway: No Best Solution or Graph visualization is still a great research area!

39 Text

40 Text Data No Numbers (implicitly) Characters (ASCII) Strings

41 Text Data love Words visualization I. Sentences Paragraphs Chapters Lines

42 Text Data love Words visualization I. Sentences Paragraphs Chapters I love visualization. Lines

43 Text Data love Words visualization I. Sentences Paragraphs Chapters I love visualization. Lines

44 Text Data

45 Text Data Documents Books Papers Webpages s Twitter posts! Corpus: collection of documents

46 Text Data Documents Books Papers Webpages s Twitter posts! Corpus: collection of documents

47 Text Visualization For Documents

48 Tag Clouds / Word Clouds

49

50 Text Arc Wattenberg, Viegas 2008

51 DocuBurst Collins, Carpendale, Penn 2008

52 Arc Diagrams Analysis of the Characters from Les Misérables:

53 Rule-Based: Poetry Abdul-Rahman et al. 2008

54 Text Visualization For Document Collections

55

56 Document Cards (small multiples)

57 Showing Temporal Relationships: ThemeRiver (Stream Graph) Havre, Hetzler, Nowell 2000

58 Jigsaw: Many Linked Views Stasko et al. 2008

59 Jigsaw: Many Linked Views Stasko et al. 2008

60 Lec16 Required Reading

61 For datasets with spatial semantics, the usual choice for arrange is to use the given spatial information to guide the layout. In this case, the choices of express, separate, order, and align do not apply because the position channel is not available for directly encoding attributes. The two main spatial data types are geometry, where shape information is directly conveyed by spatial elements that do not necessarily have associated attributes, and spatial fields, where attributes are associated with each cell in the field. (See Figure 8.1.) For scalar fields with one attribute at each field cell, the two main visual encoding idiom families are isocontours and direct volume rendering. For both vector and tensor fields, with multiple attributes at each cell, there are four families of encoding idioms: flow glyphs that show local information, geometric approaches that compute derived geometry from a sparse set of seed points, texture i i i i Chapter The Big Picture Arrange Spatial Data ONLY!!