Nested model: Four levels of visualization design Domain characterization Design Process Information Visualization • domain situation • details of an application domain Characterize Domain Situation – who are the target users? • group of users, target domain, their questions, & their data Task Abstraction • abstraction –varies wildly by domain domain domain – translate from specifics of domain to vocabulary of visualization abstraction –must be specific enough to get traction Map Domain-Language • what is shown? data abstraction Map Domain-Language Task • domain questions/problems Data Description to to Abstract Task • why is the user looking at it? task abstraction idiom Data Abstraction Tamara Munzner –break down into simpler abstract tasks – often must transform data, guided by task algorithm Department of Computer Science • idiom University of British Columbia – how is it shown? • visual encoding idiom: how to draw Identify/Create Suitable Idiom/Technique Lect 3, 14 Jan 2020 • interaction idiom: how to manipulate [A Nested Model of Visualization Design and Validation. Munzner. IEEE TVCG 15(6):921-928, 2009 • algorithm (Proc. InfoVis 2009). ] https://www.cs.ubc.ca/~tmm/courses/436V-20 [A Multi-Level Typology of Abstract Visualization Tasks Identify/Create Suitable Algorithm – efficient computation Brehmer and Munzner. IEEE TVCG 19(12):2376-2385, 2013 (Proc. InfoVis 2013). ] 2 3 4 Example: Find good movies Abstraction: Data & task Example: Find good movies Example: Find good movies • identify good movies in genres I like • map what and why into generalized terms • identify good movies in genres I like • one possible choice for data and tasks, in domain language • domain: –identify tasks that users wish to perform, or already do • domain: –data: combine audience ratings and critic ratings –find data types that will support those tasks –task: find high-scoring movies for specific genre –general population, movie enthusiasts –general population, movie enthusiasts domain • possibly transform /derive if need be • abstractions? abstraction • task: what is a good movie for me? –attribute: audience & critic ratings –highly rated by critics? • ordinal –highly rated by audiences? – levels: 3 or 5 or 10... –successful at the box office? –attribute: genre one possible idiom –similar to movies I liked? • categorical –stacked bar chart for ratings –matches specific genres? – levels: < 20 • data: (is it available?) –items: movies • items: millions –yes! data sources IMDB, Rotten Tomatoes... –task: find high values? 5 6 7 8 Example: Horrified Task abstraction: Actions and targets Actions: Analyze Analyze • same task: high-score movies • very high-level pattern • consume • {action, target} pairs Consume –discover vs present • slightly different data – discover distribution Discover Present Enjoy – compare trends • classic split –14K rated horror movies from IMDB • actions –l ocate outliers • aka explore vs explain • very different visual encoding idiom –analyze – browse topology –enjoy –circle per item (movie) • high-level choices Why: Task Abstraction • newcomer Produce –search –circle area = popularity • aka casual, social Annotate Record Derive • find a known/unknown item –stroke width/opacity = avg rating tag –query –year made = vertical position • produce • find out about characteristics of item • interaction idiom –annotate, record –lines connect movies w/ same director, –derive on mouseover • crucial design choice http://alhadaqa.com/2019/10/horrified/ 9 10 11 12 Derive Analysis example: Derive one attribute Means and ends Actions: Search • Strahler number What? • don’t just draw what you’re given! • what does user know? Search – centrality metric for trees/networks –decide what the right thing to show is Why? – target, location Target known Target unknown – derived quantitative attribute –create it with a series of transformations from the original dataset • lookup How? – draw top 5K of 500K for good skeleton Location Lookup Browse –draw that – ex: word in dictionary known [Using Strahler numbers for real time visual exploration of huge graphs. Auber. Proc. Intl. Conf. Computer Vision and Graphics, pp. 56–69, 2002.] What? • one of the four major strategies for handling complexity • alphabetical order Location Locate Explore unknown • locate Task 1 Task 2 Why? .74 .74 .58 .58 exports – ex: keys in your house .64 .64 How? .84 .84 .54 .54 .74 .84 .74 .84 – ex: node in network imports .84 .84 .24 .24 .64 .64 .94 .94 trade • browse What? In Out In In Out balance + Tree Quantitative Tree Quantitative Filtered Tree – ex: books in bookstore attribute on nodes attribute on nodes Removed Why? unimportant parts • explore trade balance = exports − imports What? Why? Why? How? What? How? https://bl.ocks.org/heybignick/3faf257bbbbc7743bb72310d03b86ee8 In Tree Derive In Tree Summarize Reduce – ex: cool neighborhood in Derived Data Out Quantitative In Quantitative attribute on nodes Topology Filter Original Data new city 13 attribute on nodes 14 15 16 Out Filtered Tree
Example: Horrified vs stacked bars Actions: Search, query Example: Economics Task abstraction: Targets Search • horrified: browse/explore • what does user know? • task: compare and derive All Data Network Data Target known Target unknown • stacked bars: locate/lookup –target, location • data: derive change Location Trends Outliers Features Topology Lookup Browse known • how much of the data • which is better? Location matters? Locate Explore unknown –depends on goals / task Paths –one, some, all Attributes • enjoy, social context, lots of time • find 2nd-best rated movie of all time One Many Query http://alhadaqa.com/2019/10/horrified/ • independent choices – Jeopardy call, < 10 seconds to respond! Spatial Data Distribution Dependency Correlation Similarity Identify Compare Summarize for each of these three Shape levels Extremes –analyze, search, query –mix and match 17 18 The Economist 19 20 Abstraction Examples: Job market Exercise: Rating Charts for Tasks Exercise: Rating Charts for Tasks • these {action, target} pairs are good starting point for vocabulary • trends • task A: sort attributes • task B: compare pair of attributes (Direct vs Distributor) –but sometimes you'll need more precision! –how did job market develop since recession overall? • rule of thumb • outliers –systematically remove all domain jargon –real estate related jobs • interplay: task and data abstraction –need to use data abstraction within task abstraction • to specify your targets! • but task abstraction can lead you to transform the data –iterate back and forth • first pass data, first pass task, second pass data, ... https://www.nytimes.com/interactive/2014/06/05/upshot/how-the-recession-reshaped-the-economy-in-255-charts.html 21 22 https://www.perceptualedge.com/articles/dmreview/radar_graphs.pdf 23 https://www.perceptualedge.com/articles/dmreview/radar_graphs.pdf 24 Exercise: Rating Charts for Tasks Exercise: Rating Charts for Tasks Exercise: Rating Charts for Tasks Exercise: Rating Charts for Tasks • task C: compare pair of attributes (Distributor vs OEM) • task D: present trends across all attributes • task E: spot outlier attributes • task F: enjoy / engage https://www.perceptualedge.com/articles/dmreview/radar_graphs.pdf 25 https://www.perceptualedge.com/articles/dmreview/radar_graphs.pdf 26 https://www.perceptualedge.com/articles/dmreview/radar_graphs.pdf 27 https://www.perceptualedge.com/articles/dmreview/radar_graphs.pdf 28 Exercise: Task abstraction in genomics I Exercise: Task abstraction in genomics I Exercise: Task abstraction in genomics, I Example: Genomics II Derive • ... only some samples show the desired effect • ... which genes are likely to cause the difference, and • goal: control data quality for gene splicing data You have been approached by a geneticists to help with a visualization what role they play in that pathway . • tasks –derive two groups of samples problem. She has gene expression data (data that measures the activity of Search – judge magnitude of sample the genes) for 30 cancer tissue samples . She is applying an experimental Target known – compare samples, identify within-group variance & outliers • ... the difference between the samples is caused by differential expression –locate the outlier in the network drug to see whether the cancer tissue dies as she hopes, but she finds – compare groups, identify between-group variance Location (different activity) of genes in a particular pathway. She would like to Lookup that only some samples show the desired effect . She believes that the known understand which genes are likely to cause the difference difference between the samples is caused by differential expression Location Locate –identify those genes unknown ( different activity) of genes in a particular pathway , i.e., an interaction –compare gene expression of pathway genes between two groups network of genes. She would like to understand which genes are likely to –identify the outliers Query Outliers –explore the topology Network Data cause the difference , and what role they play in that pathway. Identify Compare Topology proposed idiom –Vials [Strobelt et al 2016] Paths 29 30 31 32
Recommend
More recommend
