• Key tool for visual analytics
• Much more efficient than static graphics

Examples:

• Navigation (panning, rotation, zooming)
• Selection (highlighting)
• Connecting (linked views)
• Filtering (sample)
• Reconfiguring (change aesthetics)
• …

• Current visualization systems contain certain interaction
  • Limited Features
• Why do we need a theory?
  • To understand what can be interacted and how
  • To see limitations of existing visualization software
  • To be able to propose new relevant interaction tools missing

Navigation operator:

• Camera location
• Viewing direction
• Level of details (e.g. hierarchical representations)

Australian crabs

• User isolates a subset of objects
  • Highlighting
  • Masking
  • Focusing
• How to implement?
  • Click
  • Click+hold
  • Bounding box, lasso

Australian crabs

• Related observations are linked in corresponding views
• Selection operator+Connection operator = Brushing
  • Persistent and transient

• Challenges
  • How to link data with various mappings
    • Ex: Histogram and scatter plot
    • Ex2: Contour plot and bar chart
    • Ex3: link in which direction? (Hierarchically connected)
  • How to define the corresponding link (key)?
  • Allow for disconecting views?

• Australian crabs:
  • Which sex do the upper-cluster-crabs have?

• Reducing data acc. to specifications

• Transforming data
  • Sorting rows, reorder columns, MDS
  • Change aesthetics mapping

• Changing the visualization type
• Changing aesthetics
  • Another color map
  • Change shapes
  • …

• Distoring objects locally or globally

• What can the operators be applied to?
• What actions does it imply?

Screen space:

• Navigation: pixel-wise actions
• Selection: sets of pixels (boxes, polygons,…)
• Abstraction: distortion of image (fisheye)
• Filtering: removing some pixels

Data value space

• Operate observations instead of pixels
• Navigating: translate the axis range
• Zooming: increase/decrease axis range
• Filtering: sample the data, sample dimensions
• Reconfiguring: sorting observations, dimensions, nonlinear transformations

Data structure space

• Different data structures exist: matrix, list, graph,…

• Navigation operator: how navigate the view in the long tree?

• Selection operator: node in the tree is selected -> subbranches must be selected

• Filtering operator: Social network: click on node - nodes that are X links away disappear

• Abstraction/Elaboration: histogram with zooming - recompute bars?

• Working with aesthetics

• Navigation: change range of aesthetics to certain interval (show certain range of colors)

• Encoding operator: change shapes of symbols, non-linear color mapping

• Selection: highlight certain ranges of aesthetics (highlight stars)

• Dashboards, scatter plot matrices
  • How to navigate user in these?
  • Which components of dashboard can user hide, move, rearrange?
  • Distortion of certain elements

• Measurements of fleas and their types
• Which operators are available here?
• Which clusters do you see?

Without Shiny: key ingridients

• Create cross-talk objects from your dataframes by d<-SharedData$new(data, key, group)
  • key parameter should point to same observations in dataframes
    • Not specified: row id used
  • Group - a unique name related to the same data

Without Shiny: key ingridients

1. highlight-function:
   • applied to Plotly object
   • parameters:
     • on: ‘plotly_click’, ‘plotly_selected’,…
     • persistent: TRUE/FALSE
     • dynamic:T/F - enables color selector
     • selectize: T/F text field for selection

d <- SharedData$new(crabs)
scatterCrab <- plot_ly(d, x = ~CL, y = ~RW) %>%
  add_markers(color = I("black"))

barCrab <-plot_ly(d, x=~sex)%>%add_histogram()%>%layout(barmode="overlay")

subplot(scatterCrab,barCrab)%>%
  highlight(on="plotly_selected", dynamic=T,
            persistent = T, opacityDim = I(1))%>%
  hide_legend()

plotly_data(barCrab)

## # A tibble: 200 × 9
##    species sex   index    FL    RW    CL    CW    BD .crossTalkKey
##    <chr>   <chr> <int> <dbl> <dbl> <dbl> <dbl> <dbl> <chr>        
##  1 Blue    Male      1   8.1   6.7  16.1  19     7   1            
##  2 Blue    Male      2   8.8   7.7  18.1  20.8   7.4 2            
##  3 Blue    Male      3   9.2   7.8  19    22.4   7.7 3            
##  4 Blue    Male      4   9.6   7.9  20.1  23.1   8.2 4            
##  5 Blue    Male      5   9.8   8    20.3  23     8.2 5            
##  6 Blue    Male      6  10.8   9    23    26.5   9.8 6            
##  7 Blue    Male      7  11.1   9.9  23.8  27.1   9.8 7            
##  8 Blue    Male      8  11.6   9.1  24.5  28.4  10.4 8            
##  9 Blue    Male      9  11.8   9.6  24.2  27.8   9.7 9            
## 10 Blue    Male     10  11.8  10.5  25.2  29.3  10.3 10           
## # ℹ 190 more rows

• Link can be one-to-many or even dynamic (scatter/barchart)

m <- SharedData$new(mpg)
p1 <- ggplot(m, aes(displ, fill = class)) + geom_density()
p2 <- ggplot(m, aes(displ, hwy, fill = class)) + geom_point()
subplot(p1, p2) %>% highlight("plotly_click") %>% hide_legend()

Applications:

• Articles, books
• Emails, blogs, websites
• Program Logs
• Collections (corpus) of books, blogs,..

Analysis:

• Understanding structure/context of text
• Group similar documents
• Development of contents/topics over time

• A text document is a sequence of characters->processing needed
• R package: use tidytext
• Raw text

• Tidy text format: one token per row

-Data in tidy format can be analysed by dplyr, tidyr, ggplot2, plotly

-n-grams: tokens of consecutive n words:

• Visualizes word frequencies
• Words are placed in 2D
• Layout decided algorithmically
• Size of words=Word frequency
• Another aesthetics: color

Issues:

• Stopwords need to be removed: a, the, is, actually…
• Words sharing the same stem aggregated
• Synonyms
• “Satisfied”/“not satisfied” example
• Incorrect spelling?
• Hyphens and apostrophes

• If stop words not removed, they are most common in each document
• Solution
  1. Count how frequent the word is among all documents
  2. Weigh the word frequency by the inverse of this number
• TF-IDF model
  • \(Tf(w)\)=frequency of word, \(df(w)\)=frequency of word in all documents, \(N\)=#documents

\[ TFIDF(w)=Tf(w) \cdot log(N/df(w)) \]

• Aim: understand the emotional context of the text
• Each token is given a label/score (positive/negative, [-3,3] )

## # A tibble: 2,477 × 2
##    word         value
##    <chr>        <dbl>
##  1 breathtaking     5
##  2 hurrah           5
##  3 outstanding      5
##  4 superb           5
##  5 thrilled         5
##  6 amazing          4
##  7 awesome          4
##  8 brilliant        4
##  9 ecstatic         4
## 10 euphoric         4
## # ℹ 2,467 more rows

• A possible strategy:
  1. Count total sentiment per text chunks
  2. Analyse Total Sentiment versus Chunk index

• Root= word of interest
• Branches = contexts the word used
  • Type a combination of words and observe the tree, check its branches

https://www.jasondavies.com/wordtree/

• Graph that analyses tri-grams (n=3)

  • ‘A and B’, ‘A of B’, ‘A is B’,…

• Understanding of context without reading

• Example “Romeo and Julia” (is, are)

• Size of the word = word freq
• Thickness of the connection =co-occurance freq
• Color: dark colors -> word often to the left

Analysis:

• Analyse most frequent
• Analyse connections and paths
• Analyse strength of paths and intensity of colors

• Analyses thematic changes in document collections over time

https://hrbrmstr.github.io/streamgraph/

• Example: #movies by genre,different years
• Perception problem: angles in areas

• Other tools for document collection:
  • Place similar documents close in 2D (MDS, clustering, SOM)
• Graph/network visualization can be used for text
• A plenty of other visualizations for text:

http://textvis.lnu.se/

• Chapters 10,11

• Plotly book, ch. 4

• Text Mining with R book, ch 2.1, 2.2, 2.5, 3.1, 3.3, 4.1.1, 4.1.4

• Van Ham, F., Wattenberg, M., & Vi?gas, F. B. (2009). Mapping text with phrase nets. IEEE transactions on visualization and computer graphics, 15(6).

• Wattenberg, M., & Vi?gas, F. B. (2008). The word tree, an interactive visual concordance. IEEE Transactions on Visualization & Computer Graphics, (6), 1221-1228.