Let’s draw some more graphs.
Here’s the matrix of form and genre rendered graphically:
<tt>## Load the tab separated values for the table of artworks
artwork<-read.delim("./visual_art/artwork.tsv")
# Get rows with both genre and form
## This loses most of the data :-/
art<-artwork[artwork$art_genre != "" & artwork$art_form != "",
c("art_genre", "art_form")]
## Drop unused factors
art$art_genre<-as.factor(as.character(art$art_genre))
art$art_form<-as.factor(as.character(art$art_form))
## Get table
art.table<-table(art) ##as.table(ftable(art))
## Strip rows and columns where max < tolerance
tolerance<-3
art.table.cropped<-art.table[rowSums(art.table) >= tolerance,
colSums(art.table) >=tolerance]
## Print levelplot
## Levelplot is in the "lattice" library
library("lattice")
## Rotate x labels, and set colour scale to white/blue to improve readablity
levelplot(art.table.cropped, xlab="Genre", ylab="Form",
scales=list(x=list(rot=90)),
col.regions=colorRampPalette(c("white", "blue")))</tt>
The highest frequencies leap out of the graph. We should do a version without painting to look for subtleties in the rest of the data.
And here’s some of the basic frequencies from the data:
<tt><br></br>## Load the tab separated values for the table of artworks
artwork<-read.delim("./visual_art/artwork.tsv")
## Function to plot a summary of the most frequent values
topValuePlot<-function(values, numValues){
## Get a count of the number of times each value name appears in the list
values.summary<-summary(values)
## Draw a graph, allowing enough room for the rotated labels
par(mar=c(10,4,1,1))
barplot(values.summary[1:numValues], las=2)
}
## Artists
topValuePlot(artwork$artist[artwork$artist != ""], 20)
## Subject
topValuePlot(artwork$art_subject[artwork$art_subject != ""], 20)
</tt>
The dataset is clearly dominated by Western art.