Artist's statement
The original dataset separates the world into ten top-level categories, each with the same 5 subcategories. The top-level categories are: As child at Home, As child at Play, As part of a Magical world, As an adventurer in Nature, As a Good/Bad child, As a Student, As a Professional, As member of High Society, As a Soldier , As part of Empire/Religion. The subcategories are: people, goods, actions, emotions, qualities. ‘Objects’ is a sixth but minor category that is also included.
I created four plots in order to show the relationship between the top-level categories and the subcategories as well as the frequency of the top-level categories within each decade. The dataset provided ranked each book’s tag in the top-level category, and subsequently the subcategory. In order to process the data for representation, first the duplicates in the dataset were removed. There were many repetitions in the data making it difficult to process; thus the duplicates were removed. The removal of duplicates made the dataset compact enough for use with online chart tools. Next, to gather more information from the dataset, I wrote a Python script to count in each top-level category the number of instances that each subcategory occurred. This allowed for charts to be created that had not only the type of subcategory, but also the relative value.
Figure 1 is an interactive histogram that was created was using Plotly, an online interactive chart creator. We used Plotly to plot the frequency of each top-level category by the decade. In order to do this, each top-level category was selected as a distinct group, and the number of each category occurred was plotted. Since, each decade was separate in the dataset, I was able to use Plotly to count the top-level categories. I chose to plot by decade instead of year or century, since year was too specific, and century too broad. Within the chart we can interact with the bars and see within each decade the ranking of each category. By using this chart, we can observe how over time most of the top rankings are of ‘As a Student’, however, the second highest ranking changes overtime and we see the rise of ‘As part of a Magical world’ after 1900. Another feature to note, is that this chart can also compare in isolation categories by selecting or deselecting them on the legend, which may be useful when comparing two top-level categories and the trends. Some data was not present; for example, from 1940 to 1960 there was no input from the dataset and the chart is blank here. In addition, this type of chart is most effective when there are hundreds of data points, and some of the entries in the dataset only had 40 or 50.
Figures 2 and 3 are interactive tree-maps which were created and coded with JsFiddle (online Javascript chart editor), and show the hierarchy relationship between the top-level categories and subcategories. The design is adapted from the Google Charts Treemap example code. First, the data is taken from the Python script that counts the frequency of the subcategories. Then, we use JsFiddle to create the plot. The size of the top-level categories is all the same, however the size of the subcategories is in relation to the frequency within each top-level category. To show this we use a treemap because each subtree can be shown individually. To move down the subtree we click on the top-level category. The size of each of the size subcategories can be see when the mouse is over the subcategory. We used Javascript to create this chart because the treemap is best viewed with interaction and Javascript can enable these features. Further, the color intensity increases as the size of the category increases. In addition, two plots were created in order to show the the frequency of each subcategory as a value. However, since some top-level categories have more subcategories we also created a chart (figure 3) that graphs the percentages of each of the subcategories instead of the raw count.
Figure 4 represents the high-level separation between the top-level categories and subcategories. We use an alluvial flow chart (produced with Raw) in order to show the largest (most frequent) subcategories over the entire space of top-level categories. The ranking from top to bottom shows the ranking of the most prominent top-level categories and subcategories. This high-level overview of the dataset allows us to make connections within the categories, and between the levels. For example, using this chart we see how overall besides 'As a Student', 'At Play' and 'At Home' rank highest among the top-level categories.
In conclusion, by using these 4 charts I think the original dataset can be understood more clearly, and observations about not only hierarchy, but also the categories over time can be understood more. In addition, I created four charts because I think that there are multiple features of the dataset that can be illuminated through different data representations. I also think the ability to interact with the charts adds to the ability to understand the data. Through interaction we can observe multiple things from a single chart and we have the ability to isolate categories. The dataset provides a unique break-down of the tagging data, and by using these charts we can understand the separated categories.
Figure 1
Figure 2
(Click 'JavaScript' then click 'Result'). Left-Click to descend, Right-Click to ascend .Frequency of Top-level Categories and Subcategories Treemap.
Figure 3
(Click 'JavaScript' then click 'Result').Frequency of Top-level Categories and Subcategories Treemap as Percentages.
