The second part of this post deals with the program created by Michael Correll (email@example.com, http://pages.cs.wisc.edu/~mcorrell/) named “Docuscope Viewer”. With it, this program gives a way to analyze Docuscope’s output without using JMP or any type of multivariate-affected methods of visualization. A quick demo of the program is shown below using the same data set that I fed into JMP to create the visualization in the first part of this post. (Seen here)
This is the initial view of the Docuscope Viewer. This is using the Corpus Viewer function and after adding the output provided by Docuscope in an Excel spreadsheet. This portrays the first level of analysis, the 17 Clusters, according to the same color scheme that Docuscope has in its viewer pane. If the mouse hovers over a particular box the value for that attribute for that play (item being observed) is shown. In addition, the Docuscope Viewer also creates an Averages row at the bottom of the diagram that incorporates all of the other rows present (unless subdivided into groups on the Excel Spreadsheet)
The next view allows for a similar view however it is now subdividing the Clusters into their respective Dimensions (51) through line breaks and slight color differentiation.
As you may have guessed, this next picture is a further subdivision into LATs or Language Attribute Types (evident in the Level of Detail choice on the far right of the screen, with 0 being Clusters, 1 being Dimensions, and 2 being Lats). This kind of revisualization allows for new interpretations of Docuscope’s findings, without necessarily imposing multivariate filters or restrictions.
Another interesting tool this application has is the reordering of the attribute/cluster you wish to look at. (For this diagram, Emotion has been moved to the farthest left in order to analyze the spread more easily than if it were in the middle of the diagram and justified). This can yield very interesting results immediately. For instance Imogen has an Emotion rating of 4.82 which is higher than Prospero, Leontes, and the Averages which are below Imogen on the diagram. This can immediately call into question whether or not Imogen’s Emotion score is higher because she is a female character, a stock character, etc, etc.
And while this program doesn’t lack much in functionality, another great addition to its user interface is its use of deletion within the data set present.
Shown above is the same data set, unaffected within the hard drive of the computer, yet modified within the temporal space of the program. This allows the user to select which elements he or she wishes to look at in more detail and it can function at the same time as any of the previous functions shown (so analyzing Emotion instead of Description at the Dimension level is easily possible). So for the sake of this diagram, I chose to look at Winter’s Tale, Cymbeline, and Tempest in their “with” and “without” forms including their isolated characters as well. And to conclude with amazement, it even calculates the new average of the items shown above.
Michael Correll’s (firstname.lastname@example.org, http://pages.cs.wisc.edu/~mcorrell/) program is indeed a very functional and very beneficial program, and I am grateful to have access to it. Be sure to look for more visualizations from the Docuscope Viewer being incorporated alongside the JMP diagrams.