Author: Kyle Adams

For this assignment, I continued my investigation of journalistic sources involving the firing of Joe Paterno at Penn State University in 2011. In addition to the university newspaper sources I analyzed for our previous assignment, I added several non-collegiate journalistic sources to my corpus (articles from The New York Times, The Chicago Tribune, The Seattle Times, The Los Angeles Times, The Denver Post, and The Dallas Morning News). In terms of metadata, I chose to classify my data into the following categories: institution, institution location, institution geolocation (lat,long), institution connection to Penn State, article author, author gender, article title, article word count, date of publication, article classification upon original publication (ex: ‘Opinions’), and a quantitative sentiment valuation for the article (taken from Jigsaw and altered by adding one to the value given in Jigsaw to put all scores in a numerical set ranging from one to two). While my metadata certainly contained values for many constituent aspects of my articles, I believe each could be visualized in an epistemological way. When it comes to sentiment and word count (my two numerical categories), each of these facets could be visualized in such a way that aids in generating arguments regarding quantity and quality of discourse surrounding Joe Paterno’s firing. In addition, these values could be paired with author gender or geolocation to determine which areas (in general, as there is still subjective interference in article selection) or genders are speaking more or less and more or less pleasantly about the event. Although publication date did not turn out being too helpful for my data visualization, a larger corpus may foster some sort of conclusions tying article publication (denoting temporal proximity to the actual firing) to geographic location, institution classification, or institution connection to Penn State. I also believe it is important to keep article title and author name present in certain visualizations, as these things can both become pertinent to discursive analysis (understand particular author bias if a name is familiar and get a sense of article tone by looking at title).

In both the Table and Gallery views in Palladio (Left: Table View, Above: Gallery View) I found the most important function of the platform to be keeping researchers from being too far divorced from the metadata and corpus in question. One of Johanna Drucker’s most poignant comments from her chapter mentions that “almost all information visualizations are reifications of mis-information” (Drucker 245). In my opinion, these two windows in Palladio prevent this from happening on some level. While they do not necessary alleviate concerns regarding data as capta (constructed and gathered by biased subjects), it does allow for those who engage with a corpus and set of metadata in the platform to see for themselves exactly where visualizations are coming from. In the Table view, this is accomplished by presenting metadata exactly as it was uploaded. The Gallery window (when linked to URL’s) not only does the same thing, but can allow (at least in my case) for researchers to return to the component articles/documents of a corpus with the click of a button. Essentially, each of these views in Palladio helps to increase familiarity with a data set/corpus for the benefit of better understanding visualizations in other views.

In the Graph view (Pictured Twice Above), I chose to visualize connections between author gender (M, F, and N/A in the case of collective publication) and article title with node sizing governed by Jigsaw sentiment score. While node sizes were not remarkably illustrative in an argumentative, epistemological, or hermeneutic sense, this could potentially be attributed to the small range of achievable sentiment scores keeping nodes at similar sizes. In addition, because the metadata being visualized comes from a journalistic corpus, it is understandable that wild swings in sentiment from one article to the next are not necessarily evident in the visualization (most articles must conform to a customary journalistic prosaic style). However, when node sizes are combined with the sentiment conclusions rooted in article title (as this conveys tone), one can use these visualizations to begin to discern a connection between gender and article “feel.” It appears to me that female authors, based on article title in particular, were more willing to tie the Paterno firing back to Sandusky’s sexual abuse explicitly, whereas male authors were more likely to use their platform to honor Paterno or frame the event through a less criminal lens.

I also found the Map view (Above) to be quite interesting in terms article sentiment and publishing institution geolocation. When I began to construct this corpus, I theorized that article sentiment would become more negative as proximity to Penn State decreased. However, according to the map, it seems that some of the highest Jigsaw sentiment scores (denoting more “positive” vocabulary in the source document) were found further from State College on the map (except in the southern United States). This could have had something to do with closer, Big Ten affiliated, schools being more saddened by the symbolic loss of one of their conference’s most recognizable icons. It is also important to remember when looking at this visualization that the documents in question are journalistic, and, for that reason, may not be the most perfect texts to analyze in terms of sentiment (although some articles are opinion pieces).

All in all, I believe my visualizations presented above can be perceived as flawed in some sense, but also intellectually constructive in another. Importantly, Palladio’s provision of the Table and Gallery views is helpful for preventing researchers from being barred from a conception of the “translation” process that occurs in the visualizations in the Graph and Map windows. That being said, I believe my visualizations, at least in terms of sentiment analysis, are hindered by forces of concealment and reduction that come along with an imperfect (and hidden) Jigsaw sentiment computation (in some sense these visualizations can be seen as simple representations or mis-representations of sentiment). However, what I do believe my visualizations were successful in doing is putting on display some of the ambiguity or more nuanced aspects of my corpus. For this reason, I believe some of my visualizations would be considered successful knowledge creators to Drucker, for their “denaturalizing” effect on viewers. These visualizations are certainly novel ways of interrogating journalistic discourse surrounding Joe Paterno’s firing, and thus can help to serve as stepping stones for new understanding.

When it comes to Drucker’s discussion of spatialization creating meaning, I believe the map view can be used as a tool for knowledge creation. Due to the fact that article sentiment is tethered to proximity to the focal event, one can begin to create arguments regarding geographic location governing attitudes toward the Paterno firing in local newspapers. Article titles could also be added to these visualizations (found by hovering over a point) to help with this aforementioned pattern recognition.

Working with Jigsaw and Voyant

a.) After accepting the fact that corpus construction and visualization was, indeed, an iterative process (and that I would have to add nuance to my corpus throughout the semester), I decided to begin collecting journalistic articles from university newspapers regarding the firing of Joe Paterno. Although it appears that each article does not necessarily focus on Paterno’s firing, I decided it would be interesting to incorporate each publication’s first mention of the event (whether that was the overt focus of the article or not). I chose to gather articles from institutions that are connected to Penn State University either by proximity (UPenn, Bucknell, Pitt), athletic conference affiliation (Ohio State, Michigan, etc.), or level of football prestige (Oklahoma, Georgia, etc.). As my research with this corpus continues, I hope to remain cognizant of the subjectivity lurking behind article selection. My role as an archival author is just one factor that allows for the visualizations produced from my corpus to remain perspectival (simply by adding technology into the equation, essentialist understandings of this journalistic prose cannot be reached).

b.) Throughout my time working with the Voyant platform, I have appreciated it not only for its ability to foster more intimate understanding of meaningful entities, but also provide individuals with the opportunity to easily generate comparative research questions at the document level. In the two Voyant visualizations embedded (one being a Bubblelines visualization and the other a StreamGraph) one can begin to analyze occurrences of key entities as they relate to one another within specific documents and in the corpus as a whole. In particular, the raw word frequencies displayed in the StreamGraph visualization allowed me to to understand how different university news outlets framed the Paterno firing (ex: some did not even mention the board responsible for his firing and others). Two other interesting conclusions that can be drawn from the visualizations are that the victims of Sandusky’s abuse (the cover-up of which led to Paterno’s demise) rarely appear in this journalistic prose, but also that external campus news outlets (non-State College) do not shy away from connecting Paterno and Sandusky, leading to an implicit association shrouded in negativity.

c.) One of the most meaningful visualizations I was able to produce using the Jigsaw platform was the WordTree displaying appearances of “Paterno” throughout my corpus. By expanding my capacity for entity contextualization, this tool afforded me the opportunity to begin to discern community level sentiment displayed in nominal data consisting of journalistic prose of publications external to PSU. Meirelles effectively summarizes the value of a WordTree visualization for analysis in her sixth chapter, “Besides preserving the context in which the term occurs, the method also preserves the linear arrangement of the text” (Meirelles 200). While functioning in this way, the WordTree visualization manifests itself as a more intermediate step in distancing oneself from a text. By quickly allowing me to understand the context in which JoePa’s name showed up, the Jigsaw platform helped me to see that university newspapers outside of State College did not shy away from criticizing Joe Paterno for his role in the Sandusky scandal, but also felt the need to show appreciation for his accomplishments as a coach (things that may aid in blinding the Penn State community to his wrongdoing). The image of Jigsaw’s Document View exemplifies the platform’s ability to further remove researchers from their corpus and garner an understanding of entity frequency and classification all while summarizing documents (although the selection of these statements by the program may also be an example of non-objectivity in data provision).

d.) In terms of commonalities, each of these programs has the ability to generate meaningful visualizations that allow for the contextualization of entities (ex: seeing connected words, etc.) and understanding of word frequency (which can be extrapolated to fit perspectival analysis). Obviously, each platform also possesses characteristics which its counterpart lacks. For example, Jigsaw can easily generate visualization of sentiment while Voyant offers remarkably unique and varying visualizations of word frequency (ex: bubble sizes, word sizes, stream width). It has also become rather apparent to me that individuals choosing to work in the Jigsaw platform must have a more intimate understanding of their corpus in order to create visualizations, as more “work” must be done to produce meaningful images. However, once one is familiar with the platform, I am under the impression that Jigsaw allows for more advanced visualization of entity context and document connection. All in all, Voyant provided me a more preferable method for discerning patterns regarding word frequency (allowing me to address sentiment and entity relationships). Jigsaw, on the other hand, allowed me to produce detailed WordTree visualizations that worked as a sort of additional approach in bolstering conclusions regarding outsider perception of Paterno (bigger picture as opposed to entity-level).

e.) Simply put, by engaging with both my corpus and the textual analysis platforms Jigsaw and Voyant, I have learned there are many opportunities for argumentation and subjective involvement in digital humanities research. In this way, I have learned to appreciate the role of technology in expanding the capabilities of nominal data analysis, but also understand that the implementation of digital methods into digital humanities does not generate essentialist interpretations (I now actually see where I could have gone in different interpretive directions). As Clement puts it, “Sometimes the view facilitated by digital tools generates the same data human beings (or humanists) could generate by hand…At other times, these vantage points are remarkably different from that which has been afforded within print culture and provide us with a new perspective on texts that continue to compel and surprise us by being so provocative and complex — so human” (Clement 12). What can be gleaned from this quotation, and my work on this assignment is that digital humanities does, in fact, involve a give and take between subjective and objective construction. By utilizing differential reading strategies and accepting the interaction of close and distant reading, researchers such as myself will be properly equipped to read and contextualize readily accessible and more distant digital findings (and acknowledge subjective intervention in text selection, platform construction, etc.).

Comparison of “The Genealogy of Pop/Rock Music” and “IBM Watson News Explorer”

I chose these visualizations for their collaborative function as exemplar of the fact that networks can manifest themselves in remarkably different ways. When analyzed in tandem, “The Genealogy of Pop/Rock Music” and the “IBM Watson News Explorer” exemplify several distinctions between static and dynamic visualization. As a dynamic visualization, the latter allows individuals to interact with a massive news/media database that simultaneously sorts articles by location, keyword, actors, etc. By giving those engaging with the platform the freedom to “play” with the visualization of sources (ex: alter how they are presented and what is presented) as well as updating information in real time, this IBM interface demonstrates characteristic elements of dynamic visualization that the more static (cannot be altered or “played with” but does include a temporal dimension) musical genealogy does not. Both visualizations, by allowing for the studying of patterns of connections between elements (musicians, news sources, or whatever else) allow data to be viewed in novel ways. However, only the news explorer provides individuals with the opportunity to actively interact with the data in multiple ways (must also keep in mind that the genealogy’s multiple dimensions allow for pluralistic interpretive interaction with data). Although essentially static, the musical genealogy does serve a heuristic function for its ability to depict artistic longevity and influence that may not immediately be apparent to researchers (operates similarly to Priestly’s timeline). The IBM platform utilizes more dynamic qualities to create new knowledge about the interconnectedness of major global figures, organizations, etc. (and even can be used as a way to interrogate the media as an institution). In either case, the visualization does a magnificent job of contributing to novel ways of understanding their fields and many others (understand both relational and attribute data).

Analysis of “Six Degrees of Francis Bacon”

I selected this visualization based on its direct ties community network designs discussed by Meirelles and Lima. This network visualization entitled “Six Degrees of Francis Bacon” utilizes dynamic characteristics such as a mutable structure and variable presentations of biographical and connective information that can be accessed at the discretion of users (both things static visualizations do not necessarily present). More specifically, this visualization allows users to both interpretively interact with the data as well as more actively mutate the visualization (can be radial, force-directed, etc.). By visualizing myriad informational dimensions and including analytical ambiguity, this community network succeeds in allowing individuals to view the data from different perspectives and arrive at unique and valuable humanistic conclusions. “Six Degrees of Francis Bacon,” much like other community networks we have encountered, effectively contributes to new understanding of intellectual, political, and general social influence in England (shows connections, timelines, biographical information). If this particular visualization did not exist, it would seem nearly impossible to readily generate conclusions regarding the impact or magnitude of this particular social network tethered to its central node, Francis Bacon.