As I’ve been working through the many complications of finding, downloading, cleaning, uploading, and analyzing my data set, I took a moment to create the above word cloud using Wordle which itself was a little complicated as it requires a Java Plugin that is no longer supported on any of the computers I’ve been using today.  After visiting three browsers and two computers to finally have access to adding the plugin (since I cannot easily download or run updates on my work computers due to ITS protocol), I made the above from a list of Subject Headings provided by the Library of Congress.   I actually used a spreadsheet found here.  When working with data, it seems even the simplest of tasks come with many complications!  Regardless, I am not deterred!

The word cloud represents one aspect of what I am trying to look at with my data – a set of call numbers from the McEngegart Hall Library’s Reference Collection at St. Joseph’s College, Brooklyn.  Call numbers have a direct relationship to their subject headings, so I will be using a data set of call numbers and their subject headings which I will use to determine if there is a fair representation of monographs, based on their call numbers and associated subject headings, in the library’s collection for each of the academic programs offered at the College.

Initially, I wanted to look at the entire print collection, but as I began to pull the data, I thought it might be too big (68,794 lines in Excel) of a set (at least for the time being since I plan to use the larger set if I can get the subset to make sense).  So, I’ve chosen to look at the reference collection (1,926 lines in Excel) as a subset, which may offer some insight, but I believe isn’t the best representation of the collection.  First and foremost because of the nature (and cost) of reference collections, my library has stopped adding paper bound titles and relies more heavily on online databases (whose titles are not represented here), so there will be many gaps.  Also, this past summer, we moved several of the reference titles into the circulating collection to give them a better opportunity to be found while students are browsing the collection.  Nonetheless, not an ideal representation, but a start!

Why would I want to know about this aside from having to pick a data set to work with??  Many academic libraries have to defend their need for the large spaces they take up on campus and as many libraries before mine have had to accept, we are losing a quarter of our space to new classroom labs, so I’ve been asked to condense the collection which required hours of measuring and project planning.  For now, I at least do not have to remove any books from the collection (and there’s a little room for growth).  Taking a look at the collection by analyzing the representation of items through their call numbers and subject headings seems to be an interesting way of interpreting the collection in addition to trying to work out the gaps in the collection and see where there may be too much in any particular area.  As I was in the stacks measuring–and I’ve known this for quite some time–there’s a great need for weeding the collection (library speak for removing titles that are superfluous to the collection and do not support the users’ needs).

I know this project won’t change the way library collections are seen on the macro-scale, but it will help define and defend my library’s collection.  I have been asked by two administrators in the last few years if I actually think physical books belong in a modern library.  This is not a question I love being asked as I believe there is still a great need for physical materials, but I think with this project, I can at least counter these inquiries with some impressive data and hopefully some nice visualizations.  It is worth mentioning here that I wanted to start with an open data set available through NYC’s Open Data portal.  I was inspired for this project by a data set listing Brooklyn Public Library’s Collection uploaded to the portal, but I wanted to work with something more directly related to my day to day work (not that I don’t love public libraries and all that they do for this city!).

Pulling the Data:

I thought this would be the very easiest part, and I was right!  With a simple query in an Access Database, I was able to pull call numbers and other bibliographic data that I could export to Excel.  As already mentioned, I planned to use the entirety of the collection as a set, but have since decided to just start with the reference collection subset data.  I came to this decision after grappling with time constraints on cleaning the data.

Cleaning the Data:

I thought this would be a difficult part of the project and I am finding that to be true.  First, LoC Call numbers do not sort very nicely (read: quickly) in Excel as a result of their alphanumeric configurations.  I did find some work arounds and am still experimenting with some of my options.  Another tool I’d like to look into using is OpenRefine.  Second, I also have to remove several of the duplicates due to multiple volumes of one title (or decide if that’s the right thing to do).  I also need to pull and add data on the subject headings that are assigned to each call number.  I am still working out how best to do this: a) do I rerun a query in Access to include assigned subject headings using the MARC records or b) do I work out a way to assign a subject heading based on the call number of the item (the latter could potential be achieved by writing a Python script to assign subject headings based on a call number range).  I still need to experiment with both of these possibilities.

Data Tools to Use:

I intend to first experiment with relational analysis tools listed on DIRT (DIgital Research Tools): http://dirtdirectory.org/tadirah/relational-analysis.  Also, I still have to work out how to draw a correlation between the volume of subject headings represented in the collection to the academic programs.  By taking a look at the programs offered, I will need to determine how these programs can be looked at in terms of subject headings.  Do I develop a list of each program and then determine which subject heading(s) would be of use to this program?  For example, Child Study is a major at the college and one such LoC heading and its corresponding call number range is: “Education — Theory and practice of education — Child study / LB1101-1139” but there are several other headings too.

Conclusions:

So, I still have a lot of work to be done and many more questions to be asked & answered.  I think two complementary projects for this first project would be to look at usage statistics for the titles in the collection.  Once I know what is in the collection and how it represents the programs, I can compare that to what students are actually using.  What are the students searching for in the catalog and what books are they borrowing?  Also an important outgrowth of this initial analysis of paper materials  would be looking at data from the library’s various ebook collection subscriptions, such as title lists from each database and what titles have disappeared over the years of subscribing.  With these additional points of data, an even greater holistic picture could be drawn.  Since Michelle’s presentation on maps a few weeks ago was so thought provoking, I would also like to include some physical representation of the collection using a catalog addition such as LibraryThing’s or StackMap since GPS coordinates wouldn’t work here.

Scope creep is becoming extremely apparent as I wrap up this post and realize that I have talked about a variety of tools I want to work with and tangential data sets that can be looked at and I am starting to feel like I’m overreaching–especially for a project that I do not feel will have some lofty impact on the library world.   However, this data set will change my day to day work world a little and that’s worth it.

Update as of 11/9/16: – I came across this article in my work inbox: http://crl.acrl.org/content/77/6/765.full.pdf+html  I plan to use it as a model to further my own project.

Aside from the juvenilia, many of On Kawara’s most famous works are documents of movement and time. For his I Got Up series, Kawara sent two postcards every day between May 10, 1968 and September 17, 1979. Each postcard is stamped with the words “I GOT UP AT” with the time that Kawara got up. In his I Met series, also done during this period, Kawara documented all the people he met. There is a page for each day with a list of each person he encountered. The series I Went, done between June 1, 1968 until September 17, 1979, consists of Kawara tracing with a red line on a photocopy of a map the his daily movements. During the 1970s, Kawara sent telegrams to his friends with the words “I am still alive.” On Kawara, in a canonical history of art, is classified amongst the Conceptualists. If so, he was certainly one of the most meticulous and disciplined. The postcards, telegrams, and lists could be seen as early metadata of a life, one part of the projects of walking, waking, and meeting. Yet he was distinct from his contemporaries in the movement, as Jeffrey Weiss points out, for his devotion to painting.

For this data project, I would like to focus on his Date Paintings, known collectively as the Today series. It consists of nearly 3,000 works. For the project, Kawara painted according to a strict set of rules. The painting of a given date must be completed by the end of the day or be destroyed. He has eight possible dimensions and three possible colors: red, gray, or blue. There is variety in the colors since they are hand mixed. The date, also the focal point and representative image, is rendered according to the language of the country Kawara was in (he was an avid traveler).

Image of On Kawara’s Time series via Phaidon

My fantasy is to derive data from all of Kawara’s documented works and establish an interactive database of all his works. The Today series is, in part, a meditation on the day’s events. The craft and care in their painting are astonishing. At the same time, I am approaching this project under the belief that these paintings can facilitate a generative interaction—even when the actual paintings are left out.

I have attached a csv I made using this checklist from an exhibit of Kawara’s date paintings that was on display at David Zwirner gallery from January 6 to February 11, 2012. I used Excel and made a column for the title of the painting, the month, year, dimensions, city, language of the text, the country, the date, and the caption. I briefly considered using Photoshop to identify the colors of each painting but there are too many variables and adding a column containing that information.

On Kawara CSV from Zwirner Checklist

In putting together this csv file, I have already faced several complications and inconsistencies. For some entries in the checklist, the city appears to be in the original language while others are listed in English. I may edit this document to display cities in English without accents. Most cities have identifiable countries, however, based on my Google-ing, Karija (on line 63) may be a city in India or Yemen. I do not know what country Simantogava, the location for 10 FEB. 2006. Then there is the issue of what to do about the captions. After a certain point, Kawara stopped using newspaper headlines as a subtitle and began to simply indicate the day of the week when each painting was rendered. I would like to add to my dataset. I know that he kept meticulous calendars with the dates and colors of his paintings and I am still trying to get access to reproductions of them.

Moving forward with this dataset, the cleaning of which is proving to be far more complicated and time-consuming that I anticipated, I would like to include latitudes and longitudes for each city. I will definitely be adding a column for a color category (either red, gray, or blue).

Just to test out my dataset, I have mapped the paintings using CartoDB with years indicated by color .

I have also used Google to generate histograms of the paintings by language and dimension included below.

Moving forward, I would like to use R to create subsets of my dataset using color and/or particular years. I can also use it to generate maps with weighted data points, which I might size according to the painting dimensions. Using R would also allow me to examine multiple variables in single visualizations so I would like to represent each date with a weighted datapoint in the color the painting background was rendered in. While my dataset is limited, I am excited to see what patterns in the work can be brought forth through visualizations.

As I mentioned, part of Kawara’s process involved the construction of a box with a clipping from the day’s newspaper. I don’t know how yet but I would like to somehow incorporate major international events of particular days into this project. It would be great to have headlines from major newspapers for each particular day of my dataset, particularly on the days from the period after Kawara stopped using the headlines as a subtitle. If anybody has suggestions on how I can make this happen, I would love to hear it. Looking forward to your feedback! Thanks!

Who knew that Finnegans Wake would one day be reduced to cells in a spreadsheet?

For a long time, I wanted to do experiments with Finnegans Wake and data visualizations. This recent assignment gave me this chance so I quickly started to try to think about it. The first thing I had to do was figure out my dataset. Obviously Finnegans Wake is fiction and relies heavily on an idioglossia of Joyce’s design so it might be tough to pinpoint distinct data points for the book. This meant that I had to take a step back and look at the book from a very removed perspective to start -what better a dataset for this book than its own lexicon and frequencies? Studying the Wake in the past led me to remember a couple of different online tools like Fweet, which is a search engine for the book, and Finwake which is an online annotated version. However, the most useful gathered data would have to be from Eric Rosenbloom’s Concordance of Finnegans Wake which he compiled apparently in the late 1990s. Throughout this project, there will definitely be certain data constraints considering the fact that no major datasets have really been constructed for the Wake.

To continue reading, please visit my blog HERE.

First off, let me say that this was not a tutorial. I followed along on some of work Micki Kaufman did in Tableau and Gephi, but this session was not meant to be a tutorial. Instead, we learned from Micki’s tremendous experience the practices that make effective, scholarly presentations of data. I will highlight some of the things I learned and insights I had.

Micki spent some time on color acuity, which is obviously very important when using color to represent data. Apparently red and green make for an awful pair, and Excel often defaults to a red and blue that some may struggle with. The best pair to use is blue and yellow, as most people can distinguish it. When possible, using different textures and lines can also be very helpful.

Micki emphasized that we do not have to be experts in computer science to use advanced data-manipulation techniques and feel confident in our work. For one thing, she railed on “brogramming,” the idea that one needs to use command line to achieve anything. Instead, it is probably better to use GUIs that have been created for more general use. I do feel that with most software, limitations will eventually surface as you get really specific in what you need to do. Relatedly, one only needs to know so much of the backend and underlying principles to confidently make their argument. This reminded me of some of our early readings in Debates in the Digital Humanities, and I think some scholars may disagree with the adequacy of our tools.  Micki instead highlighted visual subjectivity and the need to be aware of how a visualization exaggerates or de-emphasizes phenomena.

For instance, a word cloud represents how often words are used in a dataset, but their proximity is entirely random, and this can have unintended consequences. Micki’s work obviously goes far beyond just measuring the magnitude of words. She demonstrated some of the techniques she has applied and experimented with, such as topic modelling, collocations, and concordance.

Some of Micki’s other points related strongly to our class discussions–her use of monochrome in one visualization because of severity of death her topic touches, The History Manifesto, and the hermeneutic, cyclical process of visualization. Perhaps the most poignant thing Micki said was that a groundbreaking analysis or visualization must at once be dramatically different and similar to what the observer expects. Replicating past arguments is hardly compelling, nor is finding data that supports a completely bewildering idea. Instead, there must a contextualized challenge to a belief other scholars’ hold in the field. It may seem reductionist to suggest how to make an impact with research, but Micki’s example did well to illustrate this point: if a young, 11th century scholar invented a new telescope, the learned astronomer would expect it to replicate his ideas about the universe, but show him something new.

We did get some instruction on using Gephi, which was helpful to me as I had previously been completely overwhelmed. I was eager to learn a practical method, a new technical tool. Instead I got substantial insight into how to best employ the skills I hope to learn. Micki is a very capable and clear presenter, so her ideas will stay with me until I have use for them. Meanwhile, I have signed up for Baruch’s mapping class and am looking forward to getting in-depth with a tool.