Tag Archives: dataset

Part Two, Mapping the Icelandic Outlaw Sagas Narratively

Dear digitalists,

In my last post, I shared a rather lengthy write-up of a geospatial data project I’ve been working on–I hope that some of it is helpful!

Aiming for brevity in this post (and apologies for hogging the blog), I’d like to see if anyone has feedback for part two of the mapping project I’m working on currently. To summarize the project in brief, borrowing directly from my last post: “Driven by my research interests in the spatiality of imaginative reading environments and their potential lived analogues, I set out to create a map of the Icelandic outlaw sagas that could account for their geospatial and narrative dimensions.”

While you can check out those aforementioned geospatial dimensions here, the current visualization I’ve created for those narrative dimensions seems to be lacking. Here it is, and let me describe what I have so far:

Click through for interactive Sigma map

I used metadata from my original XML document, focusing on categories for types of literary or semantic usage of place name in the sagas. I broadly coded each mention of place name in the three outlaw sagas for what “work” it seemed to be doing in the text, featuring the following categories: declarative (Grettir went to Bjarg), possessive (which included geographic features that were not necessarily a place name, but acting as one through the possessive mode, such as Grettir’s farm), affiliation (Grettir from Bjarg) and whether the place name appeared in prose, poetry, or an embedded speech. Using open-source software Gephi, this metadata was transformed into nodes and edges, then arranged in a force algorithm according to a place name weight that accounted for frequency of mentions across the sagas. I used the JavaScript library Sigma to embed the Gephi map into the browser.

While I feel that this network offers a greater degree of granularity on uses of place name, right now I feel also that it has two major weaknesses: 1) it does not interact with the geographic map, and 2) I am not sure how well it captures place name’s use within the narrative itself.

My question to you, fellow digitalists: what are ways that I could really demonstrate how place names function within a narrative? Should I account for narrative’s temporal aspect–the fact that time passes as the narrative unfolds, giving a particular shape to the experience of reading that place names might inform geographically? How could I get an overlay, of sorts, on the geospatial map itself? Should I consider topic modelling, text mining? Are there potential positive aspects of this Gephi work that might be worth exploring further?

Submitting to you, dear readers, with enormous debts of gratitude in advance for your help! And even if you don’t consider yourself a literary expert–please chime in. We all read, and that experience of how potentially geographic elements affect us as readers and create meaning through storytelling is my most essential question.

Mapping the Icelandic Outlaw Sagas

Greetings, fellow digitalists,

*warning: long read*

I am so impressed with everyone’s projects! I feel like I blinked on this blog, and all of a sudden everything started happening! I’ve tried to go back and comment on all your work so far–let me know if I’ve missed anything. Once again, truly grateful for your inspiring work.

Now that it’s my turn: I’d like to share a project that I’ve been working on for the past year or so. I’ll break it down into two blog posts–one where I discuss the first part, and the other that requests your assistance for the part I’m still working on.

A year ago, I received funding from Columbia Libraries Digital Centers Internship Program to work in their Digital Social Sciences Center on a digital project of my own choosing. I’ve always gravitated towards the medieval North Atlantic, particularly with anything dark, brooding, and scattered with thorns and eths (these fabulous letters didn’t make it from Old English and Old Norse into our modern alphabet: Þ, þ are thorn, and Ð, ð are eth). Driven by my research interests in the spatiality of imaginative reading environments and their potential lived analogues, I set out to create a map of the Icelandic outlaw sagas that could account for their geospatial and narrative dimensions.

Since you all have been so wonderfully transparent in your documentation of your process, to do so discursively for a moment: the neat little sentence that ends the above paragraph has been almost a year in the making! The process of creating this digital project was messy, and it was a constant quest to revise, clarify, research, and streamline. You can read more about this process here and here and here, to see the gear shifts, epic flubs, and general messiness this project entails.

But, to keep with this theme of documentation as a means of controlling data’s chaotic properties, I’ve decided to thematically break down this blog post into elements of the project’s documentation. Since we’ve already had some excellent posts on Gephi and data visualization, I’ll only briefly cover that part of my project towards the end–look for more details on that part two in another blog post, like I mention above.

As a final, brief preface: some of these sections have been borrowed from my actual codebook that I submitted in completion of this project this past summer, and some parts are from an article draft I’m writing on this topic–but the bulk of what you’ll read below are my class-specific thoughts on data work and my process. You’ll see the section in header font, and the explanation below. Ready?

Introduction to the Dataset

The intention of this project was to collect data on place name in literature in order to visualize and analyze it from a geographic as well as literary perspective. I digitized and encoded three of the Icelandic Sagas, or Íslendingasögur, related to outlaws from the thirteenth and fourteenth centuries, titled Grettis saga (Grettir’s Saga), Gísla saga Súrssonar (Gisli’s Saga), and Hardar Saga og Hölmverja (The Saga of Hord and the People of Holm). I then collected geospatial data through internet sources (rather than fieldwork, although this would be a valuable future component) at the Data Service of the Digital Social Sciences Center of Columbia Libraries, during the timeframe of September 17th, 2013, to June 14th, 2014. Additionally, as part of my documentation of this data set, I had to record all of the hardware, software, and Javascript libraries I used–this, along with the mention of the date, will allow my research to be reproduced and verified.

Data Sources

Part of the reason I wanted to work with medieval texts is their open-source status; stories from the Íslendingasögur are not under copyright in their original Old Norse or in most 18th and 19th century translations and many are available online. However, since this project’s time span was only a year, I didn’t want to spend time laboriously translating Old Norse when the place names I was extracting from the sagas would be almost identical in translation. With this in mind, I used the most recent and definitive English translations of the sagas to encode place name mentions, and cross-referenced place names with the original Old Norse when searching for their geospatial data (The Complete Sagas of Icelanders, including 49 Tales. ed. Viðar Hreinsson. Reykjavík: Leifur Eiríksson Pub., 1997).

Universe

When I encountered this section of my documentation (not as a data scientist, but as a student of literature), it took me a while to consider what it meant. I’ll be using the concept of “data’s universe,” or the scope of the data sample, as the fulcrum for many of the theoretical questions that have accompanied this project, so prepare yourself to dive into some discipline-specific prose!

On the one hand, the universe of the data is the literary world of the Icelandic Sagas, a body of literature from the 13th and 14th centuries in medieval Iceland. Over the centuries, they have been transmuted from manuscript form, to transcription, to translation in print, and finally to digital documents—the latter of which has been used in this project as sole textual reference. Given the manifold nature of their material textual presence—and indeed, the manuscript variations and variety of textual editions of each saga—we cannot pinpoint the literary universe to a particular stage of materiality, since to privilege one form would exclude another. Seemingly, my data universe would be the imaginative and conceptual world of the sagas as seen in their status as literary works.

A manuscript image from Njáls saga in the Möðruvallabók (AM 132 folio 13r) circa 1350, via Wikipedia

However, this does not account for the geospatial element of this project, or the potential real connections between lived experience in the geographic spaces that the sagas depict. Shouldn’t the data universe accommodate Iceland’s geography, too? The act of treating literary spaces geographically, however, is a little fraught: in this project, I had to negotiate specifically the idea that mapping the sagas is at all possible, from a literary perspective. In the latter half of the twentieth century, scholars considered the sagas as primarily literary texts, full of suspicious monsters and other non-veracities, and from this perspective could not possibly be historical. Thus, the idea of mapping the sagas would have been irrelevant according to this logic, since seemingly factual inconsistencies undermined the historical, and thus geographic, possibilities of the sagas at every interpretive turn.

However, interdisciplinary efforts have been increasingly challenging this dismissive assumption. Everything from studies on pollen that confirm the environmental difficulties described in the sagas, to computational studies that suggest the social patterns represented in the Icelandic sagas are in fact remarkably similar quantitatively to genuine relationships suggest that the sagas are concerned with the environment and geography that surrounded their literary production.

But even if we can create a map of these sagas, how can we counter the critiques of mapping that it promotes an artificial “flattening” of place, removing the complexity of ideas by stripping them down to geospatial points? Our course text, Hypercities, speaks to this challenge by proposing the creation of “deep maps” that account for temporal, cultural, and emotional dimensions that inform the production of space. I wanted to preserve the idea of the “deep map” in my geospatial project on the Icelandic Sagas, so in a classic etymological DH move (shout out to our founding father Busa, as ever), I attempted to find out more about where the idea of “deep mapping” might have predated Hypercities, which only came out this year yet represents a far earlier concept.

I traced the term “deep mapping” back to William Least Heat-Moon, who coined the phrase in the title of his book, PrairyErth (A Deep Map) to indicate the “detailed describing of place that can only occur in narrative” (Mendelson, Donna. 1999. “‘Transparent Overlay Maps’: Layers of Place Knowledge in Human Geography and Ecocriticism.” Interdisciplinary Literary Studies 1:1. p. 81). According to this definition, “deep maps” occur primarily in narrative, creating depictions on places that may be mapped on a geographic grid that can never truly account for the density of experience that occurs in these sites. Heat-Moon’s use of the phrase, however, does not preclude earlier representations of the concept; the use of narratives that explore particular geographies is as old as the technology of writing. In fact, according to Heat-Moon’s conception of deep mapping, we might consider the medieval Icelandic sagas a deep map in their detailed portrayal of places, landscape, and the environment in post-settlement Iceland. Often occurring around the locus of a regional few farmsteads, the Sagas describe minute aspects of daily Icelandic life, including staking claim to beached whales as driftage rights, traveling to Althing (now Thingvellir) for annual law symposiums, and traversing valleys on horseback to seek out supernatural foes for battle. Adhering to a narrative form not seen again until the rise of the novel in the 18th century, the Íslendingasögur are a unique medieval exempla for Heat-Moon’s concept of deep mapping and the resulting geographic complexity that results from narrative. Thus, a ‘deep map’ may not only include a narrative, such as in the sagas’ plots, but potentially also a geographic map for the superimposition of knowledge upon it–allowing these layers of meaning to build and generate new meaning.

To tighten the data universe a little more: specifically within the sagas, I have chosen the outlaw-themed sagas for their shared thematic investment in place names and geography. Given that much of the center of Iceland today consists of glaciers and wasteland, outlaws had precious few options for survival once pushed to the margins of their society. Thus, geographic aspects of place name seem to be just as essential to the narrative of sagas as their more literary qualities—such as how they are used in sentences, or what place names are used to obscure or reveal.

Map of Iceland, by Isaac de La Peyrère, Amsterdam, 1715. via Cornell University Library, Icelandica Collection

In many ways, the question of “universe” for my data is the crux of my research question itself: how do we account for the different intersections of universes—both imaginative and literary, as well as geographic and historical—within our unit of analysis?

Unit of Analysis

If we dissect the element that allows geospatial and literary forms to interact, we arrive at place name. Place names are a natural place for this site of tension between literary and geographic place, since they exist in the one shared medium of these two modes of representation: language. In their linguistic as well as geographic connotations, place names function as the main site of connection between geographic and narrative depictions of space, and it is upon this argument that this project uses place name as its unit of analysis.

Methodology

Alright, now that we’re out of the figurative woods, on to the data itself. Here are the steps I used to create a geospatial map with metadata for these saga place names.

Data Collection, Place Names:

The print text was scanned and digitized using ABBYY FineReader 11.0, which performs Optical Character Recognition to ensure PDFs are readable (or “optional character recognition, as I like to say) and converted to an XML file. I then used the flexible coding format of the XML to hand-encode place name mentions using TEI protocol and a custom schema. In the XML file, names were cleaned from OCR and standardized according to Anglicized spellings to ensure searchability across the data, and for look-up in search engines such as Google–this saved a step in data clean-up once I’d transformed the XML into a CSV.

 

KinniburghXML

Here’s the TEI header of the XML document–note that it’s nested tags, just like HTML.

Data Extraction / Cleanup

In order to extract data, the XML document was saved as a CSV. Literally, “File > “Save As.” This is a huge benefit of using flexible mark-up like XML, as opposed to annotation software that can be difficult to extract data from, such as NVivo, which I wrote about here on Columbia University Library’s blog in a description of my methodology. In the original raw, uncleaned file, columns represented encoding variables, and rows represented the encoded text. I cleaned the data to eliminate column redundancies and extraneous blank spaces, as well as to preserve the following variables: place name, chapter and saga of place name, type of place name usage, and place name presence in poetry, prose, or speech. I also re-checked my spelling here, too–next time, no hand-encoding!

 

KinniburghCSV

Here’s the CSV file after I cleaned it up (it was a mess at first!)

KinniburghCSVKey

I saved individual CSVs, but also kept related info in an Excel document. One sheet, featured here, was a key to all the variables of my columns, so anyone could decipher my data.

Resulting Metadata:

Once extracted, I geocoded place names using the open-source soft- ware Quantitative Geographic Information Systems (QGIS), which is remarkable similar to ArcGIS except FREE, and was able to accommodate some of those special Icelandic characters I discussed earlier. The resulting geospatial file is called a shapefile, and QGIS allows you to link the shapefile (containing your geospatial data) with a CSV (which contains your metadata). This feature allowed me to link my geocoded points to their corresponding metadata (the CSV file that I’d created earlier, which had place name, its respective saga, all that good stuff) with a unique ID number.

Data Visualization, or THE BIG REVEAL   

While QGIS is a powerful and very accessible software, it’s not the most user-friendly. It takes a little time to learn, and I certainly did not expect everyone who might want to see my data would also want to learn new software! To that end, I used the JavaScript library Leaflet to create an interactive map online. You can check it out here–notice there’s a sidebar that lets you filter information on what type of geographic feature the place name comprises, and pop-ups appear when you click on a place name so you can see how many times it occurs within the three outlaw sagas. Here’s one for country mentions, too.

 

KinniburghLeaflet

Click on this image to get to the link and interact with the map.

Takeaways

As the process of this documentation highlights, I feel that working with data is most labor-intensive when it comes to positioning the argument you want your data to make. Of course, actually creating the data by encoding texts and geocoding takes a while too, but so much of the labor in data sets in the humanities is intellectual and theoretical. I think this is a huge strength in bringing humanities scholars towards digital methodologies: the techniques that we use to contextualize very complex systems like literature, fashion, history, motherhood, Taylor Swift (trying to get some class shout-outs in here!) all have a LOT to add to the treatment of data in this digital age.

Thank you for taking the time to read this–and please be sure to let me know if you have any questions, or if I can help you in mapping in any way!

In the meantime, stay tuned for another brief blog post, where I’ll solicit your help for the next stage of this project: visualizing the imaginative components of place name as a corollary to this geographic map.

Journaling Data (what else?)

“Day 1”

Challenge number one is finding a dataset I can work with. I want to do something with text. There are plenty of books available digitally, but they always seem to be in formats, such as PDF, that require a fair amount of cleaning (which I would rather not do) before they can be consumed by any of the tools.

I did find a number interesting datasets on the NYC Open Data website. Many of them are easy to download. I downloaded one small one that contains the average SAT scores for each High School. I was able to bring the dataset into R, too. It’s such a simple dataset–basically just school id plus the scores of each section of the SAT–that there isn’t much analysis one can do (mean, median, standard deviation…?). I would like to combine it with some other datasets that could enrich the data. For example, if I could get the locations of the schools, I could map the SAT data or if I could get demographic data for each school, I could correlate SAT performance to various demographic variables.

On a parallel track, I am working with Joy, who is very interested in motherhood and wants to explore datasets related to that subject. If she can find a compelling dataset, we will work together to analyze it.

 

“Day 2”

So it turns out that Project Gutenberg has books in TXT format that you can download. It also appears that there is a website www.zamzar.com that will convert PDF files to text files for free. Question for further thought: if the PDF is a scanned image, will converting it to text get me anywhere? I doubt it. Best way to find out is to give it a go.

I am going to download Great Expectations  from Project Gutenberg and see if I can pull it into R and, perhaps, Gephi or Antconc to see what can be done that might be interesting.

“Day 3”

Stay tuned…

Mapping Data: Workshop 3/3

Hi all,

Just to follow up on Mary Catherine’s post about finding data, I wanted to recap the final session of this workshop series that took place tonight.

The library guide on mapping data (by Margaret Smith) can be found here: http://libguides.gc.cuny.edu/mappingdata

As in the other two workshops, Smith emphasized thinking about who would be keeping this data and why as a part of the critical research process. It’s especially interesting given the size of these data sets and maps, meaning that the person (or corporate entity, NGO, or government agency) likely has a very specific reason for hosting this information.

She brought us through a few examples from basic mapping sites, like the NYT’s “Mapping America” which pulls on 2005-2009 Census Bureau data, to basic mapping applications like Social Explorer (the free edition has limited access, but the GC has bought full access) and the USGS and NASA mapping applications. The guide also includes a few more advanced mapping options, like ArcGIS, but the tool that seemed most useful to me, in the short-term anyway, is Google’s Fusion Tables, which allows you to merge data sets that have terms in common. The example Smith used was a data set of demographic data (her example was percentage of minority students) organized by town name (her example was towns in Connecticut) and a second set of data that defined geographic boundaries by the same set of towns. Fusion Tables then lets you map the demographic data and select various ways to visualize and customize your results.

My main takeaway from this series was that each of these tools is highly particular and unique, and you have to really dig into playing with the individual system before you’ll even know if it is the right tool for your work.

That, and also learn R.

Dataset Project: Who do you listen to?

The basis of my dataset is my iTunes library. I chose this because it was easily accessible, and because I was interested to see what the relationships in it would look like visualized. My 2-person household has only one computer (a rarity these days, it seems) which means that everything on it is shared, including the iTunes library. Between the two of us, we’ve amassed a pretty big collection of music in a digital format. (Our physical (non-digital) music collection is also merged but it is significantly larger than the digital one and only a portion of it is cataloged, so I didn’t want to attempt anything with it.)

I used an Apple script to extract a list of artists as a text list, which I then put into Excel. I thought about mapping artists to each other through shared members, projects, labels, producers etc, but after looking at the list of over 2000 artists (small by Lev Manovich standards!), I decided that while interesting, this would be too time consuming.

My other idea was easier to implement: mapping our musical affinities. After cleaning up duplicates, I was left with a list of around 1940 artists. Both of us then went through the list and indicated if we listened to that artist/band/project, and gave a weight to the relationship on a scale of one to five (1=meh and 5=essential). It looked like this:

 

Screen shot 2014-10-27 at 9.57.03 PM

Sample of the data file

Ultimately I identified 528 artists and J identified 899. An interesting note about this process, after we both went through the list, there were approximately 500 artists that neither of us acknowledged. Some of this can be chalked up to people on compilations we might not be that familiar with individually. The rest…who knows?

Once this was done I put the data into Gephi. At the end of my last post I was having trouble with this process. After some trial and error, I figured it out. It was a lot like the process I used with the .gdf files in my last post. The steps were: save the Excel file as CSV and uncheck append file extension, then open that file in TextEdit and save as UTF-8 format AND change the file extension to .csv. Gephi took this file with no problems.

The troublesome process of coding and preparing the data for analysis done, it was time for the fun stuff. As with my last visualization, I used the steps from the basic tutorial to create a ForceAtlas layout graph. Here it is without labels:

w_I_u_2

The assigned weight to each relationship is shown in the distance from our individual node, and also in the thickness of the edge (line) that attaches the nodes. It can be hard to see without zooming in closely on the image, since with so many edges it is kind of noisy.

Overall, I like the visualization. It doesn’t offer any new information, but it accurately reflects the data I had. Once I had the trouble spots in the process worked out, it went pretty smoothly.

I am not sure if ForceAtlas is the best layout for this information. I will look into other layout options and play around with them, see if it looks better or worse.

I made an image with the nodes labeled, but it become too much to look at as a static image. To this end, I want to work on using Sigma (thanks to Mary Catherine for the tip!)  to make the graph interactive, which would enable easier viewing of the relationships and the node labels, especially the weights. This may be way beyond my current skill level, but I’m going to give it a go.

ETA: the above image is a jpeg, here is a PDF to download if you want to have better zoom options w_I_u_2

Finding Data: Preliminary Questions

Hello, all,

As promised, here’s a link to the “Finding Data” library guide on the Mina Rees Library site. Apologies if someone has posted it already!

http://libguides.gc.cuny.edu/findingdata

The guide was created by the wonderful Margaret Smith, an adjunct librarian at the GC Library who is teaching the workshops on data for social research. There’s one more–Wednesday, 6:30-8:30pm downstairs in the library in one of the computer labs–and I’m sure she’d be happy to have anyone swing by. Check out the Library’s blog for details.

Within this guide, the starting questions that Smith provides, in order to get you thinking of your dataset theoretically as well as practically, are very helpful–and I wish I had them years ago! Here are some highlights, taken directly from the guide (but you should really click through!):

HOW TO FIND DATA:

When searching for data, ask yourself these questions…

Who has an interest in collecting this data?

  • If federal/state/local agencies or non-governmental organizations, try locating their website and looking for a section on research or data.
  • If social science researchers, try searching ICPSR.

What literature has been written that might reference this data?

  • Search a library database or Google Scholar to find articles that may have used the data you’re looking for. Then, consult their bibliographies for the specific name of the data set and who collected it.

HOW TO CONSIDER USING IT:

Is the data…

  • From a reliable source? Who collected it and how?
  • Available to the public? Will I need to request permission to use it? Are there any terms of use? How do I cite the data?
  • In a format I can use for analysis or mapping? Will it require any file conversion or editing before I can use it?
  • Comparable to other data I’m using (if any)? What is the unit of analysis? What is the time scale and geography? Will I need to recode any variables?

And another thought that I really loved from her first workshop in this series:

Consider data as an argument.

Since data is social, what factors go into its production? What questions does the data ask? And how do the answers to these questions, as well as the questions above, affect the ways in which that dataset can shed light on your research questions?

All fantastic stuff–looking forward to seeing more of these data inquiries as they pop up on the blog!

(again, all bulleted text is from the “Finding Data” Lib Guide, by Meg Smith, Last Updated Oct. 15 2014. http://libguides.gc.cuny.edu/findingdata)

MC

Data Set: Topic Modeling DfR

Hello Praxisers, I’m writing today about a dataset I’ve found. I’ll be really interested to hear any thoughts on how best to proceed, or more general comments.

I queried JSTOR’s dfr.jstor.org Data for Research for citations, keywords, bigrams, trigrams and quadgrams for the full run of PMLA. JSTOR gives this data upon request for all archived content. To do this I had to request an extension of the standard 1000 docs you can request from DfR. I then submitted the query and received an email notification several hours later that the dataset was ready for download at the DfR site. Both the query and the download are managed through the “Dataset Requests” tab at the top right of the website. It was a little over a gig, and I unzipped it and began looking at the files one by one in R.

Here’s where I ran into my first problem. I basically have thousands of small documents, with citation info for one issue per file, or a list of 40 trigrams from a single issue. My next step is to figure out how to prepare these files so that I’m working with a single large dataset instead of thousands of small ones.

I googled “DfR R analysis” and found a scholar, Andrew Goldstone, who has been working on analyzing the history of literary studies with DfR sets. His GitHub  contains a lot of the code and methodology for this analysis, including a description of his use of Mallet topic modeling through an R package. Not only is the methodology available, but so is the resulting artifact, a forthcoming article in New Literary History. My strategy now is simply to try to replicate some of his processes with my own dataset.