A team of scholars and technologists at the Emory Libraries led by Rebecca Sutton Koeser and Brian Croxall are developing tools for identifying and marking up names, places, and organizations in Emory’s collection of materials associated with the poets known as the Belfast Group. Tagging these entities will make it possible to examine and present some of these writers’ social and geospatial networks. Because it connects identifiers to the semantic web, the tools created for the Networking the Belfast Group project will give users access to much more data than documents like finding aids regularly provide.

In order to compare the team’s tools with doing the same work by hand, I tagged the people, businesses, and places in the Frank Ormbsy finding aid with identifiers. These identifiers connect people to resources of linked data, providing more information about “Frank Ormsby” and establishing that we’re always talking about the same entity throughout our finding aids. That might sound easy enough with someone named “Ormsby,” but when you start trying to establish which “Robert Johnson” a particular finding aid is referencing, it becomes much more complicated. Marking up the 14,000 lines of XML took even more time than we expected—about fifty hours. The vast majority of my time was spent determining who people were in the database we are using for our unique identifying numbers, the Virtual International Authority File (VIAF).

The other issues I encountered centered on how language and networks function. The first was, bizarrely enough, a question about parts of speech. If a group is said to be Northern Irish, do I mark that phrase as a geographic place name? If the text said, this group was from Northern Ireland, I would have not even questioned marking it with the identifier for “Northern Ireland.” When it was an adjective, I hesitated and conferred with the other team members. Is this because adjectives describing groups of people have more wrapped up in them than geography? Is it just a categorization hiccup?

The second kind of question I asked was about relationships. If a poem has Belfast in the title, should this be given a geographical identifier? If a poem title has the name T.S. Eliot in it, does T.S. Eliot get marked as the person? The relationship is quite different from most of the marks, where T.S. Eliot would have authored materials in the collection or where Belfast would have been a site of production or distribution rather than a subject. When something is about, rather than by a person, should it be tagged? And whatever we decide on that question, should the same policy apply to places when they are used either as subjects or sites? Similarly, there were reviews of books where I tagged the authors of both the text being reviewed and of the review itself. The actual collection houses only the reviewer’s work; however, this seemed a different kind of “about” than a poet writing about another person.

I also ran into questions about how to differentiate members of families. When a letter is from the Longleys—a husband and wife—whose identifier do we use? When a Longley is referred to without the first name, how do I decide which Longley is meant? Sometimes it was obvious that Michael Longley was meant, because he is a poet and the collection includes more of his materials; however, I could see how making assumptions about who to identify might reduce the presence, particularly of women, in these networks.

At the center of these questions is how we understand networks. Knowing that two entities are related is quite different from knowing how two entities are related. These questions have implications not only for the functions of the tools created by the team, but also for how researchers will use and understand the networks we produce using them.

Saxifraga cochlearis

In his poem A Sort of a Song, William Carlos Williams wrote “no ideas but in things” and “saxifrage is my flower that splits the rocks.” What I’m doing here almost certainly isn’t what he meant– in fact, I may be doing the reverse, in that I am taking a poem and words and, in a sense, converting it back to, or at least representing it as, its component “things.” Even though it isn’t quite what Williams intended, these lines kept coming to mind as I worked ont his post, and it seems related to the things in poetry I’m discussing here.

Early on, near the beginning of this project, when we were experimenting with some of the tools and technologies we thought we might use to improve the process of identifying and tagging names in XML text, I noticed some strange output when I ran some of the poetry from the Belfast group sheets against the DBPedia Spotlight annotation service. Because I wasn’t restricting the identified resources to persons, places, or organizations (which is what our tools usually do when we’re trying to identify names to be tagged, e.g. in the NameDropper OxygenXML plugin we’re developing), it was identifying things like “potato”, “rock”, “eye”, “mouth”, “hand”, and “root” in the text. We’re now at the point in the project that we’re starting to shift towards using the tools we’ve been developing to enhance the EAD and TEI XML associated with the Belfast Group, and as I’ve begun working on tagging some of the poetry I was reminded of this and thought it might be worth a little more investigation and thought.

For this experiment, I restricted myself to Seamus Heaney’s poem Digging, as it appears in the draft on one of the Belfast group sheets (there are some slight wording differences from the published version).

Below are the things that DBpedia Spotlight identifies in the poem. I’m using the DBpedia thumbnails (or Wikipedia thumbnails, in the few cases where the DBpedia thumbnail image link was broken) to emphasize the “thingness” of the entities that Spotlight recognizes. Each image links to the corresponding DBpedia resource, and if you hover your mouse over the image you should see a snippet of the poem where the entity was recognized. I’ve sorted them out into three groups semi-manually, since I’m still having difficulty filtering based on support and similarity scores without losing useful data, although in this case it seemed like very few of the identified resources had high certainty, I suspect due to the poetic language.

First, the things that DBpedia Spotlight recognized accurately, in the order that they occur in the poem.

Finger

Thumb

Pen

Gun

Window

Sound

Spade

Sink

Rhythm

Potato

Drill

Boot

Knee

Root

Brightness

Scattering

Potato

Hardness

Hand

God

Handle (grip)

Spade

Grandparent

Cutting

Peat

Bog

Milk

Bottle

Paper

Drink

Sod

Shoulder

Good and evil

Potato

Slapping

Peat

Cutting

Spade

Man

Finger

Thumb

Pen

It’s sort of an odd way to read a poem, but it’s also kind of intriguing. Among other things, I think this highlight how full of actual physical items, especially body parts, the text is.

Second, a few of the resources that aren’t quite correctly matched up to the text, but are still interesting and semi-relevant.

Squatting

Coarse fishing

Lugger

Shaft mining

Tell

Fell

I actually found these mis-identifications somewhat thought-provoking. To some degree, they betray the extent to which DBpedia is thing-centric, so that verbs and adjectives are mis-identified as nouns (again, with low confidence or support scores). But I find the notion of the poet’s pen “squatting” between thumb and finger, in the sense of taking up residence in an abandoned space without permission, rather appealing and fascinating. In the case of some of the other mis-identifications, it seems that Spotlight is picking up the context of digging and working outdoors, hence the mountains and archeological entities. And in the case of the lugger ship, this mis-identification actually drove me back to the text, and when I looked at “lug” in context I discovered that I didn’t actually know what it was, and had to go looking to figure out that the lug and shaft are parts of a shovel or spade.

Third, some of the mis-identified things that are humorously, obviously wrong. In this case have actors and musicians or bands, conceptually unrelated items, and even a video game. I’m including these here partly because they make me laugh, but also to demonstrate that the technology still has limitations and we need to be careful how we apply it.

Doris Day

Toner

Vomiting

Common cold

Molding (process)

The Edge

The Roots

Dig Dug

For those who are interested, here are some technical notes on how I generated this post.

1. Got a copy of the TEI xml for the Heaney Belfast group sheets from the current Beck Center Belfast Group site (now available on GitHub!)
2. Ran the NameDropper lookup-names python script on the TEI file, restricting it to the poem I was interested in and setting the certainty pretty low, to generate a CSV file.

       lookup-names heaney1.xml --tei-xpath '//t:body[@xml:id="heaney1_1045"]' -c 0.1 \
        --scores --csv /tmp/heaney-digging.csv
     

3. Wrote a simple python script to iterate through the CSV file and generate the HTML I wanted for each item, pulling the label and thumbnail from DBpedia, and using the context pulled from the poem.
4. Manually sorted out the entities I wanted into the three groups, preserving order, and fixed missing thumbnails where I could (some of the DBpedia thumbnail references are invalid; I’m guessing this is because they have been updated on Wikipedia since the last time the current DBpedia data was regenerated).

What follow is a poster proposal for the 2013 Digital Humanities conference that both Brian and I wrote, as project manager and lead scholar/developer, respectively. Our fingers are crossed, and we’ll know more around 1 February 2013.

Networking the Belfast Group through the Automated Semantic Enhancement of Existing Digital Content

There is increasing work on and interest in social networks in the digital humanities community (Meeks 2011). Analysis is frequently done on digital content—including images (Akdag Salah et al. 2012); email (Hangal et al. 2012); and citation networks (Visconti 2012)—because the data lend themselves to aggregation, conversion, and analysis. Yet despite this flurry of activity, the possibility exists for an exponential jump in network analysis. After all, the holdings and catalogs of galleries, libraries, archives, and museums (GLAMs) include traces of vast paper-based networks, but the data are locked away in forms that don’t easily lend themselves to analysis. What if we could open up that content? In this poster, we will report on an attempt to provide tools for archivists to expose the information embedded in the descriptions of their collections as well as a test case for analyzing that data: an examination of the networks of the Irish poets collectively known as “the Belfast Group.”

Our goal is to develop software tools and design a workflow to enhance TEI and EAD—documents that are already commonly created and maintained by archivists and text centers—without radically increasing the time and effort involved. The software tools (http://github.com/emory-libraries-disc/name-dropper) consist of a plugin for the Oxygen XML editor and command line scripts that will, first, make use of DBPedia Spotlight to identify and annotate recognized names and other resources within the text and, second, connect to linked-data systems (starting with the Virtual International Authority File [VIAF]) to provide authoritative, scholarly identifiers.# The scripts will allow technical users to inspect and tune the results or to automatically tag high-certainty resources, and the plugin will provide a user-friendly interface to review and accept suggested names while editing a document. The enhanced documents should provide significant benefits to GLAMs, allowing them to connect disparate types of content (e.g., digitized texts or photographs from an archival collection) and augment with data from other linked data systems. Furthermore, the enhanced documents will make it possible to expose these data in more machine-readable and research accessible formats. Our tools and workflow could be applied to resources held by different archives (for a different approach, see Blanke et al. 2012). What’s more, enhancing these documents helps GLAMs provide a means for researchers to do non-consumptive, social network research on the metadata of collections that might otherwise be closed or problematic in other ways (e.g., restricted correspondence from living authors).

Although our tools are not yet complete, we have already begun preliminary visualization and analysis of network relationships using data that mirrors what we will generate automatically by Summer 2013. The difficulties of defining “the Belfast Group” make for a compelling test case for our attempt to understand networks via data that are newly machine readable. The Group is a contentious network since the label has been variously applied to a weekly writing workshop that ran from 1963-1972, the most famous poets who attended that workshop—including Seamus Heaney, Michael Longley, and Paul Muldoon—or more loosely applied to all of the writers who “put Belfast on the literary map” (Clark 6). The significance of the writing workshop is debated by critics and often rejected by the poets themselves, sometimes vehemently. In contrast to a more formalized group, some scholars identify “an informal community” of poets evidenced by their letters, promotion of each other, and poems dedicated to each other (Drummond 32), connections which are richly documented by archival materials held at Emory University.

Using preliminary data manually generated from a subset of the correspondence EAD, our data suggests a wider set of connections in the Group than traditional scholarly approaches. The latter selectively emphasize the relationships of the most prominent authors and the role of the writing workshop (see fig. 1). Since our data is based on a much larger set of artifacts, as well as their complete metadata, we find that the locus of poetic activity in Belfast is not so oriented around the workshop (see fig. 2). Once we collect the full dataset via our completed tools and workflow, we will compare it with models generated by traditional scholarly methods, to identify significant gaps and discrepancies in either model.

Providing not only this new analysis of the Belfast Group’s network and a report on the development of our tools, our poster presentation at DH 2013 will also include a hands-on demonstration of the software tools and interactive visualizations of network data.

Figure 1. Graph of relationships inferred from Heather Clark’s Ulster Renaissance. Nodes are sized by degree and colored by hub score. The writing workshop is the strongest hub; the trio of large nodes represent Michael Longley, Derek Mahon, and Seamus Heaney.

Figure 2. Relationship graph based on preliminary correspondence data, sized and colored as in figure 1. Based on this data, the writing workshop does not function as a hub at all, and Paul Muldoon becomes the largest node.

References