Mapping Cassity's Poetry
Having confirmed the general sense that Cassity's poetry invokes a lot of place names, and that his critics tend to be interested in his geographic references, I wanted to isolate all of the location names from Cassity's born-digital poetry collection and place them on a map. Perhaps readers would want to locate poems that reference a specific place, or perhaps readers would want to scan points on the map to encounter poems by chance. The map visualizes the diversity of locales, and allow readers to navigate this collection of poetry in a new way.
The points on this map indicate locations mentioned in the poetry in Turner Cassity's born-digital materials. When you click on the pin, you should see an excerpt from the poem where the location is named, and, in many cases, a link to the full text of the poem.
What can this map and this view of Cassity's poetry offer? This view confirms that he invokes many and a wide geographic range of proper place names. When I look at the excerpts, I notice that Cassity often uses location names in lists, or close proximity within a line, even when, or maybe especially when, the locations are far afield.
How This Map Was Made
Text Analysis of Born-Digital Poetry
The digital format of born-digital archives means that a researcher who has access to the text files can easily use digital humanities tools to analyze them. Unlike manuscripts or typescripts in the paper collections, born-digital collections do not need to be transcribed or OCR-ed to be analyzed with digital tools.
In order to map the place names in the over 300 poems included in Cassity's born digital materials, I used the Namedropper lookup-names python script, which uses DBpedia Spotlight to pull out location names. Namedropper was developed by senior software engineer Rebecca Sutton Koeser at Emory Libraries. Koeser wrote about a similar mapping project that she did with Around the World in Eighty Days, running the lookup-names script against the novel's text. The script creates a CSV file (essentially a spreadsheet) of the recognized place names, a link to the DBpedia entry for the location, and an excerpt from the text including the place name. Sara Palmer, Electronic Text Specialist at ECDS, and Rebecca Sutton Koeser, LITS Senior Software Engineer, helped to prepare the data for the map.
I also experimented with Stanford Named Entity Recognizer (NER) to pull out place names, following William Turkel's NER tutorial. The results I got with the NER script were many fewer, and often inaccurate, so I've set them aside for now. Both Namedropper and NER provide results that are far from perfect--they each recognize some things as places that are not places (for instance, "alas" was identified as "Alaska"), and overlook actual place names. But, these named entity recognition tools allow us to quickly and automatically pull out location names, which is helpful for processing large amounts of text.
Since the time Rebecca Sutton Koeser wrote up her process for mapping locations from a text, the Google Maps API has changed, and now includes Google's own geolocator information. For the map of Cassity locations, I just provided the names of the locations that Namedropper generated, and relied on Google Maps to match the name to a particular longitude and latitude.
I suspect that Google Maps must have taken my location in the United States into account as it assigned geocoordinates to the place names, because the most noticeable mistakes occurred with mis-placements within the U.S. For instance, Luxor, the location in Egypt, was originally mapped as the Luxor hotel in Las Vegas, Nevada. Similarly, the Strip, which in Cassity's poem refers to the Gaza Strip, was located as the Las Vegas Strip. Other challenges arose with historical placenames. (For instance, "Ottoman" from a poem that mentions Ottoman Arabia, was originally placed in Covington, Kentucky, where a store called Ottoman Imports is located.) The map isn't perfect, but serves as a proof of concept for one possible application of digital analysis of born-digital archival materials.