Charter Map

I have decided to share the map of East Frankish Carolingian charters that I produced for my comprehensive field in Digital Humanities in 2019. I think it will even work on mobile, but your mileage may vary!

Here is my original discussion of how I made the map:

The Origin of the Project:

​This project largely arose out of a singular question: what is the relationship between kings and their charters? It first began with my MPhil dissertation at the University of Cambridge, which looked at the places where charters were redacted (drawn up). There it became clear that kings, especially in East Francia, had several preferred places for drawing up charters (Frankfurt and Regensburg). However, it was also clear that only looking at the places of redaction ignored who these charters were being written for. As such the next step was to look at recipients of charters. The number of differing recipients made a GIS approach much more favorable, because without it the geographic distribution of the recipients would be more difficult to understand. Indeed the role of computing in digital humanities, in the words of Nicholas Bauch, is not “an advance in computing, but an advance in the application of computing tools to traditional humanistic questions”. [1]

Possibilities:

​An important consideration when approaching a digital humanities project is what tool to use. In theory, one could apply mapping, network analysis, statistics, etc. to a wide variety of sources, but that does not mean one should. In this instance I chose to use mapping software, because the questions I was asking were best answered through a geographic approach. That is, I fit the tool to the question, not to the data. Below, I present some observations on other directions I could have taken this project by using other digital humanities tools.

Network analysis can be a powerful tool in understanding relationships which traditional historical methods cannot grasp. [2] For instance Cornell Jackson used the ‘People of Medieval Scotland’ database to investigate relationships between witnesses of charters using an affiliation network. Jackson’s preliminary findings revealed that Duncan II, the Earl of Fife was co-witness in a significantly large number of charters, making his role in Scotland even more important than historians previously thought. [3] In the case of Carolingian East Frankish charters, there are some difficulties: first, Carolingian royal charters do not include witnesses, and there are less results than in Jackson’s case. Carolingian charters do contain information about the intercessor, so a network analysis could use that data instead, but would still be limited by the number of charters.

Another area where digital humanities has been applied to medieval texts has been philological investigations of texts. [4] Using content analysis and other methods, historians can get a better understanding of these texts and even make claims about authorship using stylometric measures. Mike Kestemont, Sara Moens and Jeroen Deploige were able to convincingly argue on this basis that the texts Visio ad Guibertum missa and the Visio de Sancto Martino were reworked by Hildegard of Bingen’s secretary Guibert to such an extent that a computer could attribute them to Guibert. [5] There are a variety of other approaches used on texts, such as asking how accurate medieval scribes and reconstructing stemmata. [6] In the case of our charters, this linguistic approach is likely of limited usefulness, because generally they require longer texts and defined authors (which in many instances charters are neither long or the author is known). [7] However the case of Hildegard and Guibert does offer the possibility of using a computational linguistic approach to understanding scribal innovations or getting a better understanding of which scribes wrote certain charters. 

As should be clear from the preceding discussion, when seeking to understand the relationship between a charter recipient and the king, a map based approach was the right tool for the job. However, using a map also entailed decisions about how to collect the data, how to present the data, and how to understand the data.

Debating the Map:

Before even getting to the map, the data had to be collected. This involved working with the primary sources in order to gather the relevant information, as outlined in the Making the Map section. The process of entering information is not without pitfalls, and my approach largely conformed to Hadley Wickham’s concept of “Tidy Data” where each observation (in this case a charter) forms one row while each variable forms a column. [8] This process involved making a series of choices, about how to render names, but more fundamental questions about the nature of the charter and what was going to be presented on the final product. The “cleaning” of the data, in the sense of getting the data from the charters and into a format usable by Leaflet, was the most time consuming part of the project. [9]

​Care had to be taken when devising the map, because there are a variety of considerations when deciding on the type of map, the style, the data, etc. As the work of J.B. Harley has shown, maps were, and are not, neutral objects even if they claim to have accurate and precise data [10] Supposedly “true” maps can present their own arguments by the clever use of framing, colors, etc. [11] Because this map was meant to show the different geographical distributions of charter recipients, some choices had to be made.

First, by showing the recipients of charters it meant forgoing any sort of depiction of the relationship between a charter’s place of redaction and its recipient because it would quickly be overwhelming visually. Even with one king it can often be somewhat confusing to interpret, let alone with five kings all with their own networks overlaid on a map.

Second, the density of certain places necessitated combining points into a cluster, which is represented by a circle marker with a number on the map. While this allows a user to click the cluster and see all of the charters given to that location, it does not allow a quick glance to establish the spread of charters.

What may appear as a somewhat “basic” map was the result of a series of choices (sometimes rather difficult) about what data to collect and how to present it. Even the basemap is not without issues: do you present modern borders for data from a period before they existed? Or do the modern borders help to orient the user in regards to potentially unfamiliar places? In the end I retained several basemaps for users to select from. It is hoped that this map will not only support one interpretation but allow for users to visualize different combinations of data and to produce patterns from old data.[12] For instance, how do the charters of Louis the German and Arnulf compare? Charles the Fat and Louis the German? Charles the Fat and Arnulf?

Conclusion:

​Some conclusions present themselves when viewing the map: first, there were some very important recipients in East Francia, namely St. Gall (thirty charters), Regensburg (its various churches received twenty-two charters), Fulda (eighteen charters), and Lorsch (fourteen charters). Second, a good proportion of Carloman’s charters were destined for Italian recipients, which further cements the strong connection between Bavaria and Italy in this period. Third, most recipients of charters in Italy were concentrated in the Po valley, near Bergamo and Piacenza.

References:

​[1] N. Bauch, “Digital Geohumanities: Visualizing Geographic Thought”, International Journal of Humanities and Arts Computing 11, no. 1 (2017): 1-15; See also L. Espinha da Silveira, "Geographic Information Systems and Historical Research: An Appraisal", International Journal of Humanities and Arts Computing 8, no. 1 (2014): 28-45.

​[2] See S. Weingart, "Demystifying Social Networks" (revised 2012) for an overview: http://journalofdigitalhumanities.org/1-1/demystifying-networks-by-scott-weingart/ and also S. Borgatti, A. Mehra, D. Brass, and G. Labianca, "Newtork Analysis in the Social Sciences", Science 323 (2009): 892-895.

​[3] C. Jackson, "Using Social Network Analysis to Reveal Unseen Relationships in Medieval Scotland", Digital Scholarship in the Humanities 32, no. 2 (2017): 336-343; See also C. Wetherell, "Historical Social Network Analysis", International Review of Social History 43 (1998): 125-144.

​[4] The journal Literary and Linguistic Computing has a variety of articles on this topic. See also J. Binongo and M.W.A. Smith, "The Application of Principal Component Analysis to Stylometry", Literary and Linguistic Computing 14, no. 4 (1999): 445-465.

​[5] M. Kestemont, S. Moens, J. Doploige, "Colloborative Authorship in the Twelfth Century: A Stylometric Study of Hildegard of Bingen and Guibert of Gembloux", Literary and Linguistic Computing 30, no. 2 (2015): 199-224.

​[6] See M. Spencer and C. Howe, "How Accurate Were Scribes? A Mathematical Model", Literary and Linguistic Computing 17, no. 3 (2002): 311-322; M. Spencer and C. Howe, "Estimating Distances Between Manuscripts Based on Copying Errors", Literary and Linguistic Computing 16, no. 4 (2001): 467-484; T. Roos and T. Heikkilä, "Evaluating Methods for Computer-Assisted Stemmatology Using Artificial Benchmark Data Sets", Literary and Linguistic Computing 24, no. 4 (2009): 417-433.

​[7] See M. Eder, "Does Size Matter? Authorship Attribution, Small Samples, Big Problem", Literary and Linguistic Computing 30, no. 2 (2015): 167-182, where the author argues that a text less than 2500 words for Latin is likely too small to be reliable; But also see K. Luyckx and W. Daelemans, "The Effect of Author Set Size and Data Size in Authorship Attribution", Literary and Linguistic Computing 26, no. 1 (2011): 35-55 on how small text size can be controlled if the number of authors is limited.

​[8] H. Wickham, “Tidy Data”, Journal of Statistical Software 59, no. 10 (2014): 1-23; For a discussion on the limitations of using structured data with historical sources see J. Bradley, "Databases and GIS: A Critical Approach: Silk Purses and Sow's Ears: Can Structured Data Deal with Historical Sources?" International Journal of Humanities and Arts Computing 8, no. 1 (2014): 13-27.

​[9] On the importance of cleaning (albeit with an argument against using the term) see K. Rawson and T. Muñoz, “Against Cleaning” (2016), http://curatingmenus.org/articles/against-cleaning/

​[10] J.B. Harley, “Maps, Knowledge, and Power”, in P. Laxton (ed.), The New History of Maps: Essays in the History of Cartography (Baltimore, 2002), pp. 52-82; J.B. Harley, “Deconstructing the Map”, Cartographica: The International Journal for Geographic Information and Geovisualization 26, no. 2 (1989): 1-20.

​[11] S. Schulten, “The Cartography of Slavery and the Authority of Statistics”, Civil War History 56, no. 1 (2010): 5-32; M. Monmonier, How to Lie with Maps, Second Edition (Chicago and London, 1996).

​[12] On data visualization see S. Sinclair, S. Ruecker, M. Radzikowska, “Information Visualization for Humanities Scholars”, Literary Studies in the Digital Age. MLA Commons (2014), https://dlsanthology.mla.hcommons.org/information-visualization-for-humanities-scholars/

Collecting the Data:

​Charters present challenges to creating a data model due to inconsistencies in spelling, corruption, interpolation, etc. As such care must be taken when dealing with their provenance and transmission history. Spelling differences presented the most obvious issues, and have been condensed into the modern form regardless of differences in the sources (that is, where the locations are the same, I have not maintained the differences). However, this method by nature privileges two types of data: those where the location can be identified with a modern location, and those charters which were granted to a location (that is, usually to a church or monastery). Because of this the dataset is not necessarily representative of all charters, but those given to institutions. In some instances this presents problems because land may have been given to an individual, but on their death the land went to the institution. How should this charter be categorized? Generally, charters to individuals have been excluded, but in those instances where a beneficiary institution can be identified, they have been included. One additional issue is bishops: do charters to them get included? For instance if the bishop of Paderborn received a grant, does this get placed on the map at Paderborn? In most instances I have excluded these charters, unless the charter also specifies that the land went to the bishopric as well. At all levels choices were made which inevitably shapes the data presented, no dataset is “neutral” no matter how tidy it looks.

 The basic outline for the data collected (the “data model”) is as follows:

The MGH number of the charter in the appropriate volume

The date of the charter (where possible, separated by day, month, year)

The location the charter was redacted

The location of the recipient (where possible, see above)

The name(s) of any intercessors

If the charter is an original or not

Any other notes on that charter

Mapping the Data:

​From this model only the date, redaction location, and recipient location are incorporated into the map. The decision was made to map the recipient location because the purpose of the map was to see if there were any geographic patterns to charter recipients. Further, where a location received more than one charter from a king, it has been condensed into one point which will present all relevant information when selected. A similar project could be done for the location of redaction, which may reveal patterns in which centers were used frequently by Carolingian rulers. This, however, is somewhat simpler to do without physically mapping the data because Frankfurt and Regensburg dominate as the place of redaction. In the case of recipient locations the variety is much more pronounced and a mapping approach better suits this data. However, a combination approach would also be very informative: by showing both the recipient location and connect that to the places of their redaction, the relationship between certain centers and certain regions could be explored in a way that would be difficult without a map. A further addition to the project seeks to add this functionality.

Before actually mapping the data, it was necessary to get the latitude and longitude of each location so that it could be plotted on a map. The relevant data then needed to be converted from a CSV file into a geoJSON file so that Leaflet could interpret it. Luckily, the online tool geojson.io does just this, which allowed for a quick conversion process. From there the geoJSON file could be incorporated into Leaflet.

There are several issues with combining various kings into one map: the different lengths of reigns and different areas of rule. Louis the German ruled for over fifty years (822 to 876), while his sons ruled for between four and eleven years (Carloman ruled 876 to 880, Charles the Fat ruled 876 to 887, Louis the Younger ruled 876 to 882). Further compounding this issue is that they did not always rule the same geographic region. Louis the German and Arnulf bought ruled East Francia as a sole ruler, while Louis the German’s sons split the kingdom into different kingdoms under each of them. Charles the Fat also ruled the entire kingdom from 882 to 887, but he also ruled West Francia from 884 to 887. Because of this I made the decision to not separate out the rulers into different time periods, which are revealing but may be misleading without further context.

Software Used:

 1) Leaflet.js: a lightweight Javascript package for producing maps. Can be accessed here. I also used several Leaflet plugins:

• Custom colored icons by pointhi which can be accessed here
• Leaflet Marker Cluster Group which can be accessed here
• Leaflet Feature Subgroup by ghybs which can be accessed here

​2) Geojson.io: This website allows a user to take a CSV file of coordinates and turn that into a GeoJSON file, which is one of the standard file formats for Leaflet. Can be accessed here

​​The code itself can be found over on my Github page: https://github.com/jondell518/CharterMaps.

​Spot a mistake or want my data? Contact me at 02dellisola@cua.edu.

Data Sources:

​The data was taken from the respective volumes of the Monumenta Germaniae Historica:

Louis the German, Louis the Younger, and Carloman: Die Urkunden Ludwigs des Deutschen, Karlmanns und Ludwigs des Jüngeren, P. Kehr(ed.), MGH Dip. Germ. I (Berlin, 1934). Digital version.

Charles the Fat: Die Urkunden Karls III, P. Kehr (ed.), MGH Dip. Germ. II (Berlin, 1937). Digital version.

Arnulf: Die Urkunden Arnolfs, P. Kehr (ed.), MGH Dip. Germ. III (Berlin, 1940). Digital version.