Session 1: Canon


David J. Wrisley

Monday 17 September, 9:15 AM

New York University of Abu Dhabi [AE]

Recasting the can(n)ons: Towards a New Generation of Computational Medieval French


One might claim that the notion of a canon of medieval French began with Third Republic learned societies and the launching of their publication series. In my paper, I address visions of the canon of medieval French over time up to their evolution into computationally tractable corpora in the late twentieth century. I look to the New Philology debates of the 1990s, in particular, to reassess their competing visions of a future of computer-assisted reading. Using textual experiments of my own in both computational alignment and stylometry, I will demonstrate how we can expand our vision of medieval French literary history beyond well-known texts and authors. I argue that innovative computational research in literary history requires a willingness to recast canons into corpora that address questions of mutual scholarly interest in the field. Yet, the few hundred texts in medieval French currently available for different modes of computer-assisted criticism represent only a fraction of known works that are already edited. Finally, with quantified publication histories as evidence and an action-based agenda in mind, I suggest some textual starting points for the field to craft a new generation of computational medieval French.


Jean-Baptiste Camps & Julien Randon-Furling

Monday 17 September, 10:00 AM

École des chartes [FR] & Panthéon-Sorbonne University [FR]

Was There a “Medieval Literary Canon” in the Middle Ages?


How did a handful of medieval texts gain a lot of traction, resulting in abundant manuscript traditions, while many others are lost or barely preserved in a single witness? Inside the tradition of one text, how did many branches go extinct, while one, sometimes two specific versions were copied again and again?

In a context marked by textual variation, modifications inherent to the process of manuscript transmission, diffraction, rewriting, prosification or re-versification, and by esthetics favouring varietas, and variation on a given theme sometimes termed variance or mouvance, how did some texts or text-versions gained authority, up to the point were we consider them vulgate, and what led some authors and scribes to try to reconcile varying versions, sometimes in a pre-philologic fashion?

This communication will articulate the questions related to the distribution of medieval witnesses, the dynamics of manuscript populations (“birth” and “death” rates), the shape of stemmata and the processes of contamination, to try to gain insight on how medieval literary reception could have included the elaboration of canons of specific texts or text-versions, and how the coexistence of strongly represented versions could have created dissonances for a medieval audience.

Examples will be taken from XIIth and XIIIth century French and Occitan literature, chiefly chansons de geste and lyric poetry.


Julie Orlemanski

Monday 17 September, 11:00 AM

University of Chicago [US]

How to Read Exempla: Challenges of a Text Type


This paper argues that we still know very little about how to read, or to study, medieval exempla, the short illustrative tales that arguably constituted the most widely disseminated narrative form of the Middle Ages. Should the exemplum be considered a genre, a rhetorical ornament, or a narrative mode? How should exempla be historicised, given the stories’ durability across centuries, given the multiple traditions (classical, rabbinic, Arabic, “folk”) that converged in medieval exempla, and given the fact that the fate of exempla are entangled in, but not limited to, the vicissitudes of encompassing genres like handbooks for preachers and mirrors for princes? What kind of authority did medieval authors understand exempla to have? How did the stories’ combination of truth and polysemousness operate in practice? Such questions do not have singular, all-purpose answers, of course – but together they indicate why the exemplum can act as the spur for rethinking the medieval literary canon and how we read it. Exemplary stories demand scholarly attention to matters of scale, authorship, semantic under-determination, historicization, and the nature of literary categories. My remarks will touch on two traditions of studying exempla, the folkloric and the literary-critical, and I will pose questions about how digital technologies might help to bridge those approaches. Rather than a focused case study, this paper is a provocation to reflect on how the exemplum, that distinctively medieval text type, might reshape some of the concepts and methods of literary study.


Wouter Haverals

Monday 17 September, 11:35 AM

University of Antwerp [BE]

The Measure of Middle Dutch: Empirical Assessment of Aesthetic Observations on the Rhythm of Middle Dutch Rhymed Literature


What does it mean when the rhythm of a literary text is called ‘snappy’ or ‘fluid’? The rhythmical qualities of literature are often described on an intuitive basis, while using vague terms. This especially holds true for many of the rhymed texts of our literary history’s earliest stages. Ad hoc claims about rhythmic qualities of texts abound in secondary literature, especially when it comes to the canon of Middle Dutch literature. Mostly in prefaces to text editions and books of literary history, we find editors referring to a text’s rhythm in order to boost or add to its prestige. In some cases, this even happens in the form of a comparison, where the rhythmical mastery of one author or text is set against the rhythmical amateurishness of another.

The study of the rhythm of medieval Dutch literature has much to gain from a computational approach, where biases and personal taste are ruled out. In this conference paper, I will present a computational, data-driven approach, which attempts to reconstruct and study the rhythm of Middle Dutch texts in a bottom-up fashion. Taking advantage of usage-based statistics, a typology of rhythm in Middle Dutch literature will be presented. Next, we will pursue an empirical assessment of a selection of aesthetic observations. Singling out some of the most persistent claims about Middle Dutch’s most notable works and authors (e.g. Maerlant, Velthem, Boendale, Hadewijch), we will verify whether these expert claims can be confirmed using our rhythmical analyses. Their potentially biased opinions will be considered in the light of the results achieved with our computational technique.


Session 2: Text Editions


Peter Robinson

Monday 17 September, 2:00 PM

University of Saskatchewan [CA]

Collaborative Online Editing of A Canonical Textual Tradition


Although the special problems of large textual traditions (arising from the mass of highly-structured data) make them particularly suitable to computer methods, tools to help with the editing of such traditions have been slow to arrive. This talk will discuss the theoretical and practical problems of working with canonical textual traditions, focussing on my work with Dante and Chaucer. It willl introduce and demonstrate  the Textual Communities system (www.textualcommunities.org), designed specifically to assist with the editing of many texts in many versions.


Lydia Wegener & Nadine Arndt

Monday 17 September, 2:45 PM

Berlin-Brandenburg Academy of Sciences and Humanities [DE]

The Benefit of Digital Editions for the Reconstruction of Mystical Discourse: Two Composite Treatises as Test Case


If we turn our attention to the religious culture of 14th -century Germany, we inevitably encounter the big names of ‘Rhineland Mysticism’: Meister Eckhart, John Tauler, and Henry Suso. Though the enormous impact of their vernacular writings on Late Medieval literature is beyond doubt, we still know very little about the different ways in which their treatises and sermons were read, discussed, extracted and combined to create new texts. This shortcoming  in research is essentially attributable to two factors: firstly, the perseverance of an author-centred perspective, and secondly, the lack of modern editions of highly variant compilations.

These so-called ‘mosaic-treatises’ and ‘composite treatises’ are of major importance within the literary landscape of the Late Middle Ages. Although they are not the works of canonical authors, they demonstrate how different aspects of ‘Rhineland Mysticism’ were received by a mainly monastic audience. They constitute a broad discursive field which offers us deep insights into the predilections, reservations, and terminological innovations of mystically interested male and female compilers.

In our paper we want to discuss the two related composite treatises Sant Johannes sprichet ‘ich sach daz wort in gote’ (Saint John says: ‘I saw the Word in God’) and Die drîe persône geschuofen die crêatûre von nihte (The Three Persons Created the World from Nothing). In the first part we will refer to the fluctuating transmission of the texts in the manuscripts and present the concept for a digital edition which is adaptable to different user profiles dependent on individual research interests.

In the second part of our paper we will discuss the highly speculative contents of the treatises (for example their notion of the relationship between man, God and the Godhead) which are clearly influenced by Meister Eckhart but pursue their own paths. By means of significant examples we will demonstrate how a digital edition helps us to reconstruct the vitality of mystical discourse in the Late Middle Ages which is largely independent of canonical authors and derives its authority from other sources.


Miriam Edlich-Muth

Monday 17 September, 3:45 PM

University of Düsseldorf [DE]

Using Interactive Network Graph Editions of Medieval Manuscripts to Explore the Principles of New Philology


The New Philological drive towards engaging with manuscripts has lead many scholars to embrace the growing number of digital manuscript facsimiles that can be found online. However, what is lost in our encounters with these visual images of manuscript folios is the engagement with the codex as object. This untethering and decorporealisation of the manuscript from the world of objects renders it more vulnerable to being interpreted only within the bounds of the metaphor by which it is organised online: that of a book whose pages can be turned and whose contents can be navigated with reference to indexes. As a result, students and scholars attempting to ‘return to the original’ lose the possibility of navigating manuscripts in ways that are flexible and eclectic rather than directional. And yet, there is a great deal of evidence to suggest that early medieval manuscripts were not read consecutively or navigated according to titles, but were also likely to be dipped into and explored in more open-ended ways. This paper uses the example of the Old English Exeter Book to explore how we can use PCA to create digital network graph editions of medieval manuscripts. These interactive networks have the potential to facilitate an open- ended online navigation of the varied content of medieval miscellany manuscripts, inviting readers to read the texts in different orders and constellations. Such non-directional navigation can serve to reveal the many networks of inter-reference connecting the texts in this collection and can draw readers’ attention to works that lie outside the canon of famous Old English texts contained in the codex.


Gustavo Fernández Riva

Monday 17 September, 4:20 PM

University of Buenos Aires [AR]

Naming and Renaming Texts. Distant Reading of Middle High German Rubrics in Miscellany Manuscripts


One way of conceiving a literary canon is as a list of works and authors; a subsection of all the texts composed within a literary system. In this sense, a canon presupposes a catalogue of interrelated works. This presentation focuses on such a catalogue and confronts the modern and the medieval ways of conceiving it. The literary system in question is that of the Middle High German short texts composed in rhyming couplets (Beispiele, Mären, Minnereden, etc.). These texts usually appear in miscellany manuscripts where they include rubrics that delimit, identify and characterize them. Although sometimes copied along with the core text, rubrics were considered even by the medieval scribes as part of the paratext, which would often be altered during the copying process. These rubrics also tend to be more complex than modern titles and sometimes consist of whole sentences or even some verses. This presentation analyses the rubrics in a considerable number of manuscripts using methods of distant reading. It is inspired by Franco Moretti’s1 study of titles of English novels, but adapted to the particularities of the medieval materials. The corpus is composed of about 1500 rubrics, which are analysed to gain insights into their features and properties but also into the evolution of the German literary system during the High Middle Ages. The comparison of these rubrics with the modern canonical names of the same works can also lead to an interesting reassessment of the literary canon.


Session 3: Authorship


Diane Watt & Mary Dockray-Miller

Tuesday 18 September, 9:15 AM

University of Surrey [UK] & Lesley University [US]

Women’s Patronage, Authorship, and Collaboration in the Medieval Literary Canon


In this joint presentation, we explore how a consideration of “women’s texts” and collaborative authorship can enable a widening of the medieval literary canon, before going on to consider the contribution digital humanities tools might make to this process.

The presentation begins with an overview of the work undertaken by the Women’s Literary Culture and the Medieval Canon international network [https://www.surrey.ac.uk/womens-literary-culture-medieval-canon], which was funded by the Leverhulme Trust from 2015-2017 and led by DW, and to which MD-M contributed. The phrases “women’s literary culture” and “women’s texts” encompass many possible forms of women’s relationships with textuality — women as subjects, authors, audience, patrons, scribes, editors, and archivists of various written expressions. They thus make visible the diverse contributions of women to what we now think of as the literary canon.

DW will briefly discuss her work on collaborative authorship in relation to early medieval material, including Bede’s Ecclesiastical History. MD-M will then speak in more detail about The Anonymous Vita Ædwardi, composed in the mid- to late-1060s for Edward’s Queen Edith, which exemplifies many of the relationships between women and texts outlined above. After languishing in relative obscurity for many years, castigated as both bad history and bad hagiography, the VER has recently received substantial critical attention for its historical, literary, and religious contributions to our understanding of late Anglo-Saxon culture within broader European culture. It acts today as our case study in gendered production and reception of collaboratively-authored texts.

This paper uses the techniques of Lexomics (see https://wheatoncollege.edu/academics/special-projects-initiatives/lexomics/ for a description of the digital tool) to demonstrate that the VER is actually a composite, collaborative text, with different authors responsible for specific parts of the prosimetrum. Most crucially, this analysis informs our understanding of the gendered relationship between patron and author(s), between the (dowager) Queen and the monks and possibly nuns who worked for her. Lexomics finding show Queen Edith as a substantially  engaged patron, the active guiding intellect behind text, whose voice is occasionally discernable in certain passages.


Jeroen De Gussem & Jeroen Deploige

Tuesday 18 September, 10:00 AM

Ghent University [BE]

Between Manuscript and Edition: Scribal (re)writing in Hildegard of Bingen’s Liber Divinorum Operum


All too often in a current-day mindset, one conveniently thinks of a literary canon as consisting of singular and “finished” texts that are ascribed to named authors. In such a definition of canonicity, authenticity and integrity of the text are key. After all, the text is seen in function of exemplifying —preferably intact— the talent or significance of a canonical author. The merit of the New Philology, which gained momentum in the 1990s, was that it problematised many of the anachronistic notions attached to such a definition of canonicity for medieval times. In this wider framework of the canonicity debate, the current case study finds its place. We focus on a specific manuscript, namely Ghent MS 241 of the Liber divinorum operum (1170-4) of Hildegard of Bingen (1098–1179). Ghent MS 241 is well known due to the many corrections which Hildegard’s collaborators brought on to the parchment: words are scored out, erased or replaced in order to further refine Hildegard’s Latin. Therefore, the manuscript, by virtue of its being a rare, physical witness of a medieval redaction process, plays a crucial role in our better understanding of the collaborative process by which Hildegard’s texts were composed, and of the role and recognizibility of her secretaries. In this paper, we will make use of both stylometric methods and palaeographical arguments to obtain a better insight into the collaborative process behind Hildegard’s canonised oeuvre.


Justin Stover

Tuesday 18 September, 11:00 AM

University of Edinburgh [UK]

Canonical Texts, Medieval Editors: A Computational Approach?


Sometimes a literary work survives; sometimes it does not. Between these two, there is a medial case, where the original text as written by the author does not survive, but a deliberately redacted or edited version made by a later scholar or compiler does. Sometimes we know that this is the case because the editor has told us so. Eugenius of Toledo, for example, tells us for example that he prepared a revised version of the works of Dracontius for King Chindaswinth (d. 653). Similarly, Paul the Deacon tells us that he prepared new and augmented version of Eutropius for the duchess Adelperga (d. after 787). Those are the simple cases. But in other cases, recension can only be suspected but not proven. In this paper, I will look at a couple of case-studies of possibly edited and revised corpora, exploring whether computational methods can help us begin to distinguish the hand of the (medieval or late-antique) editor from the text of the original author. This is related to, but not identical with, the problem of multiauthored corpora (see Stover and Kestemont 2016 on the Historia Augusta) and that of collaborative authorship (see the pioneering study Kestemont, Moens, and Deploige 2015 on Hildegard of Bingen). But given the fact that almost all of our ancient ‘canonical’ texts survive only through medieval copies, determining as far as possible the role of the medieval editor may well prove fruitful in accurately assessing the textual heritage which has come down to us.


Godfried Croenen

Tuesday 18 September, 11:35 AM

University of Liverpool [UK]

Guillebert de Mets: Scribe, Editor or Author? Digital Tools and the Analysis of the Working Methods of a Late Medieval Flemish Scribe.


The 15th-century Flemish scribe Guillebert de Mets has long been identified as the copyist of three manuscripts with French literary text, two of which were produced for the Burgundian ducal library, including one volume that contains a Description de la ville de Paris which Guillebert himself seem to have written, partly based on existing texts and partly on his own knowledge of the city. Research in recent years has added a substantial number of new manuscripts to Guillebert’s corpus and has also suggested that in many of these volumes Guillebert did not just act as scribe but often intervened in the texts he copied. Apart from the Description de la ville de Paris there are at least two more autograph compilations of which the authorship may also be attributable to Guillebert (the Tresor des histoires and the Livre de Charlemaine). Furthermore, the copies of Christine de Pizan’s Livre des fais d’armes et de chevalerie, Froissart’s Chronicles, the Chronique de Flandre, and the Second Lucidaire, in Guillebert’s hand all represent scribal redactions that introduce textual innovations which may also have been authored by Guillebert.

This paper will present some ways in which DH techniques can be used to identify Guillebert’s intervention in this corpus in order to assess and quantify Guillebert’s working methods and compare them across manuscripts and across texts.


Session 4: Corpus Bias


Karina van Dalen-Oskam

Tuesday 18 September, 2:00 PM

University of Amsterdam [NL]

How to Avoid Corpus Bias: Some Thoughts on Building Digital Research Corpora of Medieval Texts


Many scholars, archives, and libraries are actively digitizing texts. Computational humanists may use these, for example, for authorship attribution or genre analysis by means of stylometric analysis. It has become relatively easy to gather a large digital corpus. However, several issues make trusting on already existing digital texts risky, such as the age and type of the editions selected for digitization. For some research questions, existing digital texts can safely be used, but for others, scholars may need to become aware of an underlying bias. The digital bias builds on a millennia-long analog bias. In my contribution I’d like to point out the dangers and suggest some possible solutions. By way of example I will refer to my stylometric analysis of samples from a Middle Dutch work by Jacob van Maerlant,  Scolastica (1271), which is available in fifteen manuscripts. The aim of my stylometric approach was to test the assumption if it could help us discover the limits of scribal freedom in certain times and areas. In this case, the issue of the influence of the type of edition used on the results of the measurements will receive special attention.


Eveline Leclercq

Tuesday 18 September, 2:45 PM

Université de Strasbourg [FR]

Stylometry in Charters: A Practical Test on the Case of Cambrai


Stylometry as a method of authorship attribution and verification has taken flight in the past few years. Subsequently, it is slowly finding its way into other domains of historical research, such as medieval diplomatics. The question of authorship in these texts is a complicated and interesting one, which has received attention for over 150 years. The introduction of the computer has been slow but inevitable, for palaeography and dictamen-research alike. This last field specifically has benefitted from the use of the computer to determine if a charter has been written in a chancery or in the scriptorium of a beneficiary, and if possible to determine a personal style of its writer. The method “De Paermentier” is its most recent development. As such, stylometry might open up the way to an additional or refined method of dictamen research.

In this talk a practical test on the twelfth-centrury charters of Cambrai will be discussed. The chancery of this diocese was particularly well organised. In the current state of the art some authors have been attributed via a study of the dictamen until 1130 (Van Mingroot), completed by my own research for the remainder of the century. Especially the chancellor Werimbold III seemed to have had a strong influence in the chancery and a particular style in the charters attributed to him.

The corpus of the test consists exclusively of charters, as no other texts of these attributed writers are currently known. As a research object charters present their own set of problems and challenges for stylometry, especially in terms of the length of the text, the formulaic character of the charters and the lack of other attributed texts to compare to. The goal of the test is to see how far stylometry can be of use in the dictamen research. Could the method help to determine whether a charter was written in the chancery or no? Could it confirm suspicions of a writer as author of a certain charter? This test is conducted by a diplomatist first. As such, the point of view of a layman of the method is present, allowing for a check of how user-friendly the method and the program Stylo with R is.


Reima Välimäki

Tuesday 18 September, 3:45 PM

University of Turku [FI]

Authorship Attribution and the Late Medieval Literature: Challenges and Solutions of a Heterogeneous Corpus


In the last few years, computational authorship attribution has been proven to be a powerful tool in the analysis of features particular to medieval literature, such as shared or collaborative authorship. The particular challenge of the late middle ages is the abundance of sources but the lack of machine-readable  corpora with consistent editorial standards. While orthographical variation is possible to solve with pre- processing, e.g. with lemmatization and sequence tagging (Kestemont & De Gussem 2017), it is evident that  from the late medieval literature it is extremely challenging to build a representative corpus for AA tasks, at least compared to periods covered by Patrologia Latina, CSEL or Perseus.

In the proposed paper an additional approach is presented. Our ongoing project PROPREAU uses Support Vector Machine and Convolutional Neural Network in AA tasks, and our preliminary results with large Ciceronian and Augustinian corpora demonstrate that the classifiers give accurate results based on word uni- and bigrams and character 5-grams (classifications run both with all but 1000 most common words masked, and without masking). In February 2018 we ran a first test (SVM) with a 700 000 word corpus of anti-heretical treatises. The goal with the anti-heretical corpus is to verify the inquisitor Petrus Zwicker’s authorship of a 1390s text Refutatio errorum, an attribution I proposed based on qualitative analysis and manuscript evidence in my dissertation (Välimäki 2016). The RE exists in four different redactions, only one of which is (poorly) edited. The texts in the corpus have different orthographical standards, from classical editions to transcripts from medieval manuscripts. However, after a very crude normalization of orthography, the three test cases of RE (1 edited, 2 manuscript transcripts) performed in a similar and expected way. For the proposed paper, we’ll run additional tests both with SVM and CNN in order to test to which extent our classifiers allow ‘noise’ caused by the variation in orthography. The potential results would provide a work-economical way to widen the canon in authorship attribution.


Maciej Eder

Tuesday 18 September, 4:20 PM

Pedagogical University of Kraków [PL]

What is the Difference Between Prose and Poetry, Really? A Computer-Assisted Analysis of Latin Style


The presentation will focus on applying stylometric methodology to the study of Latin style. Usually associated with authorship attribution, stylometry can be used to trace other stylistic signals beyond authorial idiosyncrasies. These include gender, genre, chronology, and so forth. This study will focus on automatic classification of poetry and prose. The question to be undertaken is as follows: wich words are more likely to appear in poetic works, as opposed to prose? The results are rather counter-intuitive, since the strongest discriminators are certain prepositions.