=Paper=
{{Paper
|id=Vol-2717/paper02
|storemode=property
|title=Twin Talk: Bukvik+LitTerra+Colabo.Space - An Example of the DH Collaboration Across Disciplines, Languages, and Style
|pdfUrl=https://ceur-ws.org/Vol-2717/paper02.pdf
|volume=Vol-2717
|authors=Sasha Mile Rudan,Eugenia Kelbert,Lazar Kovacevic,Sinisa Rudan,Matthew Reynolds
|dblpUrl=https://dblp.org/rec/conf/dhn/RudanKKRR20
}}
==Twin Talk: Bukvik+LitTerra+Colabo.Space - An Example of the DH Collaboration Across Disciplines, Languages, and Style==
https://ceur-ws.org/Vol-2717/paper02.pdf
Twin Talk: Bukvik+LitTerra+Colabo.Space - An
Example of DH Collaboration Across Disciplines,
Languages, and Style
Sasha Mile Rudan, Eugenia Kelbert, Lazar Kovacevic, Sinisa Rudan and Matthew
Reynolds
sasharu@uio.no, ekelbert@hse.ru, lakinekaki@gmail.com,
sinisa.rudan@gmail.com, matthew.reynolds@ell.ox.ac.uk
Abstract. This paper focuses on a long-term collaboration between the two poles
of the DH-dipole; the D-pole: a CSCW (Computer-supported cooperative work)
- Computer Science scholar, Sasha Rudan, and the H-pole: a Comparative Liter-
ature scholar, Eugenia Kelbert. It involves work with a larger team as well, in-
cluding this paper’s co-authors, among others. Our research started through a
mutual interest in the digital analysis of stylistic features of fictional texts, mostly
novels. Eventually, it developed towards designing a new ecosystem for collab-
orative research in the textual and stylometric DH domains. From a practical re-
search question in stylometry in translingualism, we evolved to developing new
tools, a DH infrastructure, later a DH research collaboration ecosystem and meta-
research questions addressing challenges of DH collaboration and its practical
solutions. Here, we discuss the oppositions between the different disciplines in-
volved, the challenges we faced on the road, and how we tried to avoid them by
getting a level higher in our collaboration.
Keywords. DH collaboration, methodologies, workflows, stylometry, research
challenges
1 Introduction
The primary participants in this collaboration already had experience working out-
side their field, and were prepared for the peculiarities of an interdisciplinary collabo-
ration to some extent. Eugenia was working on a project comparing literary texts sty-
listically across languages and had reached the conclusion that a DH perspective would
be complementary to the close analysis she otherwise based her argument on. She there-
fore took a course on computational linguistics (in Python) in the first year of her PhD
program at Yale and later sought out another collaborator in Computer Science, Wil-
liam Teahan at Bangor University, with whom she worked on a conference paper in
2011, a few months before her and Sasha’s collaboration started. She had also had some
exposure to authorship attribution methods and probability theory, and was familiar not
so much with contemporary stylometry as with the pioneering work by Shannon and
the great Russian mathematician Kolmogorov. Sasha, in his turn, had always had an
Copyright © 2020 for this paper by its authors. Use permitted under
Creative Commons License Attribution 4.0 International (CC BY 4.0)"
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 15/143
interest in the Humanities, publishing poetry and performing slam poetry, as well as
being active in literary campaigns in Serbia. At the same time, he was an active member
of the DH community and worked, pre-DH, on various projects ranging from interac-
tive text and media to visualizing novels and poems in an appealing way, juxtaposing
writers, texts and facts. His dream was to get access to the archives of the Serbian Nobel
Prize winner, Ivo Andrić (Иво Андрић) and understand him through the help of DH
analysis. Similarly, he wanted to tame the wild metaphors of the Serbian neo-symbolist
and surrealist, Branko Miljković (Бранко Миљковић).
1.1 Background Reflections
This previous history is key for the positive results of this collaboration, for two
reasons. First of all, few collaborations start from scratch, and participants invariably
bring in their agendas and experience. Establishing some vocabulary in common, and
a mutual appreciation of the other field’s methods, is perhaps the one prerequisite for
any successful DH work in the long run. Such an appreciation can never be taken for
granted in a DH collaboration, where both sides, however genuinely intrigued by the
possibilities of working together, often have to overcome misunderstandings: the liter-
ary scholar may be skeptical about the extent to which scientific method can be usefully
applied at the level of literary analysis, or feel threatened by such methods, and the
computer scientist is liable to consider literary analysis to lack the formalism and the
empirical grounding of a scientific approach.
Luckily, both collaborators had a degree of understanding of the other discipline’s lan-
guage and approaches, perhaps more than many starting off in DH. For example, Eu-
genia’s knowledge of programming—albeit minimal—was invaluable. Unable to con-
tribute to the code herself, she could understand it, when explained, and discuss it in
some detail, which made a major difference to the project’s progress. In this sense, we
cannot stress enough the advantages of time invested in even the most basic acquaint-
ance with the other collaborator’s field of expertise, even if it appears meaningless (in
Eugenia’s case, for example, she may have not taken the course in Python thinking it
would not be enough to code what she wanted on her own, and therefore not a good
investment of her time; nevertheless, it was).
2 The birth of the project, or a DH research methodology
In terms of what each researcher brought to the project, it is important to note that such
‘dipole’ collaborations may be of three primary kinds. One [1] is where one of the par-
ties has a project and enlists the other fully into it (for example, in a situation where the
‘D’ researcher hires literary experts to create training sets, or an ‘H’ researcher engages
a programmer to create a tool for them). Its limitation lies in the fact that the enlisted
party has no inherent motivation, may or may not contribute original thinking to the
project, usually needs to be paid for their contribution and clearly there is no
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 16/143
interdisciplinary innovation involved. Another [2] is where one of the parties has a fi-
nalized corpus (‘H’) or tool (‘D’) the other decides to use, as for was the case in Sasha’s
collaboration with Biljana Dojcinovic (corpora of feminist literature) and Eugenia’s
collaboration with William Teahan (tools for textual compression), respectively. The
limitation here is that the preexisting corpora/tool becomes a Procrustean bed that limits
what the researcher can achieve significantly, and forces them to adapt the
knowledge/method to what is available. This is, indeed, the issue with most stylometric
projects relying on pre-existing tools, however flexible.
Finally, perhaps the most promising but also the most complex scenario is what the
present collaboration ended up to be, namely two or more researchers who each has a
stake in the mutual project and is therefore internally motivated.
2.1 Dimensions of Freedom (or Interests)
Initially, our work started with a range of different dimensions, or rather interests that
were at the same time challenging, and opened new opportunities and improved both
our individual and collaborative research processes. Below, we present some of these
dimensions and the researchers’ “place” along them.
1) Tools: the D-pole: to understand how the DH stylistic distant-reading process may
be improved to provide better and more targeted/useful results and new insights, and
the H-pole: to use available DH tools to get insights into the style of bilingual writers
compared to native-speaker writers.
2) Languages of interest: Eugenia Kelbert’s main languages of interest were English,
Russian, French, and German, and Sasha Rudan’s languages of interest were English,
Serbian (and other former Yugoslavian languages), and Russian where the former Yu-
goslavian languages were under-supported languages (in the NLP+stylometry scope).
Both of them had a general interest in languages well-supported in the NLP+stylometry
domain.
3) Collaboration scale: the D-pole was customized to a higher-scale real-time collab-
oration with various stakeholders with a high interest in inter-disciplinary collaboration.
On the other hand, the H-pole tends to support lower-scale collaborations, and less real-
time collaborative work, and is usually less used to inter-disciplinary collaboration.
4) Close reading: in our collaboration, the H-scholar’s expertise lies in the close read-
ing of bilingual writers (among others), while the D-scholar’s competence comes from
his undergraduate education, as well as from being a writer of poetry and short stories.
5) Distant reading: in our collaboration, the primary D-scholar’s expertise is in NLP
and data analysis, and system modelling especially for collaboration, and the H-
scholar’s competence lies in introductory programming courses and stronger mathe-
matical background.
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 17/143
6) Research workflow tools: the D-scholar’s research interest lies in optimizing teams’
face-to-virtual workflows and enhancing knowledge federation. On the other hand, the
H-scholar had a basic knowledge of the Python ecosystem and higher than average
computer literacy, but no exposure to elaborative digital research workflows.
7) Research methodologies: The D-scholar’s are statistical quantitative and qualitative
methodologies, comparative evaluation and participatory design and, partially, action
research. The H-scholar’s preferred research methodologies fall within the fields of
qualitative analysis and archival research.
8) Infrastructure evolution: finally, the D-scholar aimed to design, research and opti-
mize the workflow, while the interest of the H-scholar was in the availability, con-
sistency, and reliability of the research workflow.
Challenges and their resolutions
True inter-disciplinary collaboration of two researchers with equal stakes in the project
comes with several benefits, but also significant challenges. Coming from different dis-
ciplines, two (or more) researchers bring rich innovation dimension to mutual work and
much stronger overall expertise and likeliness of correct and successful project finali-
zation. On the other hand, given distinct, and usually not strongly overlapping, research
agendas, they are liable to have wildly divergent investment in the project, leading to
unexpected developments and inevitable compromise.
The challenges of collaboration in our case lay mainly in two categories; [1] the “col-
laboration” category relating to different practices and previous experience in collab-
oration and the “research-interests” category relating to different research interests in
the project and overall collaboration – for example; Sasha’s strong research interest was
in the continuous evolvement of the DH tools and methodologies through participatory
design and action research. While this is an interest Eugenia eventually came to share,
her primary interest is in using tools and conducting stylometric research. This means
that she was especially invested in workflow stability, which opposed Sasha’s research
interest. This bipolarity of the our skills and research interests, which were largely com-
plementary, introduced inevitable tension during the project’s critical milestones. How-
ever, we had respect for the methodologies each of us brought to the project, were keen
to expand the range of languages covered, and wanted to improve the tool to be both
powerful and, crucially for both, flexible to evolve over the long term as competing
technologies evolved. In other words, despite tension coming from non-aligned inter-
ests and collaboration practices, we had mutual goals in terms of the resulting set of
tools and methodologies, which largely helped with the ongoing success of the project
and the collaboration itself.
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 18/143
In terms of the DH tools and infrastructure, Sasha’s interest and that of another col-
laborator he brought into the project, Lazar Kovacevic, was in language-agnostic (when
possible, or multilingual when there were language specific requirements) solutions,
scalable to work with a high and reproducible volume of research. For example, our
LitTerra infrastructure deals with the whole Gutenberg corpus counting over 45’000
texts with various intertextual and intratextual analyses. On the other hand, what Eu-
genia wanted was a set of tools supporting her project, since existing tools either did
not satisfy her needs, or were too hard to unite into a consistent workflow and/or une-
qual to tackling large corpora in several languages systematically. Sasha’s answer to
that challenge was not to deal with and maintain every single tool in a conceptually
consistent research workflow, but rather to propose a “one ring to rule them all.” In this
way, he could avoid unhealthy maintenance of separate tools, but also provide a repro-
ducible environment for parallel experiments against multiple corpora. Sasha’s interest
as a researcher was in workflows and systems facilitating collaboration and knowledge
federation, so that the tool itself had for him an added value as a case study of such a
system. The result was that, on the level of the research workflow, the project took on
a life of its own as a workflow-based system rather than a simple toolkit, but with the
capabilities required by the initial project. In other words, a great deal of flexibility, as
well as patience, was required of both parties to accommodate each other’s research
needs. During this process, each collaborator became a contributor to the theoretical
and methodological aspects of the other’s research pursuit.
An interesting disbalance and semantical inequality of the D and H disciplines lay in
the fact that our D-related work resulted in a rather generic tool that could be used by
H-scholars without relying on a D-researcher. However, the H-scholars’ results were
not “reusable” for D-researchers. For example, for stylistic analysis of Former Yugo-
slavian authors, there was not much help (apart from certain methodological aspects)
from material associated with the writers Eugenia was interested in. On the other hand,
collaboration on designing and conducting stylometric research at generic and meta
levels helped Sasha and the other D-contributors (Sinisha Rudan and Lazar Kovacevic)
to transfer practical and tacit knowledge and conduct research on Former Yugoslavian
authors (Rudan et al, 2019-Torun) as well as ongoing research with Matthew Reynolds
on his Prismatic Jane Eyre project.
Differences in working styles when it came to collaboration proved to be another ma-
jor, and unexpected, challenge. In our case, this seemingly innocuous difference, which
one would have expected to be a lot less of an apple of discord than, say, methodolog-
ical differences, became one of the hardest issues to overcome in our work together.
We were both open-minded and willing to learn and to accommodate the other disci-
pline’s methods and approaches. We were, however, a lot less willing to change our
day-to-day workflows. A humanities scholar tends to do most of their work at their own
pace, and have entrenched ways of working, and can be resistant to the practices of
structured collaboration, brainstorming, regular meetings, documentation, etc. These
are only partially personal differences of style; they are largely down to divergent cul-
tures of research within the different disciplines. Even in writing this paper, after seven
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 19/143
years of working together, we experienced tension over Sasha expecting Eugenia to
write her part of the paper in bullet points, and Eugenia insisting to formulate her
thoughts writing in full sentences from scratch. A compromise we arrived at was that
she wrote her parts first but then highlighted the internal structure for Sasha to use the
highlights as ‘bullet points’ of sorts to integrate into the overall argument. Even minor
factors such as using different textual editors, or Sasha’s insistence on Markdown for-
mat and GitHub for documentation as a form that allowed for easier integration with
the coding environment, added to the cognitive load of adapting not to one tool (the one
we were developing together), but to several different interfaces and ways of working.
On the larger scale of project development, it was a challenge for Eugenia to write user
documentation for the program we created as the form was alien to her, and once she
learned how to do one task or another, she did not feel the need for a separate record.
This, in turn, made Sasha’s work harder, since omissions in documentation meant he
had to repeatedly not only re-teach his collaborator after a break in the project, but also
often re-teach himself, as he would also forget the parameters in running a given version
of the tool. For Eugenia, on the other hand, it was a source of frustration that the pro-
cedure of running the program and the interface—not intuitive for an H-scholar—had
to be relearned for each version as the system improved or needed to be restructured.
Finally, and perhaps crucially, both collaborators had very different tacit assumptions
about the development process (the infrastructure-evolution dimension). For Eugenia,
the very concept of the coding workflow took time to absorb. On the other hand, she
had to deal with discomfort when she recognized with time that the tool, once func-
tional, was never set in stone but kept developing, both as it grew and improved, and
also as the external libraries and tools it relied on also changed, triggering the need for
several instances of top-to-bottom refactoring. For a D-scholar, this was the normal—
indeed expected—price of a system’s evolution and progress. From an H-scholar’s per-
spective, however, it came as a surprise that our work depended on external—and
evolving—systems and that a function that already worked seamlessly could easily re-
quire an upgrade five months later.
As these brief profiles demonstrate, much in what we had to bring to the project shares
core attributes with those of an average literary scholar who is not a novice in digital
humanities (i.e. who has a traditionally humanities research agenda and experience
working with stylometrical tools, perhaps some instruction in the area) and those of an
average computer scientist interested in the humanities (personal interest and back-
ground but little formal training). Perhaps more unusual, in our case, was the focus on
stylometrical tasks across languages and, for the D-scholar, the research interest in sys-
tem architecture, which he brought to the project. On the whole, our experience, and
that of finding a mutual research language and procedure, illuminates both the core
challenges and the potential of close inter-disciplinary collaboration as a solution to
existing challenges in Digital Humanities as a field.
In this paper, we discuss our findings and the co-evolution leading us toward these
findings. As we were introducing additional collaborators to our research team, adding
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 20/143
additional projects and participating in external grants, we understood the importance
of a proper collaboration strategy and even more, of developing a collaboration ecosys-
tem.
3 Research Questions Trajectory
3.1 How a new tool has born - Bukvik (research)
While brainstorming potential DH tools for our first mutual project, the D-scholar had
the initiative to establish an internal DH infrastructure. His primary reasons were to
ensure uniform analysis of texts and user experience, which would be agnostic of the
tools used, provide continuous research workflow and work as a reproducible research
environment for comparative stylometric analysis. This is how the Bukvik infrastruc-
ture was born and presented at the SCLA Conference in Zagreb, Croatia, 2012. From
that moment, we embraced Bukvik as our internal infrastructure that helped us to in-
corporate some aspect of our collaboration in practice, and evaluate future needs. It
became the playground for our future tools, a prototype of our understanding of what a
DH-framework should be.
3.2 Initial Research Questions
Two main research questions we started our collaboration journey with were in the
domain of: 1) translingual stylometry and 2) flexible corpora analysis infrastructures.
The emerging field of stylometry is still far from being able to fully grow out of meth-
ods it inherited from authorship attribution and distant reading, which shaped it with
their own aims and priorities. This takes both time and a different generation of com-
putational tools that would focus on stylistic features for their own sake rather than for
the sake of clustering and identification. Our goal with our central project, Bukvik, has
been to fill this gap, first, in terms of relying on a custom-made tool with an initial focus
on cross-lingual textual comparison. Secondly, it extends the principle of multidimen-
sional analysis, identified by Jockers, to a potential stylistic profile: the sum total of
quantifiable stylistic features for each text or body of texts that, together, constitute a
multidimensional model of the given writer’s style with reference to a balanced corpus
of fiction in the given language. It supports, further, a novel method of textual analysis
based on the visualization of individual words in a literary text as a network. This work
relies on interdisciplinary collaboration to enable the development of original tools. The
tool’s modular structure ensures its relevance beyond the features that we are capable
of tracking today and extends the relevance of the stylistic profile model beyond the
specifics of the current project (cf. Jockers, 2013; Hoover, 2014).
The translingual stylometry aspect of the collaboration seeks practical solutions to
quantifying those of the possible stylistic markers that current language processing
tools are already capable of tracking and contextualizing this work within the
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 21/143
theoretical framework of comparative literature. Is style separate from the linguistic
norms of a given language? Is content? Eugenia’s dissertation on bilingual writers (Yale
University, 2015) strongly suggests that it is not, or at any rate not fully. The featured
bilingual authors’ corpora were used for the initial digital comparison in the collabora-
tion. We aim to extend it to other corpora, notably translated texts with their originals
and corpora in the NLP-underdeveloped languages (like Former Yugoslavian lan-
guages as part of South Slavic languages, although the scene dramatically changed in
this aspect in the past few years with dedicated research like The CLARIN Knowledge
Centre for South Slavic languages (CLASSLA) and more universal tools like Adobe’s
Cube NLP, Universal Dependencies framework and treebanks).
3.3 Secondary Research Questions: Methodology
Our study draws on an original methodology that aims to make a real contribution to
computational stylistics or stylometry. This approach complements Moretti’s more
popular method of distant reading. The system is conceived as an aid for automatic
zooming: unlike the “distant reading” approach where statistics replaces reading and
helps process large corpora, we see Bukvik as a non-automatic augmenting framework
that will ultimately aid and direct close reading. “Close reading at a distance” is one
way to describe the idea behind the methodology. This goes together with the research
approach we refer to as Qualitatively Augmented Quantitative Analysis. The goal is for
the two approaches to interact and inform each other: qualitative data will shape and
instruct the quantitative component in analysis leading to more relevant results. Having
this flexibility, Bukvik allows scholars a variety of tasks, such as the analysis and dif-
ferential parallel comparison of translations of the same book, of an original with a
translation, of corpora of two writers’ work, as well as comparing texts within a lan-
guage or across languages, and comparing variations from respective corpora in each
language.
4 The birth of further tools (and eventually, an infrastructure)
Out of this multi-dimensionality and polarity, the understanding was emerging that we
had to essentially design and structure our own collaboration in order to fulfill the
requirements and expectations of each pole of the DH-dipole. We realized, further, that
our collaboration exemplified many of the general core challenges of DH collaboration
more generally, and the need to provide a more articulated and rigid framework for DH
practice.
4.1 How a further tool was integrated - LitTerra
Soon we realized that for successful DH research we needed a “space” to map our re-
search findings and provide them to other scholars. A key element of this component
would be the visualization of findings that would facilitate both cross-references to the
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 22/143
texts analyzed and data analysis. That is how we integrated another infrastructure in
our research workflow; LitTerra - an infrastructure for augmentation of texts with var-
ious digital content, founded at a similar time by the D-scholars in the project, Sasha
Rudan and Lazar Kovacevic (Rudan et al, 2013; Rudan et al, 2019).
The most important consequence of such an integration lay in the understanding that
there was a much wider audience for Bukvik than we were aware at the beginning. In
the language of business models, we discovered additional user personae. We have also
isolated research analysis (Bukvik) from research presentation (LitTerra) making it pos-
sible to extend to other texts and corpora. This allowed finalized research to co-exist
with related texts and other relevant research and be made available for the end user to
explore holistically. Eventually, it helped us in terms of the availability and shareability
of Bukvik results.
Ongoing work with Matthew Reynolds on the Prismatic Jane Eyre project
(prismaticjaneeyre.org) enforces the standardization and scalability of the Bukvik+Lit-
Terra systems. It additionally pushes the multi-lingual and collaboration aspects as the
Prismatic Jane Eyre project involves dozens of translations of “Jane Eyre” and a large
community of researchers.
4.2 How new tools were integrated - Collaboration (dialogue and knowledge
federation)
After a few years of practice supported with the Bukvik and partially LitTerra infra-
structures, we felt the need to formalize our work and methodologies, and that is how
we approached the discipline of CSCW (Computer-supported cooperative work) for
answers. The first concept from CSCW we introduced in our practice were boundary
objects (BO), as “spaces” of common understanding, that had a reasonably clear mean-
ing for most of the stakeholders (Star & Griesemer, 1989; Star, 2010). In an inter-dis-
ciplinary collaboration such as this one, building a dialogue space is indispensable;
without such a space, however limited, no collaboration could continue. Hence, we felt,
the importance of what we have referred to above as the meta-discussion of a collabo-
rative process one is part of, and consequently, of a theoretical basis for this discussion.
To technically integrate boundary objects into our research workflow, we came to the
Colabo.Space ecosystem as a part of Sasha’s PhD dissertation and Sinisha Rudan’s
research and development, supported with Dino Karabeg’s Knowledge Federation ini-
tiative. Colabo.Space provided the knowledge federation component of the DH-
ecosystem which could natively support the concept of boundary objects together with
fuzzy-knowledge and multi-truth. This helped our collaboration in the incremental de-
velopment of the initial (fuzzy) knowledge starting from the commonly-understood con-
cepts (expressed with the boundary objects).
Additionally, integrating the Colabo.Space ecosystem was intrinsically feasible as its
main principle is puzzlebility (i.e. modularity, fig. CF-example). Thus, we could fed-
erate Bukvik and LitTerra with an instance of the Colabo.Space ecosystem adjusted to
our requirements.
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 23/143
Figure CF-example: An example of the Colabo.Space ecosystem in use
4.3 Search for sustainable research evolution (ColaboDialogue)
However, we still lacked a healthy mechanism for dialogical collaboration which would
organically evolve into a next round of research questions, actions and solutions.
To enable incremental evolution when it comes to the capacity for dialogical interven-
tion through knowledge changes and actions, we needed to introduce a reflective and
proactive mechanisms of dialogue and knowledge evolution—to balance the unbal-
anced. Unfortunately, the majority of technologies and tools used to support dialogue
(including IBIS systems and Wikimedia) lack the possibility of automatic and continu-
ous evaluation and evolution of dialogical outcomes—interpreting dialogical results
and intervening either in the knowledge space or in the real-world. In other words, the
sustainability of the dialogue-knowledge-action loop was broken.
Therefore, we embraced ColaboDialogue—a concept that unites all the three spaces
(dimensions), i.e. dialogical, knowledge and action spaces, into a single continuum
where interactions across domains are natural, fluent and frictionless. In essence, the
main or rather the most solid and long-term dimension is the knowledge dimension,
which evolves continually —it represents the collective memory of our collaborative
research effort. The aim of each DH community is to evolve its collective memory.
That evolution can run solely across the knowledge dimension, but it can be supported
by expansions into other dimensions. These expansions (based on their nature, evolu-
tion and life-time) we call bubbles.
A dialogical bubble bubbles out as a need to discuss an issue in the knowledge space,
for example, the “insight D” in the knowledge space initiated the “bubble 1” in the
dialogical space (fig. colabo-dialogue). At the same time, the dialogical bubble is re-
flective (for example “supports” reflection) on the knowledge space (as can be seen on
the same fig. colabo-dialogue). One important feature of the multidimensionality of the
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 24/143
ColaboDialogue is that the dialogical bubble lives in a separate dimension and does
not pollute the knowledge space. At the same time, it is strongly coupled with the
knowledge space and can support, change and reflect the knowledge artifacts (Insight
D, Claim A, …). After a period of time, the dialogue in the bubble matures and it can
usually be considered as "resolved." Consequently, following the real-world and social
model of artifact lifetime, it "fades out". It is important to notice that it remains avail-
able to enable arguing a particular knowledge evolution (decision) and avoid
"knowledge-wars" (well known in the Wikipedia discourse).
On the other hand, a dialogue provokes (real-world) actions and creation of an action
bubble (i.e. Question 1 → Action 1). The whole process naturally continues through
interactions across domains—an action outcome can affect the knowledge space (Ac-
tion 3 → Fact 3) or the (original) dialogical bubble (Action 2.2 → Idea 2). In this way,
actions introduce changes into the system and provide new information that calls to be
processed and understood. The ultimate goal of the process is to go the whole way back
and evolve the original knowledge space.
Figure colabo-dialogue: A detailed view of the three dimensions of the ColaboDi-
alogue - an example of the dialogical and action bubble 1
As a result, dialogue does not "hang in the air," but reflects back and transforms
knowledge and potentially neutralizes the tension or open question (in the knowledge
space) that initiated the dialogue at the first place. We can say that dialogue provides
healing support for knowledge management.
With ColaboDialogue, we could safely perform “Close reading at a distance” and in-
tegrate the Qualitative Augmented Quantitative Analysis research approach into our
collaborative workflow.
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 25/143
4.4 Seeking a mutual language - ColaboFlow
Much of the existing research in Digital Humanities relies on either scholars of litera-
ture adapting their approach to existing scientific methods and tools, or computer sci-
ence scholars working on literary texts. In both cases, competence is necessarily one-
sided, and we have not yet come to a defined language that would allow the two com-
petences to be orchestrated, together, to address the same problem. Our collaboration
is, among other things, an experiment in establishing such a language. This brings the
last key player in our research workflow: ColaboFlow, founded by the D-pole (Sasha
Rudan and Sinisha Rudan). ColaboFlow is a visual language for brainstorming, design-
ing, visualizing, and, most importantly, executing research workflows, and finally ex-
ploring and visualizing their results. It is based on an extended subset of the BPMN
language. As a visual language, it became our language of collaboration, the Lingua
Franca of DH research collaboration. Fig. ColaboFlow shows an example of the
ColaboFlow used in the Prismatic Jane Eyre Project.
Figure ColaboFlow: An example of ColaboFlow used in the Prismatic Jane Eyre Pro-
ject
The DH holistic research workflow and ecosystem presented here helped us to practice
it in the open/real-world at workshops, in various projects and campaigns.
DH being a relatively young discipline (although hand-counted authorship analyses and
Markov chains were demonstrated for the first time in literary analysis before digital
computers were discovered), many DH scholars are H-scholar new-comers from an H-
discipline (literature, history, music, art, etc). With the (fig. DH-research-workflow),
we present a standard workflow of a DH-scholar. As one can see from this diagram,
such a research flow is not that different from a similar science research flow, and rea-
sonably different from regular humanities research (e.g. a close-reading research flow).
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 26/143
Figure DH-research-workflow DH research workflow
On the one hand, this means that a DH-scholar is often faced with unforeseen chal-
lenges. On the other hand, for the H-scholar new to DH, this field represents a new
world expanding their disciplinary horizons toward new visually exciting and interac-
tive forms of research. That being said, DH researchers may find DH research reward-
ing without it always being innovative or rigid in digital terms. In the case of a trans-
disciplinary team collaborating on a DH project, this difference will bring conflict in
the way the D- and H- parts of the community work, or even in their research interests.
At the (fig. DH-research-challenges), we present a set of common challenges H-scholar
may face when they enter the DH world.
Figure DH-research-challenges: DH research challenges
It was to provide a safer environment for conducting DH research and in order to ena-
ble the dialogue across different disciplines (sub-communities) of the DH community
(sometimes represented in the single DH team conducting particular research), we
have designed and implemented Bukvik and evolved it into a DH-framework as pre-
sented in this paper.
5 Conclusion
From practical research questions in the domain of 1) translingual stylometry and 2)
flexible corpora analysis infrastructures, we came to developing new tools, a DH infra-
structure, and eventually a DH research collaboration ecosystem and meta-research
questions addressing challenges of DH collaboration and its practical solutions and pro-
totypes.i
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 27/143
There are various deadlocks a DH team can face in its lifetime, and not all such teams
survive long-term due to incompatibility, losing energy or a lack of resources (financial
or otherwise). A team may also not necessarily be interested in developing as a DH-
dipole unit, i.e. in a search for answering new research questions using DH tools, and
focus on developing and maintaining the tools they initially introduced.
All these scenarios were possible in our case, but we continued toward a successful
collaboration with external partners and external grants supporting our work. Our solu-
tions lead to new research questions to answer and a better understanding of DH chal-
lenges and possible solutions.
As already hinted in this paper, we are heading toward a DH-framework as a set of tools
and methodologies that would ultimately help other DH researchers in their work, but
this remains a topic for another paper.
6 Bibliography
Pfeiffer, S. I. (1981). “The Problems Facing Multidisciplinary Teams: As Perceived by
Team Members.” Psychology in the Schools, 18 (3), 330-333.
Matthew Jockers, Macroanalysis, University of Illinois Press, 2013
David Hoover, ed. Digital Literary Studies, Routledge, 2014
Sasha Mile Rudan, Lazar Kovacevic, Eugenia Kelbert and Sinisa Rudan, (2019) “Lit-
Terra, by augmenting literature with meaningful connections, turns readers into explor-
ers and researchers”, ELO2019, Cork
Sasha Mile Rudan, Eugenia Kelbert, Lazar Kovacevic, Sinisa Rudan, Tamara Butigan,
Miroljub Stojanovic, (2013) "Project LitTerra: New Travel Through Augmented Digi-
tal Book (the next step after digitization)," NCD'13 Conference, Belgrade, Serbia, No-
vember 2013.
Star, S. L., & Griesemer, J. R. (1989). Institutional ecology,translations' and boundary
objects: Amateurs and professionals in Berkeley's Museum of Vertebrate Zoology,
1907-39. Social studies of science, 19(3), 387-420.
Leigh Star, S. (2010). This is not a boundary object: Reflections on the origin of a
concept. *Science, Technology, & Human Values*, *35*(5), 601-617.
Sinisha Rudan, Sasha Rudan, Lazar Kovacevic, Eugenia Kelbert (2019), “Poetry on the
Road: An Intercultural And Multidisciplinary IT-Augmented Dialogue on the Topic of
Refugees and Migrants”, Comparing e/migrations: Tradition – (Post)memory –
Translingualism, 2019, Toruń
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 28/143
Kelbert, Eugenia. "Acquiring a Second Language Literature: Patterns in Translingual
Writing from Modernism to the Moderns" (PhD thesis, Yale University, 2015).
i
The results of this collaboration and the DH infrastructures involved have been pre-
sented at several international venues, such as the selected examples below:
“Visualizing Mademoiselle O’s Emigration Trajectory: a Stylometric Approach to
Nabokov,” Stockholm University, 2 March 2018, Stockholm, Sweden
Two workshops at the Digital Humanities in the Nordic Countries conference, 15-17
March 2016, Oslo, Norway: “Bukvik, a DH Scholar’s Environment for Stylistic Anal-
ysis” and “Tools and methodologies of Collaborative and Scientifically Structured DH
Research.”
“Bukvik and Cross-Lingual Stylistic Comparison,” Max Planck Institute for Empirical
Aesthetics, 14 September 2016, Frankfurt-am-Main, Germany
“Use of Digital Humanities Techniques in the Context of (Self–)translation and Bilin-
gual Writers,” Encompassing Comparative Literature: Theory, Interpretation, Perspec-
tive, 24-26 October 2014, Belgrade, Serbia
“Visualizing Dynamics of Narrative in Fiction,” 15-18 June 2015, SCALE, Malta
“Bukvik: A Literary Scholars' Environment for Running Visualized, Social-Aug-
mented, Collaborative Research,” MLA, 8-11 January 2015, Vancouver, Canada
Twin Talks 2 and 3, 2020 Understanding and Facilitating Collaboration in Digital Humanities 29/143