Abstract
Classifying and categorizing the activities that comprise "digital humanities" has been a longstanding area of interest for many practitioners in this field, fueled by ongoing attempts to define digital humanities both within the academy and in the public sphere. The emergence of directories that cross traditional disciplinary boundaries has also spurred interest in categorization, with the practical goal of helping scholars identify, for instance, projects that take a similar technical approach, even if their subject matter is vastly different. This paper tracks the development of TaDiRAH, the Taxonomy of Digital Research Activities in the Humanities developed by representatives from DARIAH, the European cyberinfrastructure initiative, and DiRT, a digital humanities tool directory. TaDiRAH was created specifically to connect people with information on DiRT and in a DARIAH-DE bibliography, but with the goal of adoption by other directory-like sites. To ensure that TaDiRAH would be usable by other projects, the developers opened drafts for public feedback, a process which fundamentally altered the structure of the taxonomy and improved it in numerous ways. By actively seeking feedback from the digital humanities community and reviewing data about how the source taxonomies are actually used in order to inform term selection, the development of TaDiRAH provides a model that may benefit other taxonomy efforts.
1. Introduction
Classifying and categorizing the
activities that comprise "digital humanities" has been a
longstanding area of interest for many practitioners in
this field, fueled by ongoing attempts to define digital
humanities both within the academy and in the public
sphere. The emergence of directories that cross
traditional disciplinary boundaries has also spurred
interest in categorization, with the practical goal of
helping scholars identify, for instance, projects that
take a similar technical approach, even if their subject
matter is vastly different. This paper tracks the
development of TaDiRAH, the Taxonomy of Digital Research
Activities in the Humanities developed by
representatives from DARIAH, the European digital
infrastructure initiative, and DiRT, a digital
humanities tool directory. TaDiRAH was created with the
short term goal of enhancing discoverability of
resources in the DiRT directory and the DARIAH-DE
bibliography while also anticipating adoption by other
digital humanities directory-like sites. To ensure that
TaDiRAH would be usable by other projects, the
developers opened drafts for public feedback, a process
which fundamentally altered the structure of the
taxonomy and improved it in numerous ways. By actively
seeking feedback from the digital humanities community
and reviewing data about how the source taxonomies are
actually used in order to inform term selection, the
development of TaDiRAH provides a model that may benefit
other taxonomy efforts.
2. Motivating factors
2.1 DiRT
Since the inception of the DiRT Wiki
in 2008, the site has used an ad-hoc set of
categories. In its original form, each category
represented a wiki page where tools were listed. In
2010, a migration to the Drupal content management
system gave each tool its own profile page. However,
even under this new structure, the original
categories persisted as a way to organize the
individual profiles. Starting in late 2012, the DiRT
Directory undertook an assessment of its categories,
with the goal of identifying any gaps, phasing out
rarely-used terms, and adding new terms to better
reflect the scope and nature of the tools presented
in DiRT. Early investigation into how the categories
were being used produced striking results. Entire
classes of commonly-used digital humanities tools
were largely rendered invisible through the lack of
an obviously matching category. For instance,
optical character recognition (OCR) tools were
scattered between the categories of "data
conversion", "transcription", "data collection" and
"annotation". The DiRT categories clearly needed
revision; when an opportunity arose to collaborate
with DARIAH-DE on a taxonomy that could be shared
across multiple sites with different purposes, doing
so was clearly preferable to undertaking another
isolated effort.
2.2 DARIAH-DE "Doing Digital Humanities" bibliography
The other partner in the development
of TaDiRAH was a working group in DARIAH, the
European initiative to build a "digital research
infrastructure for the arts and humanities" (see
[Romary 2014]). This working group was concerned with
research and education and formed part of DARIAH-DE,
the German contribution to DARIAH. The group aimed
to establish a principled overview of the research
methods and procedures shared by humanities
scholars. Presented as prose, such an overview
(published in [Reiche, Becker, Bender, Munson, Schmunk, Schöch] could help
newcomers understand the field more quickly and
could connect established and digital research by
demonstrating their fundamental methodological
similarities. Presented as a taxonomy or ontology,
the same overview would form the basis for discovery
systems for bibliographic references, tutorials or
research projects. The first use case for such a
taxonomy was DARIAH-DE's "Doing Digital Humanities"
public bibliography on Zotero [1], which became a test bed
for a keyword system that would become one source of
seed terms for TaDiRAH.
3. Digital humanities taxonomies
The taxonomy of digital research methods
described here builds on previous work towards a
structured and principled overview of the complex field
of digital humanities. The following three approaches
were particularly influential for the development of
TaDiRAH.
The goal of providing an orientation to
and a means to think about the field are at the heart of
McCarty and Short's idea of the methodological commons.
Using the metaphor and the tool of "mapping" to
represent the complex "terrain" of the digital
humanities, McCarty & Short (2002) suggest a map
which has the "methodological commons" at its center.
They define the methodological commons as "an
abstraction for the computational methods that the
various disciplines of application share", which
functions as a space of encounter between "disciplinary
groups" and "areas of learning". The research methods
are not named, but are represented through broad data
types with which they are associated, such as "narrative
text", "images", or "music". McCarty and Short's data
types are roughly equivalent to TaDiRAH's "Object"
category.
With a similar goal of structuring and
abstracting from individual research undertakings, but
using a somewhat different approach, John Unsworth
proposed in 2000 a short list of "Scholarly Primitives"
of research, particularly in the humanities. Contrary to
McCarty and Short, Unsworth defines scholarly primitives
as "activities [which] are basic to scholarship across
eras and across media"[2]; so fundamental that they
form "basic functions common to scholarly activity
across disciplines, over time, and independent of
theoretical orientation." Unsworth's tentative list was
the following: Discovering, Annotating, Comparing,
Referring, Sampling, Illustrating, Representing.
Unsworth’s formulation laid the groundwork for the
TaDiRAH team to include both methods and objects into
the taxonomy, and to keep them as separate entities with
the potential to enter into multiple and changing
relationships.
Another, very ambitious undertaking,
situated at an even higher level of abstraction, is
documented by [Benardou, Constantopoulos, Dallas, and Gavrilis 2010]. Their outline of a
"Conceptual Model of Scholarly Research Activities" is
based on an activity model which includes inter-related
entities such as research activities, research goals,
methods, procedures, tools, information objects, and
actors.[3] This model does not
analyze research methods in isolation, but
contextualizes them in the framework of the entire
research process. From the beginning, TaDiRAH was meant
to be designed in a way that would allow it to be
integrated into such larger undertakings.
4. Sources and alignment
TaDiRAH terms were derived from three
very different sources: the DiRT categories, the
DARIAH-DE tag set and the arts-humanities.net taxonomy.
The sources were not selected as part of a long process
of research and deliberation but were pre-determined
based on the particular use cases that motivated the
work. The task of alignment involved mapping the
taxonomies to one another, while attempting to address
differences in structure, scope and granularity as they
arose.
4.1 DiRT categories: tools classed by function
The categories used by the DiRT
directory were developed as an ad hoc classification
of digital research tools used in the arts and
humanities, for use on the original DiRT wiki.
Despite the positioning of DiRT as a
humanities-oriented "digital research tools"
directory, most of the tools listed in DiRT
originated in other domains, and have non-research
applications. The categories form a flat list of 32
tool functions without scope notes, and are
accompanied on the site by a rapidly growing and
unwieldy list of uncontrolled tags. As illustrated
in the frequency distribution chart below, three
categories (data analysis, visualization, and
annotation) have a significantly higher number of
tools associated with them than the other
categories; these were candidates for division into
a series of more granular categories.
Figure 1.
Frequency
distribution of the DiRT categories as applied
across DiRT tool
profilesCategories on the high end of the
distribution (e.g. Data Analysis) were candidates
for dividing into a series of more granular
categories, while those on the low end (e.g. Staying
Current) would be candidates for merging with other
related categories.4.2 Arts-humanities.net taxonomy: projects classed by method
Arts-humanities.net was developed by the Centre for
e-Research at King’s College London and had its
origins in a 2008 collaboration between the Information and Computer Technology Guides database and the Arts Humanities Research Council ICT Methods Network. Its primary purpose was to promote and provide access to information on the use of digital tools and methods for research and teaching in the arts and humanities. In support of that mission, it was home to a directory of projects, events, centers, case studies and other information. It was also home to a methods taxonomy where
"methods" refer to "computational methods
used by artists and humanists."
According to the project’s former website, the
taxonomy is based on one originally developed by
the former Arts and Humanities Data Service
(AHDS) for classifying projects. The taxonomy
structure is broad and deep containing seven top
level methods categories, which include between
eight and twenty-five specific methods each; see
below for the distribution of top-level
categories across all content on
arts-humanities.net. All but the top-level
categories, which appear to be used primarily as
guide terms to navigate through the taxonomy,
include detailed scope notes. Taxonomies such as
this, with complex hierarchies that are both
broad and deep, tend to be faced with challenges
regarding consistent application, updating and
maintenance (NISO).
Figure 2.
Frequency
distribution of top-level arts-humanities.net
categories as applied to content on
arts-humanities.net4.3 DARIAH-DE tag-set: digital humanities literature classed by research activity
The DARIAH-DE taxonomy was developed
specifically for implementation in the Zotero-based
"Doing Digital Humanities" bibliography. It was
informed both by the arts-humanities.net methods
taxonomy, Unsworth’s formulation of scholarly
primitives, and the idea of phases in the research
lifecycle. The focus was primarily on research-based
activities and objects. It originally consisted of
nine top-level interdisciplinary activities that
matched methods from the arts-humanities.net
taxonomy, as well as a second level of forty-nine
methods and a limited list of object types.
4.4 Taxonomy alignment
There are two basic approaches to
human-mediated taxonomy development: top-down and
bottom-up. A top-down approach begins at the top
level of the hierarchy, followed by the second level
and so on. It starts with developing the basic
structure or framework. In contrast, the bottom-up
approach is informed by an existing collection of
content (documents, objects, datasets, vocabularies,
etc.) with the resulting scheme emerging from that
content (see [NISO 2010]. All of the source
vocabularies described above were designed using
some combination of the above approaches. The
development of TaDiRAH, which is based entirely on
existing taxonomies, was the product of a bottom-up
process.
Adopting a bottom-up approach made
the most sense given our limited resources and
pragmatic aims. To begin with a top-down process
would have meant starting over. It also had the
advantage of being user centric and was aligned with
the principles of user
warrant and literary
warrant (see [NISO 2010] leveraging terms
already in use – terms which had been applied to
existing content by different classes of users in
the DH community: tool users, developers and
practioners adding and tagging content in DiRT and
scholars adding and tagging content in the DH
bibliography.
Given the differences in structure,
scope and granularity across the selected sources,
the team was faced with a number of challenges
including:
- Making distinctions between goals, methods, and techniques (i.e. all are related to activities).
- Specific terms that could be mapped to more than one category (e.g. modeling).
- Terms with several different meanings (e.g. visualization as an activity or as an object).
- Categories combining multiple concepts which required decisions about whether something was distinct enough to stand on its own as a separate category (e.g. storage vs. storage and dissemination).
- One to many relationships between goals/methods and techniques (e.g. mapping).
Each source taxonomy was created
with a slightly different purpose. Aligning the
sources detailed above was an iterative process that
included human-mediated matching, review and
discussion. The larger challenge was to go beyond
identifying similarities among the taxonomies, and
review differences in turn, considering the nuanced
ways in which a term or series of terms might be
interpreted and applied.
Following some general
guidelines, [4]
the team began with an analysis of the existing DiRT
taxonomy and how it was applied across DiRT content.
That was followed by a series of mappings, starting
by mapping DiRT to DARIAH-DE, identifying and
resolving points of poor alignment. It was much more
challenging to map the results of the DiRT/DARIAH
alignment to the very granular art-humanities.net
taxonomy. Fortunately, the latter included detailed
scope notes which facilitated discussions and
decision-making throughout the process. The final
step before releasing TaDiRAH for public comment was
adding scope notes to all terms in the final draft.
5. Public review process & revisions
On September 12, 2013, the team sent out
a call for feedback on the first public draft of the
taxonomy, via the Humanist Discussion Group. The public
draft[5]
was open for comments for a two-week period, where it
received over 60 comments from individuals outside the
project team. In addition, the team received multiple
emails pointing to published and unpublished work in the
area of digital humanities taxonomies. The comments
ranged from suggesting that scope notes be rephrased to
discussions about the best choice among a set of
near-synonyms to use as the label for an overall concept
("scanning" vs. "imaging" vs. "digitizing" was the
subject of particularly active debate).
Prior to the first round of feedback,
specific techniques (such as "stylometry" and "topic
modeling") occupied the second level of the taxonomy, as
did much broader terms, like "publishing" and
"annotating". Multiple commenters pointed out this
issue, and recommended that "techniques" be moved to a
separate list. This was preferable to creating a third
level of the taxonomy, as it would be difficult in some
cases to unambiguously assign a technique to only one
parent term. Having a separate list for techniques had
the added benefit of better supporting the rapid
evolution of new techniques, without requiring constant
revision to the core TaDiRAH terms; like the "objects"
list, "techniques" would be an open list.
Due to an extensive amount of detailed
and thoughtful feedback, the process of revising the
taxonomy took almost five months of periodic meetings
and asynchronous work. In early February 2014, the
taxonomy team opened a revised draft for feedback, this
time for a week. While there was still considerable
engagement (with over 20 external comments), comments in
the second round were mostly focused on smaller issues
of phrasing, and there were no fundamental challenges to
the structure of the taxonomy. Within a week after the
feedback period closed, the taxonomy coordinators were
able to incorporate the proposed changes and release the
first public version of the taxonomy, version 0.5.
6. Current version of TaDiRAH
Realizing that there needed to be some
easy way to refer specifically to this taxonomy, the
organizers devised the name "TaDiRAH", which stands for
"Taxonomy of Digital Research Activities in the Humanities". The name is also a near-anagram
of DARIAH and DiRT, the organizations responsible for
its development. The only Google result for TaDiRAH at
that time indicated that it was the name of a child’s
dragon on the virtual pet website Neopets; in honor of
that creation, the team adopted a silhouette of a dragon
as the TaDiRAH logo.
In its current version, TaDiRAH consists
of several sets of terms: two closed sets of so-called
Research Activities, one with eight top-level categories
that represent broad research goals, and below that a second set of more
fine-grained research methods. In
addition, there are two open lists, one representing
specific research techniques and
one representing research objects
to which methods and goals can be applied.
The goals roughly cover the entire
research process — Capture, Creation, Enrichment,
Analysis, Interpretation, Storage and Dissemination. An
additional "meta" category includes activities that
transcend all other categories (e.g. "Assessing" or
"Community Building"). Each goal includes three to seven
methods, with the methods section of TaDiRAH containing
40 terms in all. For example the research goal of
"Capture" can be achieved using the following
methods:
Capture
..Conversion
..Data Recognition
..Discovering
..Gathering
..Imaging
..Recording
..Transcription
There are two separate open lists of
terms that can be associated with terms from the
goals/methods section. They include 36 terms
representing a wide range of digital research objects
(e.g. "text", "metadata", "manuscript") as well as 34
terms representing specific research techniques (e.g.
"Topic Modeling", "Debugging" or "Gamification"). For
example, a tool such as SIGIL, which is used for
creating eBooks, could be tagged with the terms
"Creation" (goal), "Writing" (method), "Encoding"
(technique) and "Text" (object). A tool such as QGIS
could be tagged with the terms "Analysis" (goal),
"Spatial Analysis" (method), "Georeferencing"
(technique) and "Maps" (object).
7. Early adoption
Apart from DiRT and the DARIAH-DE
bibliography that were the initial test environments for
the taxonomy, other initiatives have shown interest in
TaDiRAH or have already begun to apply it to their
content.
Applications that emerged from the
taxonomy development team’s own work include the use of
TaDiRAH within the DARIAH Teaching Resources Registry
available on the project website [6]
and the DHCommons project directory. DHCommons is still
in the process of developing a new project profile
schema that can accommodate the projects originally
stored on arts-humanities.net; TaDiRAH categories will
be a core part of the new profiles once they are
deployed. An implementation on the DARIAH-DE portal
(DARIAH Germany’s website) is planned and currently
under way. The "Doing Digital Humanities" bibliography
on Zotero has also adopted the current version of
TaDiRAH.
In addition, a variety of projects and
initiatives have adopted TaDiRAH for structuring their
data. One example is the draft of a DH Course
Registry[7]
hosted by the Dutch CLARIAH initiative in collaboration
with DARIAH-EU. In the near future, each member country
will provide an overview of digital humanities courses
that are being offered in that country, including a
visualization and links to each of the programs. A
consistent classification built on TaDiRAH keywords will
support a well-interlinked European digital humanities
landscape. There has also been discussion about applying
TaDiRAH to the classification of in-kind contributions
within DARIAH-EU.
Individuals working on smaller projects
that deal with structured data have also requested to
use TaDiRAH. For these projects, it is especially
attractive to implement a widely-used taxonomy in order
to become or remain visible within the scholarly
community. Examples include "Zeitschrift für Digital
Humanities"[8],
a digital humanities journal that is currently being
designed in Germany and which is considering using the
taxonomy to classify contributions, as well as the
German DHd-Blog[9],
which is considering TaDiRAH as a way of tagging
posts.
8. Current and future development
8.1 Dissemination
TaDiRAH is publicly available via
GitHub under a CC-BY license. In addition to the
actual taxonomy containing activities, objects, and
techniques, the repository includes information on
coordinators, the initiatives using TaDiRAH, and
related presentations and publications. GitHub also
has support for versioning and issue tracking. When
users report issues encountered when implementing
TaDiRAH in their own projects, this will lead to
improvements in future versions.
In support of the pragmatic goal of
creating a resource that would be widely available
to a distributed community of users and which could
be applied in a variety of contexts, it was
important to provide a standards-based
machine-readable version. The taxonomy team used an
instance of TemaTres Vocabulary Server (Ferreyra
2014) hosted by DARIAH-DE[10]
It produces SKOS core and makes TaDiRAH available as
linked open data. The SKOS version is available via
GitHub[11]
and through TemaTres as a SPARQL endpoint[12].
8.2 Further revisions
The taxonomy team will remain in
close contact with the groups implementing TaDiRAH
on the DiRT directory, the DARIAH "Doing Digital
Humanities" bibliography, and DHCommons, in order to
identify opportunities to revise TaDiRAH. Growing
interest in the use of TaDiRAH may require a more
formal review of best practices for documentation,
sustainability, and governance of public value
vocabularies, particularly those made available in a
linked open data environment. Standards for the
latter are still a work in progress but important
discussions have begun among stakeholders in the
information standards community including W3C, NISO,
DCMI and others[13].
The taxonomy team intends to make
TaDiRAH multilingual, a feature requested by the
community and supported by TemaTres. There have
already been volunteers from several countries to
help achieve this. Interactive visualizations are
another potential feature that may help users
navigate the taxonomy. Here, first experiments have
been made using SKOS play![14]
which might be integrated in future versions.
8.3 NeDiMAH
Within the broader DARIAH context,
the work on TaDiRAH was originally viewed as part of
the ongoing cooperation between the research and
education working group of DARIAH and NeDiMAH, the
Network for Digital Arts and Humanities. NeDiMAH’s
goal is to "contribute to the classification and
expression of digital arts and
humanities"[15] by developing a
theoretically sound ontology that can classify work
in digital arts and humanities, thereby contributing
to its visibility and academic credibility.[16] Ontology development has
moved forward in collaboration with DARIAH as there
was a considerable overlap and synergy between the
two organizations. Developing this ontology is
necessarily more challenging and time-consuming than
developing TaDiRAH, which has a pragmatic
orientation. Nonetheless, the TaDiRAH team intends
to work with the NeDiMAH team to ensure that the
taxonomies are interoperable to the greatest extent
possible.
9. Conclusion
TaDiRAH is currently available in
version 0.5 in several human-readable as well as
machine-readable forms. It has been implemented in
several different contexts, has been presented at
several high-profile venues (including the Digital
Humanities Conference, see [Borek, Dombrowski, Perkins, Schöch 2014] and the
Dublin Core Metadata Initiative Conference, see [Perkins, Dombrowski, Borek, Schöch 2014], described in dh+lib (see
[Dombrowski2014], and is likely to be adopted by more
projects. What can be learned from TaDiRAH's brief
history so far is that a bottom-up approach such as the
one adopted by TaDiRAH can work if a small but
sufficient number of people can be brought together to
collaborate on a common goal. In the context of a
taxonomy aiming to attain a certain level of consensus
from the community of potential users, it has been
essential to quickly publish and publicize preliminary
versions of the taxonomy in order to gain the broadest
range of feedback possible. This seems particularly true
in the context of a taxonomy that prioritizes pragmatic
adoption over theoretical soundness. The TaDiRAH team
has sought to learn as much as possible from previous
practical implementations in the same domain of digital
research methods in the humanities, while seeking to
address issues arising in projects that implement
TaDiRAH. We strongly believe that the best way to move
forward and learn more about the strengths and possible
limitations of TaDiRAH lies in actual experience with
using it in various contexts. In this sense, the value
of TaDiRAH and the true potential of the development
approach undertaken here will increase as more projects
adopt it and feed their experience into future versions
of TaDiRAH.
10. Acknowledgements
The authors would like to thank Matt
Munson for his work as part of the team responsible for
the initial development of TaDiRAH.
