Abstract In this paper we compare two academic networking platforms, HASTAC and Hypotheses, to show the distinct ways in which they serve specific communities in the Digital Humanities (DH) in different national and disciplinary contexts. After providing background information on both platforms, we apply co-word analysis and topic modeling to show thematic similarities and differences between the two sites, focusing particularly on how they frame DH as a new paradigm in humanities research. We encounter a much higher ratio of posts using humanities-related terms compared to their digital counterparts, suggesting a one-way dependency of digital humanities-related terms on the corresponding unprefixed labels. The results also show that the terms digital archive, digital literacy, and digital pedagogy are relatively independent from the respective unprefixed terms, and that digital publishing, digital libraries, and digital media show considerable cross-pollination between the specialization and the general noun. The topic modeling reproduces these findings and reveals further differences between the two platforms. Our findings also indicate local differences in how the emerging field of DH is conceptualized and show dynamic topical shifts inside these respective contexts.

Citation: Puschmann C, Bastos M (2015) How Digital Are the Digital Humanities? An Analysis of Two Scholarly Blogging Platforms. PLoS ONE 10(2): e0115035. https://doi.org/10.1371/journal.pone.0115035 Academic Editor: Vincent Larivière, Université de Montréal, CANADA Received: June 24, 2014; Accepted: November 18, 2014; Published: February 12, 2015 Copyright: © 2015 Puschmann, Bastos. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited Data Availability: All relevant data are within the paper and its Supporting Information files. Funding: This work was supported by the National Science Foundation under grant number 1243622 and the German Research Foundation under grant number PU 439/2-1. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.

Introduction The advent of the Internet has profoundly affected scholarly communication [1–4]. Few scholars, whether in the sciences, social sciences, or humanities can imagine conducting research or organizing teaching without relying on email, digital library services, or e-learning environments. Formal academic publishing has undergone a series of changes with the increased availability of electronic publications, whether under an open access or toll access regime [5]. Structural changes in the dissemination of knowledge have largely been gradual and evolutionary: while the volume of scholarly publications has greatly increased in the past decades and the formal and distribution models have diversified, the form and function of research articles and scholarly monographs have remained relatively stable [6]. Meanwhile, the range of avenues available for the dissemination of informal scholarly communication has increased exponentially. In addition to formal publication venues, scholars can now communicate their findings in (micro)blogs, wikis, social networking sites (SNS) and countless other social web platforms [7–10]. Such services carry both opportunities and risks for early-career researchers, and they are used for a wide variety of purposes and with a range of motives [11–13]. While researchers are able to disseminate their findings more quickly and reach out to broader audiences than was previously possible, they also risk that their work will not be acknowledged in more traditional and hierarchical professional structures. Informal genres of scholarly communication frequently lack peer review and rely on new measures of impact, rather than the established currency of acceptance within a field [14]. As a result, researchers have overall been very careful in their acceptance of digital formats that compete with established forms of expert knowledge dissemination, largely choosing instead to focus on established formats [15]. This is especially true in the humanities, where conservatism towards new formats is particularly strong. Digital Humanities (DH) can be broadly characterized as the adoption of an array of computational methodologies for humanities research [16, 17]. During the early nineties, DH scholarship developed under the umbrella of several academic organizations dedicated to what was then commonly referred to as humanities computing [18]. These organizations brought together scholars from different fields interested in exploring computational methods for traditionally-defined humanities scholarship [19]. The suffix “digital” is increasingly used to delineate the new computational areas of humanities research (i.e. digital literature, digital archaeology, digital history, etc.). The introduction of computational methods aims among other things to supplement established humanities research routines and explore new methodological avenues, such as text analysis and encoding; archive creation and curation; mapping and GIS; and modeling of archaeological and historical data [20, 21]. Since the early 2000s the term Digital Humanities has also been used to refer to humanities research defined by a data-driven approach, in which summarization and visualization are important methodological cornerstones. Media and cultural studies, library and archival studies, digital pedagogy, and the recently emergence of MOOCs have also been referred to as Digital Humanities in a more general sense [22]. As a result, DH has evolved to incorporate a range of different definitions and is subject to considerable interpretative flexibility [23]. The central hypothesis of this study is that the variety of terms and topics associated with DH is locally configured, and that their makeup reflects different (and to a degree contradictory) conceptualizations of what constitutes DH.

DH and Social Media Because of its interdisciplinary and international character, its affinity for digital media, and its recent emergence as a scholarly movement, DH has been comparatively strongly impacted by informal communication tools such as blogs and Twitter, with junior scholars invested in DH research using such tools widely to organize, network, and collaborate. Kirschenbaum notes the important role of social media for establishing and galvanizing DH as a movement: “Twitter, along with blogs and other online outlets, has inscribed the digital humanities as a network topology, that is to say, lines drawn by aggregates of affinities, formally and functionally manifest in who follows whom, who friends whom, who tweets whom, and who links to what.” [17] Usage of Twitter and blogs has contributed to establishing DH as a brand, and it has helped to increase its visibility on a global scale [24]. While actively using social media does not make one a digital humanist, social media applications seem to be perceived as valuable instruments for intra-community communication in the DH community, rather than being used just out of curiosity or for self-promotion [11]. Crucially, there are scholars who take up blogging and Twitter because they are important channels of communication in the DH community. Such tools therefore increasingly constitute scholarly infrastructure to their users in the same sense that library services and communal mailing lists constitute infrastructure. While traditional scholarly organizations are struggling to integrate social media, DH scholars, especially junior researchers, have considerable uptake of such tools, reflected for example in the strong use of Twitter at the annual Digital Humanities conference [24, 25]. DH can therefore be characterized as an emerging digital scholarly network—a group of scholars that has integrated digital genres of scholarly communication into its communicative infrastructure from the onset. Inside such a network in which heterogeneous links connect different actors it should be possible to study the flow of ideas, trends, and discourses much more effectively through social media than purely by assessing formal publications in scholarly journals and monographs [26]. HASTAC The Humanities, Arts, Science, and Technology Alliance and Collaboratory (HASTAC) is an online community and social network that connects researchers, young scholars, and the general public interested in a wide range of subjects associated with DH and peer-to-peer learning. Founded in 2002 by Davidson and Goldberg [27], HASTAC emerged as a consortium of educators, scientists, and technology designers funded by the National Science Foundation, the Digital Promise Initiative, and the MacArthur Foundation, with infrastructure provided by Duke University and the University of California Humanities Research Institute. HASTAC differs from similar initiatives in that it is largely decentralized with content generated by a network of over ten thousand members including university faculty, students, and general public. The network platform is built on the Drupal content management system and requires an inclusive free-of-charge membership. Member participation varies widely, with many registering but passively interacting with the website by reading the content and a robust minority expressing their thoughts and communicating their interests by writing or commenting on blog posts, joining discussion forums, or contributing information about current events. According to the initiative’s website, “HASTAC members are motivated by the conviction that the digital era provides rich opportunities for informal and formal learning and for collaborative, networked research that extends across traditional disciplines, across the boundaries of the academy and the community, across the two cultures of humanism and technology, across the divide of thinking versus making, and across social strata and national borders.” [28]. While the platform is interdisciplinary in nature, it is strongly focused on learning and DH-related topics. Hypotheses Hypotheses is a publication platform for academic blogs. Launched in 2004, it is funded and operated by the Centre for Open Electronic Publishing (Cléo), a unit that brings together two major French research institutions and two universities: the Centre national de la recherche scientifique (CNRS), the École des Hautes Études en Sciences Sociales (EHESS), the Aix-Marseille Université, and the Université d’Avignon. In addition to Hypotheses, Cléo provides other tools via the OpenEdition portal: Revues.org, a platform for journals in the humanities and social sciences and Calenda, a calendaring tool. According to the Hypotheses website “[a]cademic blogs can take numerous forms: accounts of archaeological excavations, current collective research or fieldwork; thematic research; books or periodicals reviews; newsletter etc. Hypotheses offers academic blogs the enhanced visibility of its humanities and social sciences platform. The Hypotheses team provides support and assistance to researchers for the technical and the editorial aspects of their project.” [29] To publish on Hypotheses, a blog must first be admitted by the platform’s editorial team. Only researchers employed by institutions of higher learning are eligible to join Hypotheses after having been evaluated, and the criterion for positive evaluation is a consistent focus on academic issues. Through its policy the platform maintains some characteristics of a formal publication outlet, aiming to stimulate both open discussion within scholarly disciplines and exchange with the broader public. Hypotheses is based on the Wordpress content management platform, with a home page that features current contributions from participant blogs. In addition to English, a large portion of Hypotheses’ content is composed in French, German, Spanish, and other languages, but for the purpose of this study we only considered posts published in English. Similarities and differences Both platforms share strong similarities: they aim to promote new forms of scholarly communication and knowledge dissemination. At the same time, there are also considerable differences: HASTAC places a clear emphasis on learning and also mentions media and communication in its self-characterization. While Hypotheses is also interdisciplinary in character, it has a stronger slant towards traditional humanities subfields, and specifically towards history. The concept of scholarly blogging outlined on the Hypotheses website points to its role for intradisciplinary communication, whereas HASTAC is more geared towards interdisciplinary exchange. Despite these differences, the two platforms make an ideal case for comparison on the grounds of their functional similarities. Both are related to DH, both seek to integrate blogging into scholarly communication, and both are publicly funded. Furthermore, both platforms have been operational for a similar timespan and attract broadly comparable user communities.

Research Design Our aim is to characterize differences in the discourse that takes place on HASTAC and Hypotheses reflecting different cultural implementations of DH and different understandings of what constitutes DH. To this end, we formulated two research questions: How frequent are particular keywords associated with (digital) humanities on the two platforms (ᴴ1) and what are thematic differences in the distribution of topics in the two sites (ᴴ2)? We approached the first question by counting the co-occurrence of humanities-related terms and their digital equivalents (e.g. history—digital history) on blog posts. In a second step we applied topic modeling to the post content to identify substantial thematic differences between the communities in both platforms and their respective approaches to blogging. Based on the self-characterizations of both platforms, we expected there to be both overlap and variation with regards to the adoption of DH-related labels and overall disciplinary focus. Data The data from the two platforms were collected from database dumps containing the SQL table structure and the blog post content. HASTAC data included content posted between August 14, 2006 and August 14, 2013, together with the profile data of 11,284 users. Most users shared brief biographical information and identified a set of topical interests, institutional affiliation, and links to personal websites. In addition to the posts themselves, the Hypotheses data included metadata such as author information, timestamp, text, internal and external links in each post, which was collected between the 1st of July 2006 and the 30rd of June 2012. The language of posts was detected automatically using the language identification system langid.py for Python, which supports a large number of languages and achieves a high level of accuracy without requiring prior in-domain classifier training [30]. The material initially included a large number of posts published in languages other than English (45,528 posts) published over different periods of time. For the purpose of this investigation, we only considered blog posts in English published between the 1st of July 2006 and the 30th of June 2012, thus extracting 7,269 posts from HASTAC and 6,777 posts from Hypotheses. We performed a co-word analysis over these 14,046 posts [31] and subsequently extracted a random sample of 5,000 posts from each platform to perform topic modeling. Fig. 1 shows a frequency histogram of blog posts in the abovementioned period on a logarithmic scale, with HASTAC posts being comparatively more frequent from 2006 to 2010, and posts on Hypotheses being comparatively more frequent in the period thereafter. Activity on both platforms drops during the summer vacation months (July for HASTAC and August for Hypotheses) reflecting seasonal work patterns. PPT PowerPoint slide

PowerPoint slide PNG larger image

larger image TIFF original image Download: Fig 1. English-language blog posts published on both sites between 2006 and 2012. https://doi.org/10.1371/journal.pone.0115035.g001 Methods We approached our first question (ᴴ1) by means of a co-word analysis of keywords associated with humanities and Digital Humanities research [31]. We used one vector of twenty humanities areas (anthropology, archaeology, archive, art, culture, ethnography, history, humanities, learning, libraries, literacy, literature, media, pedagogy, preservation, publishing, rhetoric, scholarship, storytelling, knowledge) and another identical vector plus the suffix “digital” (digital anthropology, digital archaeology, digital archive, digital art, digital culture, digital ethnography, digital history, digital humanities, digital learning, digital libraries, digital literacy, digital literature, digital media, digital pedagogy, digital preservation, digital publishing, digital rhetoric, digital scholarship, digital storytelling, digital knowledge). These keywords include terms that describe fields or general domains associated with the humanities on the basis of raw token frequencies identified in the two datasets. This approach comes with considerable limitations. Firstly, the semantics of the terms differ considerably, as some describe fields of scholarship (history—digital history), while others are more general and tend to be polysemous (knowledge, media). The same applies to their prefixed counterparts, with digital history likely identifying a field, while digital media most likely describes certain kinds of technical media. Furthermore, issues of precision and recall arise, due to which not all discussion of the relevant phenomena is reliably captured and some of what is captured relates to other concepts. In spite of these limitations, we found co-word analysis to be useful, because it shows the entrenchment of the terms as convenient and fashionable labels on both platforms. We accept that such labels do not narrowly identify concepts, but believe that they are suitable to characterize the success of particular terms around which the DH community can rally. Using these terms we generated a series of term-document matrices for each of the networks. We visualized the association between humanities and DH by performing a multinomial logistic regression on the terms. We relied on the textir package for R [32] to convert the term-to-term co-occurrence matrix to a matrix of the log-odds ratios of co-occurrence. The resulting matrices (HASTAC and Hypotheses) scales the word similarity as a function of word frequency, with terms of similar semantic content numerically represented as being similar to one another [33]. After converting the log-odds ratios to distance matrices using cosine similarity [34, 35], we relied on multidimensional scaling [36] to visualize humanities and DH terms in a latent semantic space [37] with a two-dimensional density surface [38]. The second question (ᴴ2) was addressed using Latent Dirichlet Allocation [39] implementation for R [40]. R package topicmodels allows the probabilistic modeling of term frequency occurrences in documents and estimation of similarities between documents and words using an additional layer of latent variables referred to as topics. The package provides the basic functions for fitting topic models based on data structures from the text mining package tm [41]. Topics were modeled using a mixed-membership approach in which documents are not assumed to belong to single topics, but to simultaneously belong to several topics, with varying distributions across documents. To equally represent both platforms, we drew a random sample of 5,000 posts from each platform from the data previously described. Prior to mapping the documents to the term frequency vector, we tokenized the posts and processed the tokens by removing punctuation, numbers, stemming, and stop words, in order to sparsen the matrices. We also omitted very short documents (<200 characters) for the same purpose. Ethics Statement. The authors confirm that the study is in compliance with the Terms and Conditions of HASTAC and Hypotheses.

Discussion The results reported in this study can be summarized in two parts. Firstly, we found a substantial one-way dependency of DH terms on their unprefixed counterparts, as most blog posts dedicated to DH also included references to the corresponding humanities term (89% on HASTAC and 98% on Hypotheses). DH-related labels are considerably more frequent in HASTAC pointing to an unequal adoption of Digital Humanities-related terms in different local contexts. Secondly, we found a tendency in Hypotheses towards focused thematic areas representing disciplinary interests contrasted with a tendency to discuss more general, cross-disciplinary themes in HASTAC. In terms of institutional branches of humanities research, history is the areas with the largest number of posts across the networks for the sample of topics considered in this study. Areas that are not traditionally associated with humanities research (or institutions that support the field), i.e. library and media, also account for a considerable portion of the posts. We also found considerable topical differences between the two platforms. While traditional areas of the humanities and social sciences (History, Art, Law) are clearly represented in Hypotheses, HASTAC is topically more cross-disciplinary and less focused on single disciplines. Some of these topics show considerable overlap between the networks (i.e. Social Media and Data), highlighting the fact that there are areas in which users of HASTAC and Hypotheses have similar interests, while others are considerably more predominant in one of the networks. Although both networks are on the forefront of the Digital Humanities research agenda, they present considerable differences in how explicitly they use new disciplinary labels (HASTAC) and address well-established disciplinary themes without explicitly associating them with DH (Hypotheses). The differences we observed highlight that two platforms that attract broadly similar user communities may still differ considerably with regards to topics. We interpret the differences in adoption of Digital Humanities terminologies and topics across the networks to mirror different developments in DH. Whereas digital learning, digital literacy, and particularly digital scholarship are particularly prominent labels on HASTAC, Hypotheses is mostly focused on digital libraries, digital history, and digital archives. These differences are of qualitative and quantitative nature reflecting not just the personal preferences of bloggers and users, but may also indicate broader conceptual differences. While blog posts in HASTAC tend to raise issues suitable for (controversial) discussion, contributions in Hypotheses more closely mirror traditional expository humanities genres (e.g. book chapters or essays). Moreover, while HASTAC is a social network in which users can create profiles and interact with other users by posting and commenting on content, Hypotheses is a publishing platform with lesser emphasis on community building than HASTAC, and a closer alignment with traditional genres of publishing. The content of each network also presents considerable variation in terms of formats and style. The prominence of Topic #12 (Chatter) in HASTAC indicates that HASTAC’s blog entries are conceptually more like casual conversation rather than academic writing. As blogs serve different purposes for different users, the data necessarily includes posts of different genres comprising of short essays, conference reviews, book reports, group discussions, and general academic advertising. While HASTAC and Hypotheses are interdisciplinary in character, they have a strong slant towards the humanities, particularly towards learning and digital media on HASTAC, and specifically towards history on Hypotheses. Common to both networks is the small proportion of users producing the large majority of the content, which leads to a typical long-tail distribution of content within the platforms. In the last instance, the results reported in this study show that the variety of terms and topics associated with DH is locally configured and reflects different conceptualizations of what constitutes DH. We expect this study to be informative for future research grappling with the rapid establishment of DH in humanities departments. At any rate, it will be interesting to follow the ongoing maturation of both platforms and their respective approaches to scholarly blogging, as well as the different conceptualizations of Digital Humanities scholarship in North American and European contexts.

Acknowledgments We are thankful to David Sparks for helping with HASTAC data analysis and visualization, the HASTAC team for providing access to the HASTAC data, Ruby Sinreich for providing important feedback to this research, and Marin Dacos for providing access to the Hypotheses data.

Author Contributions Conceived and designed the experiments: CP MB. Performed the experiments: CP MB. Analyzed the data: CP MB. Contributed reagents/materials/analysis tools: CP MB. Wrote the paper: CP MB.