July 22023: DH (Digital Humanities) 2023 is in Graz, Austria this year. These notes are being written live so they are likely to have lots of typos and inaccuracies. Send me email if you find things to be corrected. The programme is here.

DH 2023 had 775 submissions and we ended up 28 workshops, 16 panels, 125 short presentations, 108 long presentations, and 112 posters. There are 874 participants of which 98 are online.

July 11, Welcome and Opening Keynote

After a day of ADHO meetings we moved to the Congress Centre. The Local Organizers welcomed us and thanked the people that made the conference possibility.

Then we had a welcome from Martin Polaschek, Austian Federal Minister of Education, Science and Research. Then we had David Ram from this. Then we had Peter Riedler, the rector followed by Susan Brown, outgoing COB President of ADHO.

Keynote: Professor Sarah Kenderdine: Two-Fold Revolutions: Computational Museology in the Age of Experience

Sarah Kenderdine is a prof at EPFL.

Kenderdine started with a video of dervishes twirling as a way of thinking about "revolutions". Her research is a combination of revolving machinery and museology. She then talked about the Imaginary Museum of Revolutions by Tjebbe Van Tijen and Jeffery Shaw. She talked about mapping revolutions and types of revolts. Jeffrey Shaw then in 1989 created a project called Revolution. The installation had a monitor that you could turn around and get different video. The visitor could participate.

She then talked about the architecture of world fairs like the 1851 Crystal Palace. Others followed. The Paris Exposition Universelle (1889) was an immersive experience. Kenderdine talked about a project to immerse people in this exposition in a cave like space with sterographic displays on the round wall.

Museums and exhibitions moulded the imaginary. The panorama is around us.

The virtual museum is the next move. The Museum Inside the Telephone Network (1991) explored how communication networks could influence the museum. It had an enormous set of contributors. In 1989 Jeffrey Shaw created one of the first virtual museum.

The next revolution was about immersion with special domes and spaces designed to extend the sensorial. All sorts of immersive spaces from the 19th century panoramic form to the new spaces like caves. There are 15 panoramas left and they are trying to argue that these should be preserved as intangible cultural property. She talked about Swiss panorama painting. They are conserving it in their lab and will then image it. They will create a digital search environment and a exhibit space with 3D surrounding screen in the round.

She then talked about the scanning of an etching of Constantinople prepared for a panorama. They can overlap the current view from the same tower.

The Digital Twins Revolution is about how objects take meaning from other things around them. Deep Fake: Art and Its Double is a project to repurpose the digitally visible through participatory interfaces. She tackles the digital materiality. Augmented replicas confront authority and expectations. Digital counterfeits are all over and challenging our ideas about art and ontology and materiality. Cultural deep fakes need to be recognized as we think about

Double Truth and Digital Twins are projects that let museums negotiate identity and decolonize

Co-creative and participatory museology is the future. Songlines: Tracking the Seven Sisters is a songline project in Australia that has been designed to be projected in a dome. They worked with the custodians of the story. It was a watershed in collaboration between custodians and curators.

The Computational Revolution offers a paradigm shift for curators. They can create new fine-grained collections and ontologies. Openness is a critical issue. She talked about small dome project which lets people visualize and explore Jazz from Montreux.

In the Future Cinema Systems project and other projects she is working with partners on various new ideas for interfaces for the archive and in the museum.

We have already arrived at the time for the revolution to come. We have arrived at moment of computational museology where all sorts of new objects and interfaces can be imagined.

It is hard to capture the density of her talk and visualizations. She repeatedly used revolving imagery and language.

There was a question about protecting copyright or art. She seems to be trying to use NFTs to protect some works, especially high rez digitizations of cultural objects held by museums.

I'm reminded of the panoramas I've seen like the (now closed) Cyclorama of the Holy Land.

Day 1: Wednesday July 12th

Exploring the Borderlands (Panel)

The first session I went to was a panel on Exploring the Borderlands. Not all of the panelists couldn't make it, which says something about the geopolitics of digital knowledge.

Domenico Fiormonte was going to talk about "Tackling the issue of epistemic injustice in the DH." He would have talked about historical epistemicide. He was going to talk about projects contributing to decolonizing DH.

Then we got a video from Padmini Ray Murray, "My DH is not your DH." ... what is My DH? She talked about setting up a DH programme in India for design students. They focuses much more on the role design plays. "All digital futures are designed futures." She talked about Design Beku, a collaboration that works at the intersection of design and digital humanities. They are inspired by design justice and try to care for the communities they work with. They call this "infrastructuring as me." They are developed co-owned feminist servers that serve communities. It is not a masculinist system paradigm, but a community centred approach. They work with health workers and use software PAPAD that allows audio contributions that can be annotated. The archive will then be made available in the spirit of happier design futures. See designbeku.in.

A video from Dibyadyuti Roy was next on "The Revolution will not be Digitized: Minority (Cultural) Viewpoints." He talked about Structures, Struggles, Strivings. He showed his Indian passport which prevented him from attending. Nesxt to it he placed the "big tent of DH." Both are ontological artifacts that provide rights and constraints. Being in the tent confers benefits and knowledge (or a framework of knowledge). For both structures there are outliers who are left out. If you are not in the big tent or don't know about it, where are you?

He then showed my diagram of Computing in the Humanities. In India the rise of computing was tied to nuclear infrastructure. Then cybercafes introduced computing to people. One need to understand the rhizome in the local culture. How is computing connected to power structures. In Europe computing was connected to different structures than in India.

He talked about Digitality os Nepantia - "A Space of Construction" without technopositivist goals. The WhatsApp he set up became a place for finding oxygen during Covid.

Dibyadyuti talked about DHARTI of which he is the Vice-President. He wants to support community based DH. Online conferences can actually support different communities in creative ways - communities that cannot meet otherwise. What sorts of communities do we want? What does local DH mean? Not localized. Emerges from sense affective fellowship. Come together around values such as openness and collaboration. (Spiro 2012)

Daniele Metilli, "Beyond Data Borders: The Sloane Lab Experience". Daniele talked about the case study on a specific collection, the Sloane Collection which is fragmented across different institutions. It has 70K objects from biological specimens to logs books. He talked about how datasets can be silos. They are trying to look at how different datasets can be linked. They are also looking at the empty spaces - what and who is left out. Sloane rarely talked about the people he paid to collect specimens.

He talked about how there are many kinds of borders. There is often knowledge in datasets that can't be accessed by anyone. It might be access information kept by the institution and kept private. There is a lot of knowledge that needs context to be understood. He talked about how a Critical Approach to Data could support multivocality, data biases and absences, usability and accesibility, and gather user requirements. There is another team trying to connect communities where they are trying to reach to communities from which the collection was drawn. They follow CARE anti-colonial approach as this collection was assembled during a colonial period and in a colonial fashion.

He concluded that despite this we will not solve the problem. The problem is structural and due to inequalities. He seemed to be suggesting that no amount of careful digital design will overcome the deep inequalities.

Quinn Dombrowski talked about SUCHO: A Case Study in Balancing Listening and Action. She had been spinning yarn on the stage until the rubber band broke and she could revolve any more. The revolution came to a stop!

SUCHO was a project that responded to the Russian invasion of Ukraine. SUCHO was set up to help archiving Ukrainian cultural heritage website (with a wide definition of heritage.) 1500 volunteers were brought together and documented what they were doing.

The problem was that little in Ukraine had been digitized. As the project went on they needed to deal with the physical archives. They didn't want to be tech-bros who wanted to only do digital. What they wanted to do was protect Ukrainian heritage. They have been getting equipment to Ukraine to help digitize stuff. It also takes a lot non-automatable labour.

They are just data-sitting. The plan is to give it back. They won't create a digital archive for people outside Ukraine.

Then Quinn held up Eli who is channeling Melissa Terras's "What makes us think we're better than this? Digital Humanities' troubled relationship with the Academy.

There are assumptions about the value of libraries. Libraries assume they are good and this leads to burnout and other issues. There are likewise arguments about access and so on. We need to fight thinking that we are brilliant and return to critique. We need more self-awareness. We need to avoid techno-utopianism in DH.

There was a question about what people in power can/should do? It is easy for those in power to talk about being critical, but isn't there some hypocrisy to that call? Some ideas mentioned:

Help those who can't afford conferences like DH afford to do it.
Datasets should be created in a participatory fashion.
Say yes when you can.
Lean out.
Mentor those left out and help them in.
Pay attention to hiring.
There is both power and privilege. Be careful about both.
Pay attention to faculty/staff

What can I as a tenured faculty do? What are the forms of power that I have that I can use.

I have the power of being able to take on leadership positions. I can use these to support marginalized research and work.
I have the power of being in the faculty caste in the uni. Pay attention to the caste system in the university. Be careful and respectful of staff and students.
I have the power of teaching. Try to teach well and care for students. Forgive them when they stumble. Try to decolonize the curriculum and the courses you teach. Help the students outside class, especially those new to Canada, but respect their privacy and autonomy. Remember to write letters of recommendation on time.
I have the power of administration and through that funding. Do your part and learn to do it well. If you have power over funding, try to be fair in giving it out.

Visualizing Text Session

The second session I went to was on Visualizing Text.

Elton Barker, Mapping Antiquity in Collaboration: The Digital Periegesis Project

Elton Barker presented the first paper. He started by saying he was a "simple text guy." He reads Pausanias's Periegesis during the day. He started with an older Arabic map as a way of thinking about non-modern ways of looking. They give us different ways of seeing that deconstructs the normative framework of seeing.

Annotation - the new technology of the annotated manuscript - is a visualization of sorts. In Perseus you have digital annotation. They are using the Recogito (from the Pelagios network) to annotate their Pausanias text with geospatial info producing linked data. Then they can get a map of all the places mentioned in the book.

He then talked about challenging the itinerary view. He showed network views (Nodegoat) and grammatical views.

He then talked about the power of linked data for a deep dive where you can go from the Pausanias text to the arachne database to see other information for the spaces discussed. This can then show you absences - things your text doesn't mention but are on shared maps.

Peripleo is Greek for journeying about. There is a new version of the Peripleo tool that he is using. You can get mashups of linked data from different people.

He talked about the collaborations under the Pelagios Network. Different people building incremental tools. Diffused sustainability.

Margherita Parigini, The Dots and the Line: How to Visualize the Argumentative Structure of an Argument.

She started by talking about the shapes of argumentation - the line, progressive, circular. We want to take arguments and spatialize them to move through them as if on a map. (See Learning to Think Spatially). This led to the idea of comparing essays using their visual form (not the text.)

Visually representing an argument is not new. We have a history of logic diagrams. They wanted a method that was transferable and comparable. They worked with connectives - the joins of argumentation. They used the Lexicon of Italian COnnectives (ILICO) to get 173 Italian connectives.

The essays they are comparing are of Italo Calvino. They converted two collections of essays. They searched for connectives in the essays and she also tracked "dubitative text", a phenomenon in Calvino that she tracked using Bert.

She designed a visualization that was circular. She developed it in Observable and then reworked with Figma (?). The circles have neat dots for connectives and wider bars where there is dubitative text. She could see which connectives introduced dubitative passages.

She closed by commenting on the collaboration between different fields. It is like the drawing on a Calvino collection with a fish and an eye.

Krister Kruusmaa, Communication Landscapes of the 19th Century: The Speed, Geographical Coverage and Content of News in the Rigasche Zeitung

Kruusmaa works at the National Library of Estonia. He is looking at historical news networks with the Crimean War as a case study. Digitized newspapers are an "Eldorado for Historians." Looking at text reuse has become a great way to track networks - who reuses what text from who. Text reuse has problems too. It is expensive and doesn't work across language borders. How can use other methods for studying networks of communication?

He worked with one paper published in Riga (Latvia), a huge trading hub of baltic German trading. The Rigasche Zeitung is on a double border of politics and language. 86 years, 289K of articles. OCRed. He used regular expression to capture place-date information. He kept 350 most common place names.

He showed how the place-date data could be used to visualize news flow. Imperial capitals, especially German ones dominated. He could also look at changes in what places were referenced over time. Berlin became more important. He could look at the speed of information (the time elapsed between date of news and the date of the event reported.) The speed of news drops as the telegraph and railways are introduced. By the 1870s the time for news to reach Riga drops to a few days. You can also see the effects of blockades on news.

He showed a map distorted to show news distance. There was an interesting change over time. With the telegraph you get news from the metropoles faster than nearby news. It creates a fold in space-time.

He then gave the case of the Crimean war, one of the first to be photographed and reported using the telegraphed. In Russia the news was censored. How and what news reached Riga? He used Top2vec algorithm and tracked 4 topics over the years of the Crimean war. He tracked where the war news came from. It often came from Constantinople, Vienna, London, Paris, and Copenhagen. It was coming from Russia's rivals, not from Russia. Reports about allied naval movements in the Baltic would have been important to Riga readers.

See bit.ly/positioning-riga for more.

I must say that all three of these presentation were brilliant in different ways. One of the better sessions on visualization that I've hear.

Theoretical Frameworks

After lunch I went to the session on Theoretical Frameworks

Susan Schriebman, Social Justice in the Digital Humanities Community of Practice

Issues of ethics, social justice, duty of care and related subjects are more and more part of DH. Susan is part of #dariahTeach which is hosting the course(s). They are developed a community driven course for which they gathered a number of example projects like "Walking the Archive of District Six Cape Town, South Africa." YEG Police Violence Archive is another project. Dariah is giving them visibility through the course.

They don't want to be gatekeepers. They have four units:

Intro
Ethical Research and Values in DH
Approaches to Ethical Decision Making
Social Justice in DH Practice - Knowledge paradigms
The Toolbox - Resource for those doing ethically aware research

She encouraged us to send case studies.

Gonzalez-Perezl, A Proposal for the Demarcation of Digital Humanities

Gonzalez-Perezl

This is part of a large team in the Hispanic Association of DH. Their motivation is questions like:

What are associations like ADHO supposed to engage with?
What should we teach?
Who is working in DH?
What proposals are eligible for a DH grant or award?

Demarcation is an old business. It can be a way of including or excluding. (My take: Do we really want to get into this again? Do people argue about what demarcates philosophy?)

GP argued that it is the digital that distinguishes DH from the humanities. There is nominal and non-nominal uses of technology. Nominal is the expected use. Non-nominative is the more innovative but also involves friction.

Usage, Analysis, and Development is the demarcation approach. Each level builds on the others. Usage is using technologies for humanistic concerns. Analysis is the more critical approach and involves understanding the technology and research. Development is where you build your own. Using WordPress to build a new site is not development. (And now making is back.) He argued that threshold 7 is when you have state of the art being advanced in both disciplines (DH and H) at the same time. This would be trans-disciplinary.

Some properties that are not definitional include:

Open and shared data.
Working in collaboration
Multi-disciplinary work

They have external reviewers to provide comments and are welcoming of comments.

They say that they are not judging value of proposals, but I don't believe it. They say they want to proceed in a consensual fashion. Hmm....

Jennifer Edmond and Pat Treusch, Human, Technology, and Culture Interaction? Mapping the Landscape of Technological ‘Sister’ Disciplines

They talked about terminology like "critical digital humanities" and how there are different senses of what that might be. It could be H criticism of D. Or it could be C of DH. So they looked wider - at projects like Full Stack Feminism, or Indigenous principles, or HCI.

In the human+ project they are trying to bring feminist science and technology studies together with DH. FSTS tries to clear the ground to make other technology futures possible.

FSTS has matters of care, material methods, and agential realism. Together with DH they also focus on power, are sensitive to constructiveness, analyze discourse, and so on.

They started some thought experiments starting with Critter Culture (asking about robots). Why do we assume robots should look and act like humans? Thought experiment 2, embodied DH? They think about infrastructure, spaces, desks ...

Rabea Kleymann, Investigating Constructivist Paradigms in Digital Humanities Scholarship

Rabea started by pointing out that everything is "constructed" in DH. She talked about multiparadigmatic DH drawing on Kuhn's ideas about paradigms. What are the blind spots of constructivism?

There is not one constructivism, but many from cybernetics to Actor-netork theory. A minimal definition would say that X constructs Y. We go from questions of what to questions of how. How do DH scholars talk about constructivism? She quoted various definitions including Johanna Drucker and her idea of "capta".

Epistemological premise of humanistic inquiries, entanglements of interpretation and construction, reflexivity.

The underlying hypothesis could be that different methods construct different epistemologies or epistemological objects. Multimethod constructions will construct multiple epistemological objects.

DH scholars (re-)construct contingencies of data settings. Looking an API output one can see all the different standards and constructions.

What are pitfalls or potentials of constructivist paradigms? Constructivism can often overlook things.

Shall we address more ontological and material questions of "what"?
How can we practice modes of criticism that go beyond constructivism?

(This paper was brilliant!)

Nabeel Siddiqui, An Undue Burden: Race, Gender, and Mobility in Digital Humanities Conferences

Siddiqui started by talking about how he could have started with a joke or anecdote. Conferences are endurance . Conferences are an arena of power, social and intellectual capital are played out. Conferences are understudied. Concerns about travel and inclusivity keep up coming up.

Siddiqui started a project looking at the database of DH conferences. He used universities as indexes and mapped travel. Affiliation was useful to build a animation of travel. He played an animation of movement showing

Gender and Ethnicity were also mapped. There are packages for rehydrating gender and ethnicities which he used to generate predictions. He found the following issues for which he has suggestions.

Most conference goers are unwilling to travel long distances. We should hold conferences in the global South.
Female researchers are less likely to travel. We should find ways to support female scholars.
More experienced authors are more willing to travel. We should incentivize newer scholars.

It would be interesting to see if there is age or status correlations.

He was asked about the prominence of English.

After the talk he told me that I come up as the person with the most papers. I'm not sure if I should be proud or ashamed.

Machine Learning

Then I went to the Machine Learning session.

Michela Vignoli, Revolution or Evolution? AI-Driven Image Classification of Historical Prints

They are looking how AI can support historical research. They are looking at travel accounts from the Ottoman empire. The project, ONiT or Ottoman Nature in Travelogues, builds on a previous project that built a corpus. These are travelogues to the Ottoman empire. They have about 1500 of them and they are sticking to trips that actually took place.

Now they are doing image classification to train a machine. The images they want are those that show nature depictions. These are labelled and they use ICONCLASS a classification from Art History. They stick to four classes, animals, plants-vegetation, landscapes, and maps-atlases.

Their experiences so far introduces new forms of heuristics. They really have to ask about representations of nature that are machine readable. They could end up with all images with some blade of grass in the corner.

AI forces them to widen methods and think out of the box. Can AI support historical interpretation.

Luis Meneses, Analysis of Cyber Threats Affecting the Survivability of Online Digital Projects

Luis talked about previous work on survival of DH projects over time. Average project is online for 6 years. Then they degrade and eventually stop working.

This paper looks at whether system vulnerabilities were responsible. They used shodan.io which has a list of vulnerabilities like open ports that are should be private. They looked at servers mentioned in DHQ and DH conference abstracts. There are lots of hosts with vulnerabilities. We need to talk about patching.

This is everyone's problem. Lots of undead out there.

They have a wiki with resources to help: https://wiki.dhlinux.org

Setsuko Yokoyama and Sai Rajan, Accented DH: Assessing Fairness of Multilingual Speech Recognition Systems

Yokoyama and Rajan are looking are looking to test "Equality of Outcome" (Phillips, 2004) They are part of a larger move to promote fairness. They work with Singlish, a creole language important in Singapore which combines English, Chinese and other languages. A fair system would recognize the mix of languages. They are using the Singapore National Speech Corpus. They are making perturbations in the speech from different groups and seeing if it affects the error rates.

In Singapore there is Chinese privilege to explain poor performance of ASR in adversarial situations. The ASR performed well for Malay speakers which suggests issues with the training data. There may have been a lot more Malay training samples.

Markus Kunzmann, AI-supported indexing of handwritten dialect lexis: The pilot study "DWA Austria" as a case study

Kunzmann talked about an experiment in transcribing handwritten documents.

Peter Bel, imgs.ai. A Deep Visual Search Engine for Digital Art History

Bel talked about how they trained a model with various image datasets that can be used to search for images, either by sample image or by prompts. They used CLIP from OpenAI to let users search by text prompt. The tool imgs.ai is here.

My laptop crashed so I lost notes on the last two.

Day 2: Thursday, July 13th

Language Models

Ismail Prada Ziegler1, and Christa Schneider, Pre-Modern Data: Applying Language Modeling and Named Entity Recognition on Criminal Records in the City of Bern

The team is looking at a fascinating "Bernese Tower Book" corpus of criminal records. They wanted to see if they could apply distant reading techniques like NER and POS to this corpus of highly irregular document. They also wanted to check if different language models would work.

What they did is to train a trained a language model using secondary raw data. Then then trained a NER model using an annotated corpus. They used the Flair framework which is character based and has both forward and backward embedding. Flair works well for German and outperformed BERT and fasttext. The got good results for people and locations, but not organizations.

In conclusion, more data that fits the context works better.

Ryan Dubnicek and Ted Underwood, Piloting A Machine Learning Approach to Identify English-Language Fiction in the HathiTrust Digital Library

Dubnicek started by talking about motivation. Cultural analytics and digital humanities has exploded. Large digital libraries have become objects of computational study even though they weren't necessarily meant for that.

There are problems

Standards change over time so metadata is inconsistent
Genre wasn't explicitly catalogued until recently
Many volumes don't have rich metadata - especially non-canonical texts

The question was whether they could leverage machine learning to identify English-language fiction which involves identifying genre or writing.

Ted Underwood talked about the NovelTM Dataset which is actually 7 datasets with different purposes. It is not, however, free from error.

They are now trying to improve the process, also deal with new volumes added to Hathi. This could lead to more fiction, more research.

They are using the HTRC Extracted Features. They train the classifier on 20% and test on 80%. Human reviewers then check.

Some errors include:

Incorrect ground truth
Volumes that blur the lines of fiction like memoirs
Non-prose fiction like poetry
True errors
A very small number of volumes where reviewers couldn't agree

Takeaways include:

It is actually hard to tell fiction in some cases (folklore, memoir, travel narrative)
Sampling logic was prominent factor
Date of publication is important - especially with OCRed text
Human-produced metadata records are not infallible

Next they will run the classifier on the 1.6 million new texts (their samples had 10K)

There was a question about whether we now have the features that predict fiction? He showed a list of words that worked for them.

Toma Tasovac and Nick Budak, Humanistic NLP: Bridging the Gap Between Digital Humanities and Natural Language Processing

He started with good old-fashioned generalizations about hadn-crafted digital textual editions that come out of philological traditions. These boutique editions are rarely annotated linguistically. This is curious given the NLP tools out there. Why do so few humanists use NLP tools?

Specialization in the humanities
We are prisoners of our disciplines
It isn't clear what to do with POS
NLP tools don't work with XML texts
NLP tools don't work perfectly and humanists don't want to annotate texts automatically
Automated annotation is considered ethically suspect
Humanists don't have tools or expertise to use NLP tools

We should be able to separate manual and automatic annotation and recognize that linguistically annotated texts are more useful. What can we do to make NLP more useful? How can we integrate NLP into DH? They have developed a number of workshops. Their approach is humanistic NLP. They did the following:

prioritised ethical engagement
focused on workflows
thought about integration into humanistic research

Then Nick Budak talked about what they did. He talked about tools, text, and labour. Labour is the most pressing challenge. A lot of attention is on the technology (tools), but the real issues have to do with how you solve labor challenges like how to manage annotators. Also, what sort of support can you get? How is the labor recognized? Nick talked about issues around annotation, especially how it can go wrong in other countries and how to train.

Then there are the challenges of texts. Then he talked about tools. Humanists are used to "fuzzy" tasks, but you need to formalize to use technology. Some of the solutions include embracing free and open source software, but FOSS places more burden on you.

He closed with outcomes and successes. With sufficient training humanists can engage in NLP.

In response to a question about TEI Toma talked about a tool Standoff Converter. It allows you to convert a TEI encoded text to standoff so that you can then use NLP meant for plain text.

Big Questions

Nathan Woods, Data Problems in the Humanities, or "When everybody is special, no one is?"

Woods set out to redefine the data problem by looking at

Data is important to humanists as outlined in a recent Ithaka report. What do humanists think of data and data management? There is a resistance to data - resistance to the "data problem." There is a rejection that humanists work with data. A view that we have capta. Finally a view that we don't know we have data. These views aren't based in what humanists really do.

The team is using a case study approach and comparative historical work to understand how humanists really use data - what scholarly practices are. They picked cases that are data intensive and scholar led.

They did document analysis, interviews etc. Their preliminary theory development includes:

Humanities scholars as infrastructure designers
Use case as a sociotechnical fact

The use case - use case modeling is used to specify situation. It Use Case Model translates specifications into an actual design. Humanists didn't set out to be infrastructure designers, instead this work emerges out of project.

Questions about data are developed as part of emergence of use case. There is a focus on humanists as information users/consumers, but little on humanists as producers and infrastructure developers. Casting humanists as users of infrastructure ignores what they are doing. The framing has consequences for how the question is posed and solved.

We need to reimagine the data problem in order to do better research policy. We need to better document what actually happens in research. They are structuring dialogues that will better shape the image of the humanities.

See humanitiesdatainquiry.org.

Stefan Heßbrüggen-Walter, A Philosophical View of the Digital History of Concepts: Four Theses And a Postscript

He started with concepts and conceptual history. This leads to concepts in philosophy which leads to concepts in digital conceptual history.

Concepts are different than words, what is the difference? How can we then examine semantic change? Can philosophy help?

Why do we need the concept of concept at all? Concepts are relevant in philosophy of mind and phil of language/logic. Are they mental entities or states? Do they exist independently from the minds of speakers? There no conclusive evidence either way.

Conceptual historians would do better to do away with the concept of concept. Ultimately we have to deal with the history word use.

So what does this mean for history of digital concepts. He wants to admonish digital conceptual historians to maintain some modesty. The object of study in DH must be operationalized. We need to be able translate concepts into something that can be formalized - coded.

What does it mean to look then at history of operationalizations? WE can look at the history of the patterns of word use. Thus we can limit our claims to histories of word use.

The problem goes beyond the approach of digital conceptual history. Concepts are relied on by various ontologies like CIDOC CRM. Ontologies presuppose a concept of concept.

But it seems to me there are ways of operationalizing other than by words. Is looking at only words enough? The ontologies could be thought of as operationalizations. FRBR assumes an abstract, but real object.

Just because we can do away with concepts does that mean we have to? We have all sorts of practices that seem to presuppose the existence concepts, why not let them be? Could that be their purpose.

His motivation is thinking about concepts on a global scale.

Cultural Motifs on #bigdata - A Semi-Automated Topic Modeling from a Socio-Cultural Constructionist Perspective

A team from Leipzig presented on a project using topic modeling on a large collection of discourse on big data.

My laptop crashed so I lost my notes on this.