We develop tools that speed up a specific workflow, for research and cultural institutions as well as for companies and public administration. Every tool is tested on your data and documented so that others can continue working on it.
A service of Digital Humanities Craft
Transcribing early printed books and manuscripts takes time, and errors of automatic recognition only become apparent in comparison with the image.
coOCR/HTR combines text recognition, correction against the page image, checks and export in one browser application. The subject expert decides on every reading.
State Research preview with recognition, correction, checks and export. An evaluation of recognition quality against reference data is still pending. Evidence
Editing TEI XML by hand is error-prone, and a custom Oxygen framework is too costly for many projects.
teiCrafter shows readable text instead of tags, validates against the schema on saving and changes the file only where it was edited. Suggestions from a large language model remain marked as such.
State Technically tested research preview. Trial use in the daily work of an editorial team is still pending. Evidence
Letter metadata are available as XML, but who wrote to whom, when and from where can hardly be seen in them.
CorrespExplorer shows a map, a timeline, a network and further views with shared filters, directly from a CMIF file or from correspSearch. Uploaded files remain in the browser.
State Running demo with all views and export. The map view currently lacks its background map, and an evaluation with users is still pending. Evidence
The indicators of the Austrian universities' intellectual capital reports (Wissensbilanzen) are spread across separate Excel analyses, and comparing universities or several years takes a great deal of manual work.
The Wissensbilanz-Dashboard brings the public indicators of the Austrian universities together in a uniform format and presents them for comparison as a time series, a table or a report.
State Working prototype from a higher education project. The frontend is being developed further. Evidence
A large digitised collection is difficult to survey through a search form as long as no search term has been settled.
The Kulturpool-Demo presents the openly licensed objects of an institution from the Kulturpool interface as an image wall with collection structure, facets, full-text search and a detail view.
State Working prototype from a public hands-on session. The object data are harvested in advance and delivered statically. Evidence
In cataloguing museum objects it is an open question how far a large language model with image understanding can produce a classification from a photograph and metadata.
Objekt-Bestimmung places the original cataloguing of each object beside the suggestions of the LLM and marks discrepancies and cases for curatorial review.
State Demonstration environment for a workshop with museum professionals, applied to a selected set of objects. Evidence
Promptotyping is our method for developing research artefacts with AI agents. Sources, data and the domain understanding of the project are held in a maintained knowledge base from which every development round draws and into which it writes back.
State Basis of our projects. An article on L.I.S.A. introduces the method, a journal article is in preparation. Evidence
Grounded Vault is a repository template for evidence-based knowledge work with AI agents. Every load-bearing claim in a report or expert opinion leads through a verifiable anchor to the source passage, so that experts can verify the text paragraph by paragraph.
State In productive use in several of our own projects. Proof that the template can be reproduced from a fresh checkout is still pending. Evidence
The Agentic Edition Pipeline is a copyable template for digital editions. It takes digital copies through transcription, review stages and TEI to a static reading and review view that requires no server of its own. Scholarly acceptance rests with the editorial team.
State Research preview without a stable release. A complete run with a current edition project is still pending. Evidence
Digital Humanities Craft is a company that grew out of research in the Digital Humanities, based near Graz. We develop research software and digital editions, teach at universities across Europe and work as a partner in funded projects. Team and projects on dhcraft.org
You show us the workflow on screen with real files. We establish where time is lost and who carries the work.
We record goal, data, limits and test criteria in writing and prepare a quotation with a clearly defined scope.
We build a first working version with your material, which you try out in your daily work.
We test the tool together with you against the agreed criteria. Your experts declare acceptance.
You receive the tool, the source code and the documentation. On request we take on maintenance and further development.
Describe your workflow to us briefly. The following details are helpful:
office@dhcraft.org
Send us what you already have. From it we build a first prototype on which the project can be discussed in concrete terms.
Billing follows the commission, by the hour or as an agreed fee.