Why HistoriaMP came into being

The starting point was both personal and fundamental: the wish to understand a historical primary source for oneself - and the experience of how quickly that wish reaches its limits.

Anyone looking at a medieval or early modern manuscript does not simply see “text”. One sees letter forms, abbreviations, damaged places, unfamiliar languages, unusual signs, later additions, gaps and traces that are difficult to interpret. The source is visible, but it is not automatically accessible.

Out of this experience came the central question:

How can someone check a historical source if they cannot read it securely - and at the same time do not want to trust someone else’s reading blindly?

The first hurdle: secondary sources

The obvious path first leads to editions, translations, commentaries and research literature.

These works are important. Many of them are careful, learned and still indispensable. But they are not the source itself. They contain decisions: readings, additions, smoothing, translations, interpretations and sometimes assumptions that are then taken over again and again.

Especially with historical standard works, this can become a problem. Some foundations are cited for decades, sometimes for more than a century. Their importance does not automatically make them wrong. But their authority can lead later work to build repeatedly on the same readings instead of consistently returning to the visible source evidence.

HistoriaMP grew out of dissatisfaction with precisely this situation: not because older scholarship is worthless, but because every reading should remain checkable.

The second hurdle: the primary source itself

The wish to return to the source does not automatically solve the problem.

The primary source is not easy to access. Anyone who cannot recognise the script, does not know the language or cannot understand the abbreviations cannot simply verify what is written there.

Digital images alone are not enough either. A high-resolution manuscript image is valuable, but it does not explain itself. One sees the source - but one does not yet know which signs are securely visible, which places remain uncertain, which abbreviations are possible and where a reading already becomes interpretation.

This is exactly where the gap appears:

One wants to return to the primary source, but cannot evaluate it securely without scholarly and technical support.

The third hurdle: AI and existing HTR systems

The next obvious attempt today is often AI.

One uploads an image, asks a model for the text - and receives an answer. Sometimes it sounds convincing. That is precisely what makes it dangerous. Fluent text is not automatically a correct or justified reading.

Established HTR systems can also be very strong for many tasks. But they often require prior knowledge: What script is present? Which language? Which model fits? Are suitable training data available? Does a separate model need to be trained? How should the output be evaluated? Where are the uncertainties?

For users without specialist palaeographical training, this creates another access barrier. To use the right tool, one already has to know much of what one is actually trying to find out.

Anyone who cannot assess script, language and model limits may receive results - but often lacks a reliable way to check those results.

The basic idea of HistoriaMP

HistoriaMP emerged from these experiences.

The project does not try to replace scholarly work. Nor does it try to generate historical truth automatically.

HistoriaMP is intended to make the path from the visible source to a justified reading traceable.

The platform does not only ask:

What might be written there?

It also asks:

A reading should not merely be output. It should be justified, checked and documented together with its uncertainty.

Why HistoriaMP works differently

HistoriaMP is not a classical OCR or HTR application. The goal is not:

Image in, text out.

A historical reading should not simply appear as finished text. It should emerge step by step from visible findings: from image areas, lines, letter forms, abbreviations, uncertainties, variants and checks.

The text is therefore not the beginning, but the result of a traceable path.

Users should not only see a result. They should be able to understand what it rests on - and where uncertainty remains.

Who HistoriaMP is for

HistoriaMP is for people who take historical sources seriously.

This includes scholars, archives, Digital Humanities projects and edition projects. The idea is equally directed at historically interested users who do not have specialist palaeographical training and still want to know what a source probably says.

For these users in particular, one thing is decisive: the system must not create false certainty. It must not pretend that every place is clearly legible. It should explain, mark, justify and warn.

Not every user has to train a model, classify a script or identify an abbreviation sign. But every user should be able to recognise whether a reading is well justified or whether uncertainty remains.

What HistoriaMP is not

HistoriaMP is not an automatic truth machine.

It is not a replacement for scholarly source work.

It is not an interface that turns a manuscript image into an apparently finished text while hiding the open questions behind it.

HistoriaMP is a research workbench. It is meant to show how a reading comes into being, which findings support it, which alternatives are possible and which places must remain open.