Project · language, culture and AI

Conteranto

Translation that understands culture, not just words.

A project about what gets lost

Meaning is more than words.
Tone is meaning.

Translation is not a neutral transfer. It is a design choice about how one person reaches another.

76 languages · politeness · formality · intimacy

The problem

Most translation systems aim for linguistic accuracy. They can get the words right while missing the relationship carried by those words. A request may become too blunt, too formal, too distant or unintentionally disrespectful when it crosses a cultural boundary.

Conteranto starts from a simple distinction: translation without tone is not communication. It is noise. What sounds neutral in one language may sound aggressive in another; a warm invitation can become cold; a careful Dutch request can become overly formal in Japanese.

What Conteranto adds

Conteranto introduces a cultural-context layer that makes the pragmatic dimensions of translation explicit and adjustable:

  • Politeness: how much respect or deference the situation calls for.
  • Formality: how the register moves from everyday speech to professional or academic language.
  • Intimacy: how warm, personal or distant the relationship should feel.

The system can initialize sensible defaults from language and context, while still allowing the user to see and change the assumptions. That combination of automatic help, user control and explanation is the important design choice.

Why it belongs in this collection

The essays ask how architectures shape what people can see, say and decide. Conteranto applies that question to the smallest unit of public life: a sentence addressed to another person.

It makes visible that language is never only a container for information. It also distributes respect, distance, authority and belonging. A translation system therefore participates in the architecture of a relationship. If its defaults are invisible, cultural misunderstanding is quietly designed into the interaction.

Conteranto is an applied example of the collection’s thesis: context is not decoration added after meaning. Context is part of meaning.

From reflection to research

The project connects an application with a research programme in explainable NLP and cross-cultural communication. Its central question is practical and empirical: how do speakers of different languages actually want to sound, and how can a translation system explain the choices it makes on their behalf?

That makes Conteranto a useful bridge between the philosophical essays and the work of building systems. It turns cultural nuance into parameters that can be questioned, compared and improved rather than leaving it hidden inside a single “natural” output.

Essays that opened the project

The project’s philosophical path can be followed through Prelude to a Thinking Future, The Luminous Weave and The Intelligence That Refuses to Be One. Together they move from AI as mirror, through dialogue between minds, toward the need for plurality and contextual judgment.

Next steps

  1. Keep cultural assumptions visible to the person using the translation.
  2. Collect feedback on how different communities want to sound in real situations.
  3. Compare model outputs and human preferences across languages and contexts.
  4. Make the system useful for ordinary communication while keeping its research questions open.

Conteranto is being developed with research and implementation partners. The current application is available at app.conteranto.com; the project site contains the methodology, research links and contact details.

Related project

AI & plurality

Experiments with multiple models and shared understanding.

Back to projects

All projects

Return to the project overview on sparla.com.