---
title: "About"
language: en
canonical: https://alman.ai/about/
---

Alman is a simplified dialect of German designed to reduce the complexity of the language while maintaining mutual intelligibility with Standard German. It systematically eliminates several complex grammatical features:

- **Gender System**: Uses a single article form for most contexts
- **Case Markings**: Removes most case-based inflections
- **Noun Declensions**: Simplifies noun endings across cases
- **Job Titles**: Eliminates gender-specific forms in occupational titles

While Alman reduces grammatical complexity, it preserves:

- Core vocabulary
- Basic word order
- Verb conjugations
- Essential syntactic structures

This makes Alman easier to learn while remaining fully comprehensible to Standard German speakers.

## Goals

The main goal of this project is to enshrine and formalize a German dialect that is as easy to learn as English or Spanish, and to eventually get it recognized by the German state.

We are developing under the following constraints:

- Maintain full mutual intelligibility with Standard German
- Preserve the essential character of German syntax
- Systematically simplify grammar without loss of expressiveness

## Language of this website

The specification and related documentation are primarily written in English, because the target audience is non-native speakers of German. The specification is also available in German and in Alman itself, through the language switcher.

## Acknowledgements

Alman was developed with [Hugging Face](https://huggingface.co/). The project uses [Hugging Face compute](https://huggingface.co/docs/hub/jobs) to train and evaluate its language models. The resulting [models and datasets](https://huggingface.co/osolmaz) are published on the Hugging Face Hub, including [GoePT-1-20M](https://huggingface.co/osolmaz/GoePT-1-20M) and [AlmanBench](https://huggingface.co/datasets/osolmaz/almanbench).

---

The Alman specification is open source and can be edited on [GitHub](https://github.com/osolmaz/alman).