Create NumidiaNLP organization card

README.md

@@ -1,10 +1,160 @@
----
-title: README
-emoji: 📚
-colorFrom: pink
-colorTo: yellow
-sdk: static
-pinned: false
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+---
+title: NumidiaNLP
+emoji: 🌍
+colorFrom: blue
+colorTo: green
+sdk: static
+pinned: false
+---
+
+<div align="center">
+
+# NumidiaNLP
+
+### Natural Language Processing for Maghrebi Arabic
+
+**Models · Datasets · Evaluation · Reproducible Research**
+
+</div>
+
+---
+
+## About NumidiaNLP
+
+**NumidiaNLP** is a research initiative focused on Natural Language Processing
+for **Maghrebi Arabic and North African social-media language**.
+
+Our work investigates computational representations of the linguistic
+variation observed across **Algeria, Morocco, and Tunisia**, including
+dialectal Arabic, non-standard orthography, Arabizi, and multilingual
+code-switching.
+
+The objective is to develop and evaluate reusable language resources that
+support rigorous and reproducible research on underrepresented Maghrebi
+language varieties.
+
+---
+
+## Research Areas
+
+Our current research focuses on:
+
+- Maghrebi Arabic language modeling
+- Regional language-model adaptation
+- Hate and offensive language detection
+- Sentiment analysis
+- Arabizi and Latin-script Maghrebi Arabic
+- Arabic–French and multilingual code-switching
+- Maghrebi dialect identification
+- Corpus construction and annotation
+- Robust and reproducible NLP evaluation
+
+---
+
+## MagharibiBERT
+
+### A Regionally Adapted Language Model for Maghrebi Arabic
+
+[**MagharibiBERT**](https://huggingface.co/NumidiaNLP/MagharibiBERT)
+is a regional language model developed to provide reusable contextual
+representations for heterogeneous Maghrebi Arabic.
+
+The model follows a continued-pretraining strategy over a large and diverse
+collection of Maghrebi text covering Algerian, Moroccan, and Tunisian
+linguistic variation.
+
+Its training data preserve important characteristics of real-world Maghrebi
+communication, including:
+
+- dialectal Arabic;
+- spelling variation;
+- Arabizi;
+- Arabic–French/English code-switching;
+- informal social-media language.
+
+The objective is not to replace country-specific models, but to investigate
+whether **regional continued pretraining can provide transferable
+representations across Maghrebi countries and NLP tasks**.
+
+➡️ **Model:** https://huggingface.co/NumidiaNLP/MagharibiBERT
+
+---
+
+## Evaluation Philosophy
+
+NumidiaNLP promotes evaluation practices that go beyond reporting a single
+best score.
+
+Our experiments emphasize:
+
+- fixed train/validation/test partitions;
+- multiple random seeds;
+- Macro-F1 as a primary metric for imbalanced classification tasks;
+- validation-based model selection;
+- comparisons with regional and general Arabic language models;
+- statistical analysis where appropriate;
+- computational-cost reporting;
+- transparent reporting of limitations and negative results.
+
+We consider cross-country robustness and stability important criteria when
+evaluating regional Maghrebi language models.
+
+---
+
+## Regional Perspective
+
+Maghrebi Arabic is characterized by substantial variation across countries,
+communities, writing systems, and communication contexts.
+
+For this reason, NumidiaNLP studies the Maghreb not as a single homogeneous
+dialect, but as a **linguistically diverse regional continuum**.
+
+Our research aims to understand when regional representations are beneficial,
+when country-specific models remain preferable, and how both approaches can
+complement each other.
+
+---
+
+## Open and Reproducible Research
+
+Where licensing, ethical requirements, and publication policies permit,
+NumidiaNLP aims to provide:
+
+- pretrained models;
+- evaluation code;
+- benchmark configurations;
+- dataset documentation;
+- reproducible experimental protocols.
+
+Research artifacts are released with appropriate documentation describing
+their intended use and limitations.
+
+---
+
+## Resources
+
+| Resource | Description |
+|---|---|
+| 🤗 [MagharibiBERT](https://huggingface.co/NumidiaNLP/MagharibiBERT) | Regional language model for Maghrebi Arabic |
+| 📚 Models | Pretrained and task-adapted Maghrebi NLP models |
+| 📊 Benchmarks | Reproducible evaluation across Maghrebi NLP tasks |
+| 🗂️ Datasets | Documented linguistic resources released when permitted |
+
+---
+
+## Citation
+
+Scientific publications and corresponding citation information will be linked
+to each released model and dataset through their individual repository cards.
+
+---
+
+<div align="center">
+
+### NumidiaNLP
+
+**Advancing reproducible NLP research for the linguistic diversity of the Maghreb.**
+
+Algeria · Morocco · Tunisia
+
+</div>