raw_text stringclasses 5
values | normalized_text stringclasses 5
values |
|---|---|
L3ezz 3lik a khoya! | laaezz aalik a khoya! |
Tbarkellah 3likom 7bibi | tbarkellah aalikom hbibi |
9ol lya chno smit had l'appli? | qol lya chno smit had l'appli? |
Chokran bzaf 3la had lvideo 5oya | chokran bzaf aala had lvideo khoya |
Mziyan bzzaf, tanmirt a gma | mziyan bzzaf, tanmirt a gma |
Decoding the Maghreb Diaspora: NLP Bridge Framework
This dataset is the first building block of the Bridge-NLP Framework, designed to address the "Linguistic Blind Spot" in current LLMs regarding North African Code-Switching.
📌 Project Context
As described in my Medium White Paper, this project aims to build empathetic educational tools for the Maghreb diaspora in Europe (Spain, France, Catalonia).
🛠 Features
- Raw Text: Authentic samples of Maghreb Arabic (Darija), Tamazight, French, and Spanish mixed in Arabizi and Latin scripts.
- Normalized Text: Processed text using my custom phonetic mapping to reduce tokenization inefficiency.
🚀 Vision
The goal is to expand this into a robust corpus for:
- Linguistic Normalization: Converting Arabizi to standard representations.
- Social Integration Tools: Helping students and families navigate host-country languages through their native hybrid speech.
📬 Contact & Collaboration
Developed by Jamal – Linguistic Engineer & NLP Specialist. I am looking for collaborators and datasets to scale this "Moonshot" for inclusive AI.
- Downloads last month
- 25