MagharibiBERT

MagharibiBERT-v1

MagharibiBERT-v1 is a Transformer-based language model pre-trained for Maghrebi Arabic dialects (Algerian, Moroccan, Tunisian).
It is initialized from MARBERT and further adapted using Masked Language Modeling (MLM) on large-scale social media text.

The model is designed for downstream NLP tasks such as hate speech detection, sentiment analysis, and dialect classification.


Associated Paper

MagharibiBERT-v1 is described in the following manuscript, currently submitted for peer review:

Ahmed Zoubir Messalti and Sadik Bessou,
MagharibiBERT: A Regionally Adapted Language Model for Maghrebi Arabic Social Media,
submitted to the International Journal of Computers and Their Applications (IJCA), 2026.

How to use

Fill-mask pipeline

from transformers import pipeline

pipe = pipeline("fill-mask", model="messalti/MagharibiBERT-v1")
Downloads last month
55
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for messalti/MagharibiBERT-v1

Base model

UBC-NLP/MARBERT
Finetuned
(26)
this model

Datasets used to train messalti/MagharibiBERT-v1