README.md

metadata

language: ar
widget:
  - text: للوقايه من عدم انتشار [MASK]

arabert_c19: An Arabert model pretrained on 1.5 million COVID-19 multi-dialect Arabic tweets

ARABERT COVID-19 is a pretrained (fine-tuned) version of the AraBERT v2 model (https://huggingface.co/aubmindlab/bert-base-arabertv02). The pretraining was done using 1.5 million multi-dialect Arabic tweets regarding the COVID-19 pandemic from the “Large Arabic Twitter Dataset on COVID-19” (https://arxiv.org/abs/2004.04315). The model can achieve better results for the tasks that deal with multi-dialect Arabic tweets in relation to the COVID-19 pandemic.

Classification results for multiple fake-news detection tasks with and without using the arabert_c19:

For more details refer to the paper (link)

	\multicolumn{5}{c}{Without Fine-tuning}	\multicolumn{5}{c}{With Fine-tuning}
	\multicolumn{3}{c}{Baseline models}	\multicolumn{2}{c}{Pretrained Covid-19 models}
	arabert	mbert
\textbf{Contains hate}	0.8346	0.6675
\textbf{Talk about a cure}	0.8193	0.7406
\textbf{Give advice}	0.8287	0.6865
\textbf{Rise moral }	0.8398	0.7075
\textbf{News or opinion }	0.8987	0.8332
\textbf{Dialect}	0.7533	0.558
\textbf{Blame and negative speech}	0.7426	0.597
\textbf{Factual}	0.9217	0.8427
\textbf{Worth fact-checking}	0.7731	0.5298
\textbf{Contains fake information}	0.6415	0.5428

Preprocessing

from arabert.preprocess import ArabertPreprocessor
model_name="moha/arabert_c19"
arabert_prep = ArabertPreprocessor(model_name=model_name)
text = "للوقايه من عدم انتشار كورونا عليك اولا غسل اليدين بالماء والصابون وتكون عملية الغسل دقيقه تشمل راحة اليد الأصابع التركيز على الإبهام"
arabert_prep.preprocess(text)

Contacts

Hadj Ameur: Github | [email protected] | [email protected]