HyperAIHyperAI

Command Palette

Search for a command to run...

Jabarti-LLM Arabic-English Bilingual Pretraining Corpus Dataset

Jabarti-LLM Arabic-English Bilingual Pre-training Corpus Dataset

Dataset Overview

The Jabarti LLM Dataset is a bilingual large language model training dataset released by bakrianoo, designed to provide corpus support for the pre-training, domain adaptation, and supervised fine-tuning of small language models in Arabic and English.

The dataset combines curated Egyptian historical corpora with general knowledge texts from multilingual Wikipedia. After cleaning, segmentation, and source organization, it balances general language coverage in Arabic and English with domain-specific knowledge of Egyptian history. The dataset provides both pre-training text corpora and question-answering fine-tuning data, suitable for bilingual language modeling, knowledge learning, and question-answering capability training.

Dataset Composition

  • Pre-training Corpus: Contains continuous text segments in Arabic and English, with each record ranging approximately from 120 to 1,500 characters, with article titles and section headers removed.
  • Egyptian Historical Corpus: Includes curated content related to Egyptian history, allowing for the filtering of historical topic data using fields such as egypt_relevance.
  • Wikipedia Corpus: Integrates general knowledge texts in Arabic and English from CohereLabs/wikipedia-2023-11-embed-multilingual-v3.
  • Question-Answering Fine-tuning Data: Provides Arabic and English question-answering samples, including information on questions, answers, language, question type, and difficulty, supporting supervised fine-tuning.
  • Data Source and Quality Fields: Records source, language, content quality, and topic domain via fields such as ds_source, language, content_quality, and topic_domain, facilitating data filtering and sampling.
  • Data Splits: Provides pre-training and fine-tuning configurations, as well as splits for training, evaluation, and stage divisions compatible with legacy workflows, supporting usage across different training stages.

Build AI with AI

From idea to launch — accelerate your AI development with free AI co-coding, out-of-the-box environment and best price of GPUs.

AI Co-coding
Ready-to-use GPUs
Best Pricing

HyperAI Newsletters

Subscribe to our latest updates
We will deliver the latest updates of the week to your inbox at nine o'clock every Monday morning
Powered by MailChimp