akbargherbal/hadith_alpaca_ft

Hadith Alpaca Dataset

Overview

This repository contains the Hadith Alpaca Dataset, comprising 4,592 meticulously processed Hadiths. The dataset removes the initial chain of transmission in Arabic and extraneous commentary, focusing on the core Hadith text. It's designed for training and evaluating language models, particularly in understanding and processing Islamic religious texts.

Data Format

The dataset is structured for ease of use and clarity. Each Hadith entry is encapsulated within markers, simplifying parsing and extraction.

Markers:

• HADITH_START| - Denotes the beginning of a Hadith entry.
• |HADITH_END - Denotes the end of a Hadith entry.

Fields:

• HADITH_NO: A unique identifier for each Hadith.
• ENGLISH_TEXT: The Hadith text in English.
• ARABIC_TEXT: The Hadith text in Arabic.

Each field is on a separate line after the starting marker.

Examples

Example 1

"""HADITH_START|
HADITH_NO: 1234
ENGLISH_TEXT: Abu Hurayrah reported that the Prophet (ﷺ) said, 'The best charity is that which is practiced by a wealthy person.'
ARABIC_TEXT: عن أبي هريرة، أن رسول الله صلى الله عليه وسلم قال 'أفضل الصدقة ما كان عن ظهر غنى'
|HADITH_END"""

Example 2

"""HADITH_START|
HADITH_NO: 5678
ENGLISH_TEXT: Narrated Anas: The Prophet (ﷺ) said, 'Facilitate things to people and do not make it hard for them and give them good tidings and do not make them run away (from Islam).'
ARABIC_TEXT: عن أنس، عن النبي صلى الله عليه وسلم قال 'يسروا ولا تعسروا، وبشروا ولا تنفروا'
|HADITH_END"""

Use Cases

• Language Modeling: Fine-tuning language models to enhance their comprehension of Hadith texts.
• Text Summarization: Developing models capable of generating concise summaries of Hadiths.
• Machine Translation: Improving machine translation accuracy between English and Arabic for Hadith content.
• Sentiment Analysis: Analyzing the sentiment expressed within Hadiths.
• Question Answering: Building systems capable of retrieving information and answering questions based on Hadith texts.