Scholarlink

<p>In natural language processing (NLP), classifying Arabic text is a basic task that involves assigning specific labels to text documents for various purposes, including topic categorization, sentiment analysis, and spam detection. In this study, we focus on the classification of Hadith, a collection of sayings and actions attributed to the Prophet Muhammad. Two components constitute a hadith: the chain of narrators, also known as Sanad, and the hadith content, also known as Matn. This paper classified Hadith according to their authenticity as sahih (authentic) or daif (weak). The proposed approach combines deep learning and BiLSTM to classify hadith. Moreover, traditional ML approaches such as Support Vector Machine (SVM), Decision Tree (DT), and Naïve Bayes (NB), are utilized as a baseline to evaluate the proposed approach. In addition to using two common datasets for evaluating the proposed approach, this paper compiles and publicly releases the authentication hadith classification dataset (AHCD). This dataset consists of nearly 10,000 Hadith texts collected from various classical Hadith books and contains approximately 7,000 narrators. Different approaches are Applied such as hybrid, deep learning, off-the-shelf LLMs, and machine learning approaches to classify Hadith. The hybrid model was applied separately to Hadith text, Sanad alone , and Matn alone; the same was then performed using deep learning (DL) and machine learning (ML) approaches. The hybrid technique outperformed the DL, Off-the-Shelf LLMs, and ML techniques when applied to (Sanad + Matn), achieving 94.43% accuracy and an F1-score of 92.73%.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel approach for Arabic text classification: Hadith as a case study

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • Badrya Dahy,
  • Majid Asker,
  • Khaled Fathy,
  • Mahamed Mostafa,
  • Mamdouh Farouk

摘要

In natural language processing (NLP), classifying Arabic text is a basic task that involves assigning specific labels to text documents for various purposes, including topic categorization, sentiment analysis, and spam detection. In this study, we focus on the classification of Hadith, a collection of sayings and actions attributed to the Prophet Muhammad. Two components constitute a hadith: the chain of narrators, also known as Sanad, and the hadith content, also known as Matn. This paper classified Hadith according to their authenticity as sahih (authentic) or daif (weak). The proposed approach combines deep learning and BiLSTM to classify hadith. Moreover, traditional ML approaches such as Support Vector Machine (SVM), Decision Tree (DT), and Naïve Bayes (NB), are utilized as a baseline to evaluate the proposed approach. In addition to using two common datasets for evaluating the proposed approach, this paper compiles and publicly releases the authentication hadith classification dataset (AHCD). This dataset consists of nearly 10,000 Hadith texts collected from various classical Hadith books and contains approximately 7,000 narrators. Different approaches are Applied such as hybrid, deep learning, off-the-shelf LLMs, and machine learning approaches to classify Hadith. The hybrid model was applied separately to Hadith text, Sanad alone , and Matn alone; the same was then performed using deep learning (DL) and machine learning (ML) approaches. The hybrid technique outperformed the DL, Off-the-Shelf LLMs, and ML techniques when applied to (Sanad + Matn), achieving 94.43% accuracy and an F1-score of 92.73%.