Scholarlink

<p>This review article asks whether credible progress in computational Hadith studies can be assessed only at the level of model performance, or whether it must instead be judged through evidence traceability, semantic infrastructure, and protocol-complete evaluation. To answer this question, the paper conducts a structured literature mapping and critical synthesis of peer-reviewed computational Hadith research published between 2008 and 2026, with particular emphasis on 2023–2026, across six bibliographic sources. The search yielded 649 records before de-duplication; after screening and eligibility assessment, 73 studies were included in the final review corpus. In addition to synthesizing the primary literature, the paper uses published shared-task results from IslamicEval 2025 as benchmark-conditioned evidence anchors rather than as new experiments. These anchors show that measurable progress is possible in retrieval and quotation-faithfulness tasks, but also that performance remains fragile under protocol-sensitive settings, especially when citation fidelity, correction, and uncertainty handling are required. The review finds three recurring field-wide constraints: fragmented datasets and reporting practices, weak coupling between matn and isnād evaluation, and insufficient semantic infrastructure for provenance-preserving analysis. In response, the paper proposes an integrated taxonomy of matn and isnād methods, defines a minimal auditable evidence packaging scheme for scholar-facing systems, and presents a staged roadmap that prioritizes open semantic infrastructure and robust narrator entity linking/disambiguation as prerequisites for reliable decision-support applications.</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

Charting the computational frontier in hadith sciences: a critical synthesis of matn–isnad methods and semantic infrastructure

  • 期刊论文

  • 发表日期：

    2026年6月16日

查看全文
  • Mohamed Atef Mosa

摘要

This review article asks whether credible progress in computational Hadith studies can be assessed only at the level of model performance, or whether it must instead be judged through evidence traceability, semantic infrastructure, and protocol-complete evaluation. To answer this question, the paper conducts a structured literature mapping and critical synthesis of peer-reviewed computational Hadith research published between 2008 and 2026, with particular emphasis on 2023–2026, across six bibliographic sources. The search yielded 649 records before de-duplication; after screening and eligibility assessment, 73 studies were included in the final review corpus. In addition to synthesizing the primary literature, the paper uses published shared-task results from IslamicEval 2025 as benchmark-conditioned evidence anchors rather than as new experiments. These anchors show that measurable progress is possible in retrieval and quotation-faithfulness tasks, but also that performance remains fragile under protocol-sensitive settings, especially when citation fidelity, correction, and uncertainty handling are required. The review finds three recurring field-wide constraints: fragmented datasets and reporting practices, weak coupling between matn and isnād evaluation, and insufficient semantic infrastructure for provenance-preserving analysis. In response, the paper proposes an integrated taxonomy of matn and isnād methods, defines a minimal auditable evidence packaging scheme for scholar-facing systems, and presents a staged roadmap that prioritizes open semantic infrastructure and robust narrator entity linking/disambiguation as prerequisites for reliable decision-support applications.