Linking Hadith Narrator Identities Across Heterogeneous Arabic Biographical Databases: A Multi-Signal Entity Resolution Pipeline

2026-07-08 · Source: cs.CL updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Digital Humanities · Depth: Advanced, extended

Summary

A two-phase entity resolution pipeline links Hadith narrator identities across three heterogeneous Arabic biographical databases: Sanadset 650K, Hadithtransmitters (hawramani), and Muslimscholars. The pipeline processes 650,986 Hadith records from Sanadset, containing 185,216 unique narrator name variants. Phase 1 matches Sanadset names to hawramani using name-only similarity after domain-specific Arabic normalization, establishing 94,628 links (51.1% of Sanadset narrators), with 39,938 at HIGH confidence. Phase 2 cross-references hawramani (100,915 entries) against muslimscholars (25,247 entries) via a weighted multi-signal function combining name similarity, death-year proximity, and reliability grade polarity. This phase yields 95,573 links (94.7% of hawramani entries), with 18,245 HIGH confidence. The chained links enable the creation of a 185,216-node, 814,093-edge directed transmission graph, enriched with cross-source biographical metadata. All linked corpora and the enriched graph are released as open resources.

Key takeaway

For NLP Engineers or Research Scientists integrating historical Arabic textual data, this pipeline offers a robust method for cross-source entity resolution. You should consider adopting a similar two-phase approach, starting with name-only matching for metadata-sparse sources and progressing to multi-signal linking for richer datasets. Implement domain-specific Arabic normalization to overcome orthographic variations. This will enable you to construct comprehensive knowledge graphs, enriching your datasets with crucial biographical metadata for deeper analysis.

Key insights

A two-phase entity resolution pipeline effectively links Hadith narrator identities across disparate Arabic biographical databases, creating a large enriched transmission graph.

Principles

• Domain-specific Arabic normalization is crucial for historical names.
• Multi-signal linking improves accuracy with diverse metadata.
• Chaining resolution phases enables broader data integration.

Method

A two-phase pipeline: name-only matching (Sanadset to hawramani) followed by multi-signal weighted scoring (hawramani to muslimscholars) using name, death-year, and reliability grade, all after Arabic normalization.

In practice

• Use token-sorted fuzzy matching for name similarity.
• Implement bigram prefix indexing for candidate generation.
• Filter candidates by death-year proximity for common names.

Topics

• Entity Resolution
• Hadith Studies
• Arabic NLP
• Knowledge Graph
• Digital Humanities
• Sanad Analysis

Best for: AI Scientist, NLP Engineer, Research Scientist

Related on AIssential

• Audience Engagement with Arabic Women's Social Empowerment and Wellbeing: A Decadal Corpus — The corpus provides a decadal view of Arabic women's social empowerment discourse on Facebook.
• Training a custom entity linking model with spaCy — Entity Linking resolves ambiguous text mentions to unique knowledge base concepts by leveraging context.
• ATLAS: Article Tracking, Linking, and Analysis of Swedish Encyclopedias — Automated pipelines can effectively restore and link structured knowledge from digitized historical encyclopedias.
• Koshur Diacritizer: A Byte-Level Sequence-to-Sequence Model for Kashmiri Diacritic Restoration — Koshur Diacritizer is a ByT5-small model that restores Kashmiri diacritics using a new 23.7k sentence pair dataset.
• 18th-century Fauna and Flora: from Named Entities to the problems of standardization — Lexical variation in historical specialized texts presents significant challenges for computational processing.
• Text-as-Signal: Quantitative Semantic Scoring with Embeddings, Logprobs, and Noise Reduction — Text can be transformed into operational, continuous semantic data points for AI engineering via embeddings and logprob scoring.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CL updates on arXiv.org.