CitePrism: AI Citation Auditor for Editorial Integrity
CitePrism is an innovative hybrid framework designed to assist in editorial citation auditing transparently. It integrates LLM-driven contextual analysis, semantic similarity through embeddings, metadata validation, integrity-focused alerts, and analyst reviews involving human oversight. The system identifies citation clusters, enhances reference metadata, calculates combined relevance scores, presents metadata and prompts for self-citation reviews, and allows for customizable threshold-based triage. Initial validation conducted on a case study manuscript containing 104 references showcases its capabilities. This tool tackles the issue that citation auditing at the manuscript level is predominantly manual, fragmented, and challenging to scale, even though editors and reviewers are tasked with ensuring the literature is relevant, accurate, current, and ethically sound.
Key facts
- CitePrism is a hybrid decision-support framework for editorial citation auditing.
- It combines LLM-assisted reasoning, embedding-based similarity, metadata verification, integrity flags, and human review.
- It extracts citation neighborhoods and enriches reference metadata.
- It computes fused relevance scores and surfaces self-citation review prompts.
- It supports configurable threshold-based triage.
- Preliminary validation was done on a single case-study manuscript with 104 references.
- Manuscript-level citation auditing is currently largely manual and fragmented.
- Editors and reviewers must ensure citations are relevant, accurate, current, and ethically appropriate.
Entities
—