The Dimensions Citation Check API reviews a submitted manuscript, identifies which references were authored by the paper's own authors, and assesses each self-citation in context to judge whether it appears justified or gratuitous.
Some self-citation is legitimate, but excessive or unjustified self-citation can inflate metrics, obscure originality, and signal misconduct — and reviewing reference lists manually is slow and inconsistent at scale. The API gives editors a fast, objective way to surface submissions that warrant a closer look, returning a clear editorial rating alongside a detailed report.
Key Features
- Automatic extraction: Upload a manuscript as a PDF and the service identifies all authors and references directly from the document.
- Dimensions-backed detection: References are matched against the Dimensions database to confirm self-citations, then located at the sentence level within the text.
- Contextual assessment: Each self-citation is evaluated in context and labelled appropriate, questionable, or inappropriate — not just counted.
- Editorial classification: Findings are synthesised into a single rating — Low, Medium, High, or Unknown — to support, not replace, editorial judgement.
- Built for integration: A lightweight REST API that fits into existing submission systems, processing up to 10 manuscripts simultaneously.
Use Cases
- Submission screening: Assess every incoming manuscript so editors can prioritise the submissions that warrant closer review before peer review.
- Editorial decision support: Give editors an objective signal — a High rating warrants editor attention, a Low rating gives confidence the submission meets expected norms.
- Integrity screening: Surface patterns of excessive or unjustified self-citation across a journal portfolio.
What You Receive
Each analysis returns an overall editorial rating, the self-citation rate, a per-author breakdown, sentence-level detail with each citation in context, an AI appropriateness assessment, and a formatted HTML report ready to attach to the editorial record. Results are typically returned within a few minutes.
