Reference-list detection
Finds the bibliography or references section, even when the PDF has multi-column text, numbering, or wrapped citation lines.
Find the boundaryPDF reference extractor guide
PDF reference extractor features and how-to guide
Use this guide to understand what LumaCite extracts from a research PDF, how the review workspace is organized, and when to resolve or export references.
Before upload
Start from the extractor control panel
The first screen is intentionally simple: choose a PDF, try the example workflow, and check the processing path before the workspace opens.
Core features
What you get from the PDF reference extractor
LumaCite is designed for the practical job after upload: turn a bibliography into records you can verify, correct, and move into the next tool.
Citation row splitting
Separates the reference list into individual records so you can review one citation at a time instead of cleaning copied PDF text manually.
Create review rowsStructured fields
Extracts title, authors, year, journal or source, volume, issue, pages, DOI, PMID, PMCID, arXiv, ISBN, ISSN, and URLs when present.
Fill citation fieldsSource-aware review
Keeps the PDF evidence, extracted record, and editable citation fields in one workspace so review work stays anchored to the document.
Check against sourceMetadata help
Uses Auto-fetch and Resolve for records that need scholarly metadata lookup or a stricter matching pass before export.
Resolve uncertain rowsReview-first exports
Exports selected references to reference managers, citation processors, spreadsheets, Word workflows, or an audit report for handoff.
Export with contextHow to use it
A short workflow from upload to export
The goal is not to click every control. The goal is to know which view answers which question while you move toward a clean export.
Choose a text-based research PDF
Use a PDF where text can be selected. The example below uses Chen et al., Cell 2021, a David R. Liu prime-editing paper with a clear reference list.
Read the extraction summary first
The summary gives the reference count, how many rows are ready, which rows need checking, and whether clean export is available. Start here before editing individual records.
Use the three-pane workspace
The left pane keeps source evidence visible, the middle pane organizes extracted records, and the right pane lets you inspect and edit the selected reference.
Resolve only the rows that need help
Use Auto-fetch for metadata lookup on uncertain records. Use Resolve when a row still needs matching review. This keeps review focused instead of reprocessing clean records.
Export after review
Export selected references when the rows you need are checked. If clean export is blocked, use the audit report to see what still needs attention.
Workspace orientation
Know what each pane is for
The workspace is intentionally dense. It is built for review, not decoration, so each area answers a different question.
- Left: Source PDF
- Use it to compare the reference row with the original document evidence when source preview is available.
- Middle: Extracted records
- Use it as the work queue. Status labels show which rows are verified, need checking, or cannot be verified yet.
- Right: Review reference
- Use it for the selected row: raw extracted text, editable citation fields, identifiers, and metadata notes.
First checkpoint
Use the extraction summary before editing rows
Open this panel first when the workspace loads. It tells you whether the reference set is complete enough to review, resolve, or export.
- Reference count
- Compare it with the paper when you need a complete bibliography.
- Rows to check
- Focus review on records with missing fields, weak identifiers, or metadata differences.
- Export status
- If clean export is blocked, use the audit report to see why.
Export
Choose the format based on where the work continues
The export drawer separates citation style, copy actions, downloadable files, and previews so users can choose the right handoff without hunting through the workspace.
Reference managers
Move reviewed records into Zotero, Mendeley, and many library tools.
Manager handoffEndNote libraries
Use EndNote XML when the next step is a shared EndNote library.
EndNote workflowLaTeX manuscripts
Export BibTeX for Overleaf, LaTeX, and technical writing pipelines.
Technical writingCitation processors
Use CSL-JSON for structured citation processors and downstream tools.
Structured dataSpreadsheet review
Send selected references into QA sheets, screening work, or team review.
Review tableAudit report
Keep warnings, source checks, and export safety notes with the handoff.
Review trailReview cautions
What to review before trusting an export
LumaCite is built to make review faster, but the final citation set should still be checked when a row is uncertain.
Scanned or low-quality PDFs
If the PDF is image-based, OCR quality can affect row splitting and field extraction. Use the source pane and audit notes more carefully.
Rows marked Check details
These rows often need a human look because a field is missing, a title is ambiguous, or metadata sources do not fully agree.
Clean export blocks
When LumaCite blocks a clean export, it is usually protecting the output from count, boundary, duplicate, or metadata risks. The audit report is the right next step.
FAQ
Common questions
What PDFs work best?
Text-based research PDFs with selectable text and a clear reference list work best. Scanned PDFs may need OCR before upload.
What are Auto-fetch and Resolve for?
Auto-fetch looks for scholarly metadata for uncertain rows. Resolve runs a stricter matching review for records that still need help.
Should I export everything?
Export the references you need after reviewing status labels and the summary. For team review, export the audit report alongside clean formats.
Where should I go for DOCX, TXT, or pasted bibliographies?
Use the Text Citation Extractor for copied reference lists, DOCX, TXT, RTF, RIS, BibTeX, CSV, or Markdown input.
Ready to try it
Extract references from a research PDF
Upload a PDF, review the extracted rows, and export only after the records you need are ready.