Metadata Review & Cleaning

Automatically retrieve external metadata from arXiv, Semantic Scholar, OpenAlex, DBLP, and OpenReview to enrich, standardize, and clean your Zotero library with visual approval diffs.

Metadata Review & Cleaning is a core capability of BibGenie. Papers imported from various academic websites, PDFs, or generic web extractors often suffer from incomplete metadata (missing abstracts, DOIs, page numbers, or journal titles), all-uppercase titles, missing journal abbreviations, unlinked preprint updates, and truncated author lists.

BibGenie pairs large language models' scholarly intelligence with multi-source authoritative academic databases. With a strictly read-only review engine, transparent candidate evidence, and visual diff approvals, you can effortlessly maintain a clean and standardized Zotero library.

100% Read-Only Review & Safe Approval

The review phase is strictly read-only and never silently modifies your library in the background. The AI generates a visual diff preview card comparing current values against proposed changes. Edits are safely written to Zotero only when you explicitly click "Approve".

Key Capabilities

Multi-Source Academic Retrieval

Aggregates arXiv, Semantic Scholar, OpenAlex, DBLP, and OpenReview, intelligently routing based on DOI, arXiv ID, or DBLP keys.

Three Specialized Review Modes

Choose between lint (local formatting), blank (fill missing facts, default), and all (full refresh and sync).

Scholarly Terminology Protection

Unlike rigid regex tools, BibGenie understands academic terms, protecting mRNA, CRISPR-Cas9, PyTorch, and Roman numerals from casing errors.

Visual Diff & Transactional Writes

Visual old-vs-new previews, type-change mapping warnings, Better BibTeX Citation Key retention, and atomic failure rollback.

Supported Academic Data Sources

BibGenie integrates specialized academic discovery channels tailored for different disciplines and publication stages:

Source	Coverage & Highlights	Best For
OpenAlex	250M+ open scholarly graph across all fields	Comprehensive lookup, missing base metadata, citation contexts
arXiv	Physics, Computer Science, Math preprint repositories	Preprint metadata, official version identifiers, published DOI links
DBLP	Computer science bibliography	Top CS conferences/journals, author disambiguation
OpenReview	Machine learning and AI venues (ICLR, NeurIPS, etc.)	Submission track verification, anonymous author protection, camera-ready tracking
Semantic Scholar	High-precision CS and biomedicine corpus	Exact paper identifiers, TLDR summaries, citation analysis
Zotero Translators	Zotero native publisher translators	Direct resolution from publisher landing pages (IEEE, ACM, Springer, Nature)

Three Review Modes

Depending on your workflow needs, you can instruct BibGenie to run under different modes:

Default mode. Queries authoritative academic sources to propose updates only for empty local fields and creators:

Preserves existing non-empty titles, authors, years, and publication titles without overwriting.
Fills missing abstracts, page ranges, DOIs, publication titles, or proceedings titles.
Ideal for the vast majority of everyday library enrichment tasks.

Local standardization mode. Skips external fact retrieval and runs built-in scholarly rules to audit and clean existing metadata:

Sentence case normalization for titles.
Journal abbreviation standardization using LTWA dictionaries.
Precise ISO-639-1 language detection.
Identifies publisher domains misplaced in metadata fields.
Ideal for items that already have complete data and only need typographical standardization.

Full synchronization mode. Queries external sources to compare and propose updates to existing fields, with full author list replacement:

Suited for fixing inaccurate or poorly parsed data from low-quality web imports.
Great for updating a preprint after it has been officially published in a journal or conference.
Note: Do not upgrade to all solely because fields are missing; use blank for missing field completion.

Safety & Ecosystem Compatibility

Better BibTeX Compatibility:
- Many researchers rely on Better BibTeX for LaTeX citation keys. BibGenie's Extra field processor strictly preserves Citation Key:, tex.*, bibtex:, and unstructured notes, never stripping or modifying your citation keys.
Item Type Change Safeguards:
- When proposing an item type conversion (e.g., from preprint to journalArticle), the approval card provides clear warnings of fields that may be mapped or discarded.
Scholarly Intelligence Over Heuristics:
- Rigid casing rules often damage technical acronyms, chemical formulas, or library names. BibGenie's LLM context filters out inappropriate heuristic suggestions.

Example Prompts

Please inspect my currently selected paper and use blank mode to fill in any missing abstract, pages, and DOI.

Help me standardize the title case and journal abbreviations for these selected items while preserving technical terminology.

This paper was previously an arXiv preprint. Please check DBLP or Semantic Scholar to see if it has been published in a peer-reviewed venue, and propose updating the item type and metadata if so.

Find items in this collection that lack an abstract or DOI, and review them to fill in the missing details.

Recommended Workflow

Select Items: Highlight one or more items in the Zotero library, and add them to context using @ or direct instructions.

Read-Only Review: BibGenie retrieves multi-source facts and evaluates rules, generating structured proposals.

Review Visual Diff: A visual Action Preview card appears in the chat, highlighting changes between old and new values.

Approve and Write: Click "Approve" to atomically apply the changes to your Zotero items. If rejected, your library remains completely unchanged.