Research Nexus Score

Measure how well your metadata contributes to Crossref's Research Nexus vision.

Live Demo • Features • Quick Start • Methodology • Insights • Architecture • Roadmap • Contributing • Citation

Research Nexus Score evaluates metadata coverage across five dimensions — Provenance, People, Organizations, Funding, and Access — giving publishers a composite score (0-100) with actionable recommendations for improvement.

Built to support Crossref's Research Nexus initiative and aligned with the Barcelona Declaration on Open Research Information.

Features

Publisher Leaderboard: Rankings for 27,830+ publishers based on metadata coverage
Composite Scoring: Single score (0-100) that captures overall metadata contribution
Dimension Breakdown: Identify strengths and weaknesses across 5 key areas
Trend Analysis: Compare current metadata practices vs historical (backfile)
Actionable Recommendations: Improvement suggestions with links to Crossref documentation
Global Rankings: See where any publisher stands among all Crossref members
Gap Fixer: Recover missing metadata from Crossref Participation Reports using open data sources
Journal Nexus: Journal-level deep analysis — article-by-article metadata coverage, OpenAlex reconciliation, PDF extraction, and metadata trend tracking
Institutional Analysis: Institution-level view of publisher deposit quality — reconciles an institution's output (via OpenAlex) against what reached Crossref, surfacing per-publisher deposit gaps and unmapped publishers
MCP Server: Integrate with Claude Desktop or other AI assistants
Core Library: Use scoring logic in your own applications

Quick Start

Web Interface

Visit nexus-score.vercel.app to search for any Crossref member and view their score.

MCP Server (for Claude Desktop)

npx @nexus-score/mcp-server

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "nexus-score": {
      "command": "npx",
      "args": ["@nexus-score/mcp-server"],
      "env": {
        "CROSSREF_MAILTO": "your-email@example.com"
      }
    }
  }
}

As a Library

pnpm add @nexus-score/core

import { CrossrefClient, calculateMemberScore } from '@nexus-score/core';

const client = new CrossrefClient({ mailto: 'your-email@example.com' });
const member = await client.getMember('286'); // Oxford University Press
const score = calculateMemberScore(member);

console.log(score.total);  // 72
console.log(score.grade);  // 'B'
console.log(score.dimensions.provenance.score);  // 18.5
console.log(score.recommendations[0].title);     // 'Increase ORCID Coverage'

Scoring Methodology

Why this matters in the AI era

Every AI research tool — Consensus, Elicit, Semantic Scholar, OpenAlex, ChatGPT and Claude with search — reads the scholarly record through deposited metadata. A paper isn't "discoverable" in the abstract; it's discoverable through the specific fields a publisher chose to deposit. Missing fields aren't decoration gaps — they're the paper going silent in the one place AI looks.

The five dimensions below aren't arbitrary. Each maps to something an AI system needs to do its job.

The score measures what publishers deposit, not what exists. Most gaps are pipeline problems: the data is upstream — in manuscripts, submission systems, PDFs — but doesn't make it into the deposit. That makes these gaps fixable, not structural. The three-layer architecture (Score → Recommend → Gap Fixer) is built around that fact.

Not an Impact Factor. Nexus Score is size-independent. A new journal can score A on day one. A three-person university press can outscore Elsevier. What you deposit is what you're judged on — not how much you publish, not how old you are, not who cites you.

Dimensions (100 points total)

Dimension	Points	What It Measures	Why AI needs it	In plain English
Provenance	25	References (15), Update Policies (5), Similarity Check (5)	Trust and traceability of the claim	Where did this paper come from, and can we trust the trail? Did the publisher tell us when it was published, what version this is, what license it's under, and what it cites? Basically — is the paper's paperwork in order.
People	20	ORCID iD Coverage (20)	Unambiguous author attribution	Do we actually know who wrote it? Are the authors real, identified humans with ORCIDs — or just names on a page that could belong to anyone? If two researchers share a name, can we tell them apart?
Organizations	15	Affiliations (5), ROR IDs (10)	Machine-readable institutional links	Do we know where the authors work? Is the university or institution properly identified with a ROR ID, or is it a free-text string like "Dept of Bio, Univ" that no machine can match to anything?
Funding	20	Funder Registry IDs (10), Award Numbers (10)	Investment traceability for funders	Who paid for this research, and can we follow the money? Is the funder identified with a registry ID? Is the grant number there? Without this, you can't answer basic questions like "what did the NIH's $40B actually produce?"
Access	20	Licenses (7), Full-text Links (7), Abstracts (6)	Whether AI can legally read and ingest the work	Can anyone actually read it? Is the full text open, or paywalled? Is there a license that tells AI tools whether they're allowed to use it? If a paper exists but no one can access it, it may as well not exist for AI discovery.

Grading Scale

Grade	Score Range	Description
A	80-100	Excellent metadata coverage
B	65-79	Good coverage with room for improvement
C	50-64	Adequate coverage but with significant gaps
D	35-49	Needs substantial work across multiple dimensions
F	0-34	Poor metadata coverage requiring attention

Data Source

Scores use pre-computed coverage statistics from the Crossref /members API. These statistics are calculated daily by Crossref and represent the percentage of works containing each metadata element.

Current: Works published in the last 2 calendar years
Backfile: Older works in the archive

Project Structure

nexus-score/
├── apps/
│   ├── web/                  # Next.js 16 web application
│   │   ├── src/
│   │   │   ├── app/
│   │   │   │   ├── leaderboard/        # 27,830-publisher rankings + filters
│   │   │   │   ├── member/[id]/        # Per-publisher score card + radar
│   │   │   │   ├── analysis/
│   │   │   │   │   └── institution/    # Institutional deposit-gap analytics (v0.1.1)
│   │   │   │   └── api/                # Search, member, leaderboard, analyze-institution
│   │   │   └── components/             # Radar, score card, gap tables, blind spots
│   │   ├── scripts/          # Leaderboard generation scripts
│   │   └── data/             # Cached leaderboard data (27,830 publishers)
│   ├── gap-fixer/            # Metadata recovery tool
│   │   ├── src/
│   │   │   ├── lib/
│   │   │   │   ├── enrichers/  # OpenAlex, ORCID, ROR clients
│   │   │   │   ├── parsers/    # Gap report CSV parser
│   │   │   │   └── scoring/    # Confidence scoring
│   │   │   └── components/   # Upload & analysis UI
│   │   └── README.md
│   └── journal-nexus/        # Journal-level deep analysis
│       └── src/
│           ├── app/
│           │   └── api/      # Enrich, trends, PDF extraction endpoints
│           └── components/   # Score card, trends, reconciliation, PDF modal
├── packages/
│   ├── core/                 # Scoring library (@nexus-score/core)
│   │   ├── src/
│   │   │   ├── crossref/     # Crossref API client
│   │   │   └── scoring/      # Score calculation logic
│   │   └── package.json
│   └── mcp-server/           # MCP server (@nexus-score/mcp-server)
│       └── src/
├── package.json              # Root workspace config
├── turbo.json                # Turborepo configuration
└── pnpm-workspace.yaml       # pnpm workspace definition

Development

Prerequisites

Node.js 18+
pnpm 10+

Setup

# Clone the repository
git clone https://github.com/aadivar/nexus-score.git
cd nexus-score

# Install dependencies
pnpm install

# Build all packages
pnpm build

# Start development server
pnpm dev

Environment Variables

Create a .env.local file in apps/web/:

CROSSREF_MAILTO=your-email@example.com

Using your email enables access to Crossref's polite pool for better rate limits.

Available Scripts

Command	Description
`pnpm dev`	Start development servers
`pnpm build`	Build all packages and apps
`pnpm lint`	Run ESLint across all packages
`pnpm test`	Run tests
`pnpm mcp`	Start MCP server in development mode

Regenerating the Leaderboard

The leaderboard data is pre-computed from all 31,000+ Crossref members:

cd apps/web
pnpm generate-leaderboard

This fetches all members, calculates scores, and saves to data/leaderboard.json.

Automated Updates: A GitHub Actions workflow runs biweekly (1st and 15th of each month) to automatically update the leaderboard data. You can also trigger it manually from the Actions tab.

Why Research Nexus Score?

The Problem

Publishers register metadata with Crossref, but there's no easy way to understand:

How complete is my metadata compared to peers?
Which areas need the most improvement?
Am I getting better or worse over time?

The Solution

Research Nexus Score provides:

Visibility: See exactly where you stand among 27,830+ publishers
Actionability: Get specific recommendations with documentation links
Benchmarking: Compare against industry leaders and peers
Trends: Track improvement over time (current vs backfile)

Barcelona Declaration Alignment

Research Nexus Score supports the Barcelona Declaration on Open Research Information by:

Making metadata coverage visible and comparable
Encouraging adoption of persistent identifiers (ORCID, ROR)
Promoting transparency in funding acknowledgements
Supporting FAIR principles for metadata

Journal Nexus

Journal Nexus goes deeper than the leaderboard — it analyzes a journal article-by-article to show exactly where metadata gaps are, what's recoverable, and how quality trends over time.

How It Works

Search for any journal by ISSN or title
Score — see the journal's Nexus Score with dimension breakdown and recommendations
Trends — metadata coverage by month, broken down by content type, with automated insights
Article Analysis — every article checked against Crossref, then reconciled with OpenAlex to show what's missing vs what's recoverable
PDF Extraction — for articles with the biggest gaps, extract metadata directly from the PDF (authors, ORCIDs, affiliations, funding, references)
Impact Summary — exportable report showing recovery potential and before/after score projections

Key Finding

Most metadata gaps are pipeline problems, not content problems. The data exists upstream — in submission systems, in OpenAlex, in the PDFs themselves — it just doesn't make it into the Crossref deposit. Journal Nexus makes that visible at the article level.

Gap Fixer

Once you know what metadata is missing, Gap Fixer helps you recover it.

How It Works

Upload your Crossref Participation Report gap CSV
Enrich each DOI using OpenAlex, ORCID, ROR APIs + Reducto PDF extraction
Score recovered data with multi-source confidence levels
Export high-confidence recoveries formatted for Crossref submission

Supported Recovery

Gap Type	Sources	Confidence
Abstracts	OpenAlex, Reducto	Up to 95%
References	OpenAlex, Reducto	Up to 95%
ORCID iDs	OpenAlex, ORCID, Reducto	Up to 100%
Affiliations	OpenAlex, Reducto	Up to 95%
ROR IDs	OpenAlex, ROR	Up to 95%
Funder IDs	OpenAlex, Reducto	Up to 95%
Award Numbers	OpenAlex, Reducto	Up to 95%
Licenses	OpenAlex, Reducto	Up to 95%

See apps/gap-fixer/README.md for detailed documentation.

Architecture & Vision

graph TB
    subgraph DATA ["Data Sources"]
        CR["Crossref API<br/><i>31,000+ members</i>"]
        OA["OpenAlex"]
        ORCID["ORCID API"]
        ROR["ROR API"]
        PDF["PDF Extraction<br/><i>Reducto / AI backends</i>"]
    end

    subgraph CORE ["@nexus-score/core"]
        SCORE["Scoring Engine<br/><i>5 dimensions, 100 pts</i>"]
        GRADE["Grading & Recommendations"]
    end

    subgraph APPS ["Applications"]
        WEB["Web App<br/><i>Leaderboard · Insights · Content-Type Filters</i>"]
        JN["Journal Nexus<br/><i>Article-level analysis · Trends · PDF extraction</i>"]
        GF["Gap Fixer<br/><i>Upload gap CSV → recover metadata</i>"]
        MCP["MCP Server<br/><i>AI assistant integration</i>"]
    end

    subgraph ENRICHERS ["Pluggable Enrichers 🔌"]
        direction LR
        E1["OpenAlex<br/>Enricher"]
        E2["ORCID<br/>Enricher"]
        E3["ROR<br/>Enricher"]
        E4["PDF<br/>Enricher"]
        E5["Your Own<br/>Enricher"]
    end

    subgraph PLANNED ["Planned"]
        API["Publisher API<br/><i>REST access to scores</i>"]
        BATCH["Batch Recovery<br/><i>Bulk Crossref-ready exports</i>"]
        TREND["Trend Tracking<br/><i>Score history over time</i>"]
        BENCH["Community Benchmarks<br/><i>Peer group comparisons</i>"]
    end

    CR --> SCORE
    SCORE --> GRADE
    GRADE --> WEB
    GRADE --> JN
    GRADE --> MCP
    GRADE --> API

    CR -->|"Gap Reports"| GF
    CR -->|"Per-journal works"| JN
    OA -->|"Reconciliation"| JN
    PDF -->|"Full-text extraction"| JN
    GF --> ENRICHERS
    OA --> E1
    ORCID --> E2
    ROR --> E3
    PDF --> E4
    ENRICHERS -->|"Recovered metadata"| GF

    SCORE --> TREND
    WEB --> BENCH
    GF --> BATCH

    style DATA fill:#e8f4f8,stroke:#2980b9,color:#000
    style CORE fill:#eafaf1,stroke:#27ae60,color:#000
    style APPS fill:#fef9e7,stroke:#f39c12,color:#000
    style ENRICHERS fill:#f4ecf7,stroke:#8e44ad,color:#000
    style PLANNED fill:#fbeee6,stroke:#e67e22,color:#000
    style E5 stroke-dasharray: 5 5
    style API stroke-dasharray: 5 5
    style BATCH stroke-dasharray: 5 5
    style TREND stroke-dasharray: 5 5
    style BENCH stroke-dasharray: 5 5

Solid boxes = shipped. Dashed boxes = planned. The pluggable enricher layer (purple) is the key extensibility point — anyone can add their own metadata source.

Roadmap

Phase	What	Status
Scoring	Publisher leaderboard with 27,830+ members, composite scoring, grading	Done
Gap Fixer	Recover missing metadata from OpenAlex, ORCID, ROR, and PDF extraction	Done
Journal Nexus	Journal-level article analysis — article-by-article metadata coverage, OpenAlex reconciliation, metadata trends, and PDF full-text extraction	In progress — evaluating with publishers
Content-Type Filtering	Filter leaderboard and insights by content type (journal-article, book-chapter, etc.)	Done
Pluggable Enrichers	Modular metadata recovery — plug in any source (OpenAlex, ORCID, ROR, PDF extraction, or your own) to fill gaps your way	In progress
Publisher API	REST API for programmatic access to scores and gap reports	Planned
Batch Recovery	Bulk metadata recovery with Crossref-ready export files	Planned
Trend Tracking	Historical score tracking — see improvement over time per publisher	Planned
Community Benchmarks	Peer group comparisons by size, discipline, and region	Planned

Pluggable Architecture

Gap Fixer recovers missing metadata (ORCIDs, funders, affiliations, abstracts, references) by pulling from multiple sources — OpenAlex, ORCID, ROR, and PDF extraction. Each source is an independent enricher module. Publishers and infrastructure providers can plug in their own sources or swap extraction backends to fit their workflow — no lock-in to any single provider.

Built-in enrichers:

OpenAlex — ORCIDs, ROR IDs, affiliations, references, abstracts, funders
Reducto — AI-powered structured extraction from scholarly PDFs (abstracts, authors, affiliations, funding, references)

In progress:

ORCID API — Author identity validation (enricher built, integration in progress)
ROR API — Organization identifier matching (enricher built, integration in progress)

On the radar:

DeepSeek, Google Gemini, AWS Textract, Azure Document Intelligence, Mistral, LlamaParse

The goal: metadata gaps are a pipeline problem, not a discipline problem. With pluggable enrichers, anyone can recover what's missing using the sources that work best for them. Everything is open source and AGPL-3.0-licensed — contributions and sponsors welcome.

Have ideas? Open an issue or jump into the conversation on LinkedIn.

Tech Stack

Framework: Next.js 16 with App Router
Styling: Tailwind CSS
Monorepo: Turborepo + pnpm
Language: TypeScript
MCP: Model Context Protocol SDK
Data: Crossref REST API

Contributing

We welcome contributions! Please see our Contributing Guide for details.

Quick Contribution Ideas

Add new scoring dimensions or metrics
Improve the UI/UX of the web application
Add more MCP tools for AI integrations
Write documentation or tutorials
Report bugs or suggest features

License

This project is licensed under the AGPL-3.0 License - see the LICENSE file for details.

Acknowledgments

Crossref for the REST API and metadata standards
Model Context Protocol for the MCP SDK
The scholarly communication community for feedback and inspiration

Citation

If you use or mention Research Nexus Score in your work, please cite it as:

@software{nexus_score,
  author       = {Varma D., Aadinarayana},
  title        = {Research Nexus Score: Metadata Coverage Scoring for Crossref Members},
  year         = {2025},
  doi          = {10.5281/zenodo.19217245},
  url          = {https://doi.org/10.5281/zenodo.19217245},
  note         = {Open-source tool for evaluating publisher metadata quality}
}

Or in text:

Varma D., A. (2025). Research Nexus Score: Metadata Coverage Scoring for Crossref Members. https://doi.org/10.5281/zenodo.19217245

Author

Aadi Narayana Varma

LinkedIn: @aadi-narayana-varma-dantuluri
GitHub: @aadivar
Email: varma2friend@gmail.com

_{Built with care for the open research community}

FilesExpand file tree

README.md

Latest commit

History

README.md

File metadata and controls

Research Nexus Score

Features

Quick Start

Web Interface

MCP Server (for Claude Desktop)

As a Library

Scoring Methodology

Why this matters in the AI era

Dimensions (100 points total)

Grading Scale

Data Source

Project Structure

Development

Prerequisites

Setup

Environment Variables

Available Scripts

Regenerating the Leaderboard

Why Research Nexus Score?

The Problem

The Solution

Barcelona Declaration Alignment

Journal Nexus

How It Works

Key Finding

Gap Fixer

How It Works

Supported Recovery

Architecture & Vision

Roadmap

Pluggable Architecture

Tech Stack

Contributing

Quick Contribution Ideas

License

Acknowledgments

Citation

Author