v0.1.2 — Typo-tolerant institution search

Institution search now falls back to OpenAlex's native Levenshtein fuzzy matching (~2 per word) when an exact query returns zero results. Misspelled queries like "IIT HYderbaad" or "Oxferd University" now resolve correctly. Well-spelled queries still hit the exact path first to preserve ranking.

Per OpenAlex search docs.

v0.1.1 — Institutional Research Visibility Analysis

Institutional Research Visibility Analysis

New /analysis/institution page measures what publishers deposit to Crossref for a given institution's journal articles over the last 90 days — observed directly from Crossref work records, never projected from publisher-wide averages. Designed to survive scrutiny in real publisher-agreement negotiations.

How it works

The analysis combines three data sources:

• OpenAlex — identifies the institution's journal articles (last 90 days) grouped by publisher
• Crossref — batch-fetches each work record and inspects metadata directly (affiliation, any-ROR, this-institution's-ROR, funder, abstract, license, ORCID)
• ROR — the machine-readable bridge; the Inst. ROR column is the strict institutional-attribution test

Scope and 90-day window chosen to respect the free rate limits offered by Crossref and OpenAlex.

Safeguards against misleading numbers

• Fixed an ancestor double-counting bug in the grouping (uses singular host_organization, not host_organization_lineage)
• PUBLISHER_MAP expanded to ~55 entries covering Springer/IEEE/T&F/WK/Elsevier sub-entities and previously-unmapped publishers (De Gruyter, Emerald, Hindawi, Karger, Thieme, Brill, eLife, NEJM, Wellcome, AAAS, JMIR, ACM, AIAA, SPIE, more)
• Removed a fabricated IEEE OpenAlex ID (P4310320430 didn't exist)
• Publishers with <10 articles at an institution are shown but not measured
• Journal articles only — no proceedings, peer reviews, or book chapters
• Individual Crossref batch failures are tolerated, not fatal

Categorized "Why N articles weren't analyzed"

Unanalyzable articles are surfaced in three buckets, every one framed as a publisher deposit gap: no Crossref DOI registered; publisher deposits but isn't mapped yet; deposit too incomplete for OpenAlex to identify the publisher.

Cost calculator

Floating corner panel with a 15-currency dropdown. Plug in your own hourly rates; the observed 90-day gap is extrapolated to an annual cost using the institution's real 1-year article count. Rates stay client-side.

Methodology cards

Three collapsible cards explain (1) how the analysis works, (2) caveats to keep in mind (amber-styled warning card — including the ROR funder-vs-institution distinction, connecting to open issue #2 FundRef→ROR transition), and (3) what each gap column means.

API / CLI

GET /api/analyze-institution?ror=<ror>&days=<days>
GET /api/analyze-institution?search=<name>

pnpm --filter web analyze:institution -- --ror 052gg0110
pnpm --filter web analyze:institution -- --search "IISc" --days 90

Vercel Analytics

Each analyzed institution gets a distinct URL (/analysis/institution?ror=<ror>) via history.pushState, so per-institution pageviews show up separately in analytics.

Live

https://nexus-score.vercel.app/analysis/institution

PR #9 — #9

v0.1.0 — Research Nexus Score

Research Nexus Score v0.1.0

The first public release of Research Nexus Score — an open-source tool for evaluating publisher metadata quality across Crossref members.

Pre-1.0 note: The scoring methodology is still evolving in public. Weights, dimensions, and signals may change as community feedback shapes the model (see open issues tagged methodology). 1.0.0 will be cut when the methodology and Publisher API are stable commitments.

Highlights

• Publisher Leaderboard: Rankings for 27,830+ publishers based on Crossref metadata coverage
• Composite Scoring: Single score (0–100) across 5 dimensions — Provenance, People, Organizations, Funding, Access
• Trend Analysis: Current vs backfile era comparisons
• Per-Content-Type Breakdowns: Filter by journal-article, book-chapter, proceedings, and 11 more
• Per-Dimension Leaderboards: Sort by any single dimension; radar chart overlays reveal publisher "shape"
• Actionable Recommendations: Every score comes with improvement paths linked to Crossref documentation
• Gap Fixer: Recover missing metadata from Crossref Participation Reports using OpenAlex, ORCID, ROR, and Reducto
• Journal Nexus: Article-level enrichment and inspection
• MCP Server: npx @nexus-score/mcp-server for AI assistant integration
• Core Library: @nexus-score/core for programmatic scoring

Why content-type and per-dimension views matter

Aggregate scores were misleading. eLife scored 31/F in aggregate but 97/A on journal articles — peer reviews were the diluter. With the content-type filter, eLife jumps from #2,581 to #2. APS scores 81/A on current journal articles but 7/F on proceedings. Same publisher, two completely different pipeline investments.

Two publishers can both score 50 and look nothing alike. Per-dimension sorting and radar profiles make those differences visible.

Metadata quality is driven by the deposit pipeline, not by discipline.

Community-driven

Features in this release were shaped by feedback from researchers, publishers, librarians, and infrastructure folks across LinkedIn and GitHub. Key contributors to the content-type filter (PR #7): Fiona Hutton (eLife), Bianca Kramer, Toby Green (Coherent Digital), Andy Byers (OLH), Colin Adcock (APS), Wendy Patterson (Beilstein), Paula Kennedy (ULP), Renisha Winston (i-manager), Kora Korzec (Crossref). Per-dimension leaderboards (PR #8): Bianca Kramer.

Tech Stack

• Next.js 16, React 19, TypeScript, Turborepo, Tailwind CSS 4
• Crossref REST API, OpenAlex, ORCID, ROR, Unpaywall, Reducto

Links

• Live: https://nexus-score.vercel.app
• Methodology: INSIGHTS.md
• License: AGPL-3.0