ECERankings.org

A metrics-based ranking of Electrical & Computer Engineering programs, built on OpenAlex and Crossref.

---

Methodology modeled on CSRankings, adapted for ECE publication patterns (journal-heavy, IEEE/Optica-centric). Primary data source is OpenAlex; Crossref serves as a fallback for conference years not yet indexed by OpenAlex.

Pre-launch. Venue registry, harvesting, and aggregation are functional end to end. The public site has not been built. Methodology and roadmap: PLAN.md.

Methodology

• Adjusted count. Each paper contributes 1.0, divided evenly among its N authors (1/N per author). An institution's credit in an area is the sum of its authors' shares. No citations, no impact factors — publication count at qualifying venues only.
• Overall score. Geometric mean of (1 + adjusted count) across the selected areas — identical formula to CSRankings.
• Venue inclusion. A venue qualifies only if it is highly selective and ≥80% in-scope for its area. Registry: data/areas.json.
• Attribution. Default mode credits each paper to the institution listed at time of publication, not current affiliation. A "current faculty" mode is planned, not yet implemented.
• Institution filter. OpenAlex type=education restricts results to universities; corporate research labs are excluded.

Areas

20 areas (18 on by default), structured to match ECE departmental organization: Circuits & VLSI, Semiconductor Devices, Power & Energy, Communications, Photonics, Control, Robotics, plus optional CS-overlap areas (Architecture, ML, Vision). Full taxonomy and inclusion criteria: PLAN.md.

Pipeline

flowchart LR
    A[OpenAlex harvest] --> D[Aggregate]
    B[Crossref / S2 harvest] --> F[Backfill via DOI] --> D
    B --> C[Normalize affiliations] --> D
    D --> E[(site/data/*.csv)] --> G[Split to per-area JSON]
    D --> H[Build institution metadata]

Loading

Crossref and Semantic Scholar know which venue a paper belongs to but carry sparse affiliations; OpenAlex has the affiliations but leaves most IEEE conference papers linked to no venue at all. The backfill step joins the two on DOI.

Path	Description
PLAN.md	Architecture, methodology, and phased roadmap.
data/areas.json	Venue registry — OpenAlex source IDs, per-year availability, and verification status for every ranked venue.
data/affiliation-map.csv	Mapping from raw Crossref affiliation strings to OpenAlex institution IDs.
data/excluded-works.csv	Curated list of non-research works (editorials, columns) that aggregate.py skips.
data/reports/	One report per data-collection run.
pipeline/	Harvest, backfill, normalization, adjudication, and aggregation scripts. Python standard library only, except normalize_semantic.py.
site/data/	Generated CSVs and JSON consumed by the frontend.
.claude/skills/	Claude Code project integrations.
cache/	Raw API responses. Tracked via Git LFS so it syncs across machines.

Running the pipeline

Requirements: Python 3 (standard library only) and an OpenAlex API key in a .env file at the repo root:

OPENALEX_API_KEY=your-key-here

# 1. Harvest works for a registered venue-year (OpenAlex)
python3 pipeline/harvest.py --venue jssc --year 2023

# 2. For conference years OpenAlex lacks, fall back to Crossref
python3 pipeline/harvest_crossref.py --venue iedm --year 2024

# 3. Recover affiliations for Crossref/S2 works by DOI lookup.
#    Always measure before writing: --pilot probes a random sample read-only.
python3 pipeline/backfill_openalex.py --pilot 500
python3 pipeline/backfill_openalex.py --apply

# 4. Resolve any remaining free-text affiliations to institutions.
#    These are ALTERNATIVES writing the same file — run one, not both.
.venv/bin/python pipeline/normalize_semantic.py --all-cached  # preferred
# or, with no .venv available:
python3 pipeline/normalize_affiliations_local.py --all-cached

# 5. Propose non-research works (editorials, columns) for exclusion.
#    Writes proposals only — promote accepted rows into data/excluded-works.csv,
#    which is the file aggregate.py reads. Nothing is excluded automatically.
.venv/bin/python pipeline/detect_editorials.py --venue scirobotics

# 6. Aggregate into per-(institution, area, year) adjusted counts
python3 pipeline/aggregate.py --all-cached

# 7. Split for the frontend, and build institution metadata
python3 pipeline/split.py
python3 pipeline/build_institutions.py

# 8. Check for regressions
python3 pipeline/verify.py --all

Each script's docstring documents the full option set. Harvests are cached under cache/<venue>/<year>/; re-runs and backfills are idempotent. The backfill checkpoints per batch to cache/.backfill-checkpoint.jsonl, so an interrupted run resumes instead of re-spending the API budget, and records fruitless DOIs in cache/.backfill-negative.txt so they are never re-queried.

data/institutions.json (~27 MB, an OpenAlex institutions dump used by normalize_affiliations_local.py) is gitignored. Not to be confused with site/data/institutions.json, which is the small name+country file the frontend loads and build_institutions.py generates.

Data sources & acknowledgments

• OpenAlex — primary bibliographic source (works, authorships, resolved institutions).
• Crossref — fallback source for conference years not yet linked in OpenAlex.
• CSRankings, by Emery Berger et al. — the methodology this project adapts for ECE.

Contributing

Current priority: venue-registry curation. Entries in data/areas.json are marked todo, candidate, or verified. Review todo/candidate entries for correct source IDs and year coverage; open an issue or PR referencing the venue key.