Step 3 · Harmonization

Harmonization is where BibexPy most strengthens the analytical validity of merged Scopus–WoS data: it resolves author identities, consolidates institution and country names, fills missing metadata from authoritative sources, and quantifies the resulting completeness.

Harmonization screen — Quality dashboard + disambiguation, address-harmonization and enrichment tools — all on the same dataset, fully logged.

The four components

| Tool | What it fixes | Deep dive | | --- | --- | --- | | Author disambiguation | "J. Smith" vs "Smith, John" vs ORCID-confirmed identities | Read more | | Address harmonization | "Wuhan Univ" vs "Wuhan University" vs department-level variants; country spellings | Read more | | Metadata enrichment | Missing DOIs, abstracts, ORCIDs, categories — filled from 7 authoritative sources | Read more | | Quality dashboard | Quantifies completeness with a bibliometrically weighted health score | Read more |

Explainable, controlled, reversible

Every operation runs on the same merged dataset with full, reversible logging.
High-confidence decisions apply automatically; borderline cases route to a review queue where you approve or reject each one.
A snapshot is taken before anything is applied — one click restores the previous state.
Optional LLM assistance (for semantic decisions like merging name variants) is off by default, only touches user-approved cases, and never invents values.

Deterministic by default

Core transformations use rule-based logic and string/context similarity — not predictive ML models. Repeated runs produce identical results.

Suggested order

Run the quality dashboard first to see which fields need attention.
Run enrichment to fill missing metadata from external sources (improves disambiguation inputs too).
Run author disambiguation, reviewing the borderline queue.
Run organization & country harmonization last, then re-check the health score.