Step 3 / 5
Harmonization
Use LLM to standardize data where semantic decisions are needed — merge variants, generate missing fields.
Records & Filter page fills fields from deterministic APIs (CrossRef, OpenAlex...).
Harmonization solves places where a semantic decision is needed: "are these two authors the same person?", "are these two keywords the same concept?"
Loading data quality…
Scan finds two things: same person written differently (merge) and same name that are different people (split). Everything needs your approval.
Splits — same name, likely different people
(0)This name splits into field-disjoint groups; each becomes a separate author (suffix a/b/c). Largest group kept plain.
No split suggestions.
Merges — same person, different spellings
(0)Variants of the same person, unified to one canonical form.
No merge suggestions.