Removing duplicates

Two kinds of match

A scan only lists pairs. Nothing is merged until you confirm it.

Exact matches share an identifier (DOI, PubMed ID or PMC ID) and are the same work. You can merge them one by one, or all together with one button that states how many pairs it will merge.

Other matches are only suggestions. Similar titles, matching authors and year, or the same journal, volume and pages often mean a duplicate, but they also catch corrections, a preprint and its published version, conference versions and two-part studies. Each of these pairs is decided one at a time.

When you import or add a paper whose identifier is already in the library, no second copy is created: the new source is recorded on the paper you already have.

How titles are compared

Titles are compared after ignoring capital letters, punctuation and extra spaces. The score is the share of words the two titles have in common, so a reordered subtitle or a missing full stop does not hide a match.

No score is high enough to merge by itself. A title match always needs a person.

Reviewing a group

Each pair shows why it was suggested and both papers side by side. Merge and choose which paper to keep, or keep both. A pair you decided is not suggested again by a later scan.

• Check the year and journal before merging on title alone. Two years apart usually means a conference paper and its journal version.
• A correction notice is not a duplicate of the article it corrects. Keep both.
• Merging keeps every source record, so you can still see that a paper was found in three databases.

Do it before screening

Removing duplicates first stops two people from reaching different decisions about the same paper. A duplicate found after screening means reconciling two decisions instead of simply merging two records, which takes longer.