Publication Centrality

When clustering publications into Topics, we have established a set of five centrality categories representing how strongly linked (or central) a publication is to its Topic based on the number of references and citations links it has to other publications in the Topic:

• Definitive
• Very good
• Defensible
• Weak
• Other

We consider a publication to be strongly linked or central to the Topic to which it is assigned if it falls within either of the first two categories: ‘definitive’ or ‘very good’. A publication is deemed central if at least 30% (with a minimum of 5 links) or at least 20 (with a minimum of 10%) of its citation links (either references or citations from later documents) are to documents in the same Topic.

The category ‘other’ includes publications with only one citation link to the documents in the same Topic and publications assigned to its Topic using a text similarity model. The text similarity model is used either when a publication doesn’t have any citation links, or when it has the same highest number of links to more than one Topic – then the text similarity model is applied to the top scoring Topics.

The centrality category is assigned to each publication and the assignment happens once the publication is added to its Topic, so that it does not change every time a publication gets a new citation link.

A Topic publication centrality changes with the year range selected, as it depends on the publications included into the year range. Please note, when analyzing recent year ranges, publication centrality is likely to be lower than when analyzing less recent year ranges, as recent publications have less citation links than older publications.

With the launch of next generation SciVal Topics in May 2024, the average Topic centrality – proportion of publications being strongly linked to its Topic, increased from about 64% for the previous generation Topics to about 72%.

Analyzing publications centrality in a Topic helps identifying central and more peripheric publications as well as understanding how condensed or diffused a Topic is overall.