Affiliations in conf.researchr.org are entered as free-text input, with autocompletion suggestions being presented as soon as the actor starts typing. These suggestions encourage the actor to choose an existing entry representing his affiliation.

However, this does not result in having just 1 variant representing everyone’s affiliation who work at the same company.
If we can treat different variants as alias of a main variant, it would give us:

• simplified filtering in programs/people index
• ability to group contributors (eg authors) by affiliation
• ability for improved data analytics
• improved autocompletion on entering affiliation (one can also use co-author affiliations to boost suggestions)
• …

From a quick analysis I observe the following cases of the same company being referenced, but entered differently:

1. Different uses of whitespaces and symbols
   • Universidad Autónoma de Madrid vs Universidad Autónoma de Madrid (leading whitespace)
   • University of Illinois, Urbana-Champain vs University of Illinois Urbana-Champaign (vs University of Illinois at Urbana-Champaign)
2. Using acronym vs using full name
   • TU Delft vs Delft University of Technology
3. Added country or not
   • Aarhus University, Denmark vs Aarhus University
4. Linguistic variants (International vs local language variant)
   • Technische Universiteit Eindhoven vs Eindhoven University of Technology
   • (Technische Universität Darmstadt vs TU Darmstadt)
5. Added department
   • Scientific Computing, TU Darmstadt vs TU Darmstadt, Real-time Systems Lab vs Institute for Scientific Computing, TU Darmstadt
6. Having multiple affiliations
   • Eindhoven University of Technology (TU/e) and Stellenbosch University
   • CWI & University of Groningen
7. Typos and (not) use of diacritics

For (6.) we may want to add support for multiple affiliation entries.

When adding the notion of aliases, we can index the main entity and its aliases for search, so we can provide suggestions for affiliations that are not associated to a main entity yet. We can provide these suggestions to website visitors that are displaying a person’s profile, who can then choose the correct main entity from these suggestions. The system administrators can later review these community-provided associations in bulk.

Submitted by Elmer van Chastelet on 27 November 2018 at 14:54