What's Wrong forum: topic
This is a public forum
Affiliations search and consistency
Richard Akerman
Wednesday, 27 August 2008 14:46 UTC
Affiliations is a problem for any complex organisation, it would be great if you provided more suggestions and did partial-matching suggestions as people entered their affiliation. I think it’s probably impossible to reflect the full structure of organisations, you might want to encourage people to affiliate at the top level first and then add more levels later.
Also when I do a search
http://network.nature.com/people/search?q=national+research+council&s=by+affiliation
My name shows up twice, once just saying “you” and once saying “You” with a “P” underneath it for some reason.
(Just as an example of organisation complexity, all of these would be valid for me:
- NRC
- NRC-CISTI
- NRC-CNRC
- CISTI
- Canada Institute for Scientific and Technical Information
- National Research Council of Canada
and many other variations and combinations )
-
Replies
-
I agree – MRC has lots of little bits across the UK. My workplace NIMR can be described as MRC NIMR or various combinations of expanded abbreviations. I think this topic did get discussed 18 months ago but I’m not sure much came of it.
We’ve seen lots of work on author idenitifers, but affiliation identifiers are still in their infancy. Elsevier have done some work with their affiliation identifier but I think it’s still a bit rough round the edges.
-
Question. If we have proper unique author IDs does the issue of affiliation go away? Or do we still need a ‘proper’ identity for each institution in the same way we need one for each person.
And does this become a nightmare rapidly as you start looking at departments, faculties, hockey clubs….
-
No, it doesn’t go away as you still want to be able to sum all activity at the level of organisation, and possibly super-levels and supersuper-levels.
Yes, it does become a nightmare when you ask “what is an organisation”. There is an 80:20 phenomenon whereby the long tail becomes uneconomic to resolve. If you have to cope with mergers and acquisitions (e.g. Beechams => SmithlKlineBeecham => GSK) or loose associations and partnerships then things become even worse.
-
I guess I was thinking more that if you were sure you had a good handle on the person then their assertion of where they worked might be adequate (with a bit of clever text matching and probably a lookup table). BioMedExperts for instance seems to do a pretty good job of figuring out where I have been even though the names are inconsistent.
but yes 80/20 (or maybe even 90/10) is a good thing to aim for in practical terms
-
Yes, that’s true but if addresses in publications are anything to go by there can be a high degree of variation in the expression of “Institution X” between different people from that institution. That makes it difficult to look at everyone from Institution X.
I’m not sure I’m being very clear there – bit bleary-eyed this morning ;-(
-
Apologies for taking so long to respond here. We recognize our affiliations feature can be improved to cut down on multiple entries for the same institution. We hope to get it fixed soon.
Our aim with this feature was to allow people to easily find others from the same institution and to find alumni as well. So even if we have unique user identifiers, I still think there’s some use in people specifying their current and past affiliations.
-
