Re: Given a university's name, retrieve URL for university's home page.

PubMed was assembled with those three assumptions (+ peer review).  The problem I referred to as "solved" was the Search Stack logic.  The result set is different from queries of Page Rank based data stores (Web Search Engines). You might have better luck with DDG (https://duckduckgo.com/ BTW) Neither knowledge universe is absolutely complete, but the relationship overlap (we hope) yields valuable insights.  As a practical matter, a data store (PubMed) with "only" 22 million entries would be a very lame basis for a Search Engine. For a different reason, your mileage may vary.
--Gannon




________________________________
 From: Sam Kuper <sam.kuper@uclmail.net>
To: public-lod <public-lod@w3.org> 
Sent: Tuesday, May 14, 2013 9:32 AM
Subject: Re: Given a university's name, retrieve URL for university's home page.
 

On 14/05/2013, Gannon Dick <gannon_dick@yahoo.com> wrote:
> Affiliation (for the text name) and domain of the lead Author's email should
> give you a little "uncertainty" with which to  resolve DBpedia.  Their rules
> are very fussy and not as much "uncertainty" as you would like, but it is a
> start.

IIUC, this strategy's success rests on (at least) the assumptions that:

[1] Each of the universities I'll be searching for is listed as an
affiliation in at least one publication within NCBI.
[2] For all such publications, the lead author's email address is
provided among the metadata for the publication.
[3] For all such publications, the lead author's email address
incorporates the domain of the affliated institution for which I
searched.

I may, as I say, be being a bit slow-minded, but these each strike me
as rather tenuous assumptions; and the likelihood of them all being
true seems even smaller.

Assumption [3], for instance, was false for the first test I ran:
affiliation searched for was "London School of Economics" but although
both authors of the first open access publication listed shared this
affiliation, the contact email's domain was "popcouncil.org" rather
than "lse.ac.uk". Assumption [2] was false for the third test I ran:
affiliation searched for was "Royal Holloway", but only the
publication's third author's email address was provided (which
happened to be for the "cam.ac.uk" domain).

I suppose I could try to narrow down the results to those with only a
single author, but that still wouldn't automatically fulfil
assumptions [1]-[3].

Perhaps I am still failing to understand the crucial insight that
enabled you to state with confidence that, "The problem is already
solved in fine detail" via the NCBI; if so, please could you share it?

Many thanks,

Sam