[Ref-Links] DOI paper on reference linking analysis

[Herbert Van de Sompel <hvds@lanl.gov>]

I finally had the chance to read Norman Paskins' document on reference
linking.  I have the following remarks that I hope you will find
constructive.  I use the papers' conventions in my description.  Since I
can't use superscript or subscript in mail, I will use a mimicing
syntax, that I guess is clear:
D1.M1 = = D2.M1 = = W1

Comments:
========

* There is a need to extend the concept of a manifestation entity.  as
far as i can see, manifestation entities in Norman's document are
referring to the paper itself. All described manifestations seem to
refer to the printed version of that paper and/or to one or more digital
versions.  This notion has been expressed in the paper as:
D1.M1= = D2.M1 = = P.M1 = = W1
This does not take into account the fact that - even in the print world
- there are many more manifestations of a work.  Meta-information,
stored in different A&I databases are actually also manifestations of
the same work, as are software files associated with the work etc...  In
this way, also the citations in a work are manifestations of the work.
They are actually manifestations of 2 works: the citing and the cited.
The bucket concept (Michael Nelson) formalizes this notion.  The
BiomedNet BundledLinks uses a similar notion.

A simple example is a paper, available in print and digital, and
abstracted in Medline and Current Contents.  There we can
write:
* D1.M1 = = P.M1 = = W1= = D2.M1 (the medline record) = = D3.M1 (the
current contents record) = +/- D4.M1 (a citation in the paper) where
D4.M1 = = W2

* Again, in the bucket or BundledLink concept we would create
identifiers somehow as follows (using a handle kind of representation):
- [D1.M1] = [W1]:1
- [D2.M1] = [W1]:10
- [D3.M1] = [W1]:15
- [D4.M1] = [W1]:20

* Although seeing a reference in a work as a manifestation of a work
might look a little awkward, it opens the way for realistic and
pragmatic ad hoc solutions.  We must agree that we are facing a big
problem when looking at the creation of "identifiers for works" when
talking about citations in papers.  I mean by this that it is not
obvious at all to insert identifier [W2] next to the corresponding
citation in - for instance - D1.M1 .  The identifiers either just do not
exist or are in practive "out of reach".   It is however pretty
straightforward to insert identifier [D4.M1] in D1.M1 ,when - as shown
above  [D4.M1] =  [W1]:20

* If  D4.M1 (the citation) is then stored in a database (which is an
underlying assumption in Normans paper anyhow) with key [D4.M1] =
[W1]:20 , the citation can be retrieved from that database.  Using an
approach as presented in my SFX paper – which presents a solution to the
Harvard problem created before it was called such -   a local linking
service can retrieve the record, and parse it in order to see which
services can be delivered for the citation.

* Even in the case where it is possible to insert identifier [W2] next
to the citation, it remains interesting to have identifier  [D4.M1] =
[W1]:20 there too, since at all times this remains referring to [W1] and
all its other manifestations.  So, one could propose to insert an
identifier like:
<citing>[D4.M1]</citing><cited>[W2]</cited> and fill-out whatever is
available.  As shown [D4.M1] could always be available and therefore at
all times services could be delivered based on the citations.

* The framework I have described in my SFX paper shows how easily such
an identification mechanism can lead to "local resolution" solutions.
Apart from the availability of the identifier(s) next to the citation,
the only thing that is required is a mechanism that redirects the click
of a user on a citation to his preferred linking-service (institutional
for example, or whatever).  As far as I can see, doing this for HTML
docs could be very straightforward without significant overhead to the
publisher.  The publisher could send “partial” URLs for each citation,
that are completed when the information is rendered by the local
browser, using information from a cookie (with a standardized name), set
by the local environment (harvard)  containing the address of the local
linking-server (sfx.harvard.edu/sfx-cgi?).  This would only require a
very limited, uniform JavaScript enhancement at the publishers’ side.  A
solution is less straightforward with PDF documents since I don’t see a
way to insert partial URLs and resolve them later.  But a pragmatic
approach would be the one – like taken by APS – where a paper is sent
via an HTML document containing meta-information at the top, a link to
the pdf in the middle and the citations (in HTML) at the bottom.  Links
can again be created in the HTML doc, using the mentioned method.  It is
noteworthy to mention how the identifying mechanism in which citations
are seen as manifestations of the document itself  ([D4.M1] =  [W1]:20 )
makes such an approach very straightforward: just provide all the
available manifestations of the document in one HTML file.

herbert


--
Herbert Van de Sompel
Head of Library Automation - University of Ghent - Belgium
Research Assistant - Research Library Los Alamos National Laboratorium
tel (505)-667-9233

Objects on mirror sites are closer than they appear.



------------------------------------------------------
Ref-Links maillist  -  Ref-Links@doi.org
http://www.doi.org/mailman/listinfo/ref-links