 |
eLINKy na eDOKUMENTY - "to pravé jahodové" |
| Sjoerd Vogt, The
Dialog Corporation, Witney, Velká Británie |
| Information available over the
web is increasing exponentially, but searchers continue to recognise the
enormous value of secondary information as a stepping stone to the primary.
The traditional abstract & index databases are able to find the proverbial
needle in a haystack where web search engines fail abysmally. However -
users increasingly expect to be able to link directly from the secondary
information source to the primary full-text, and many publishers and aggregators
already have such e-Linking technologies in place. The Dialog Corporation
has also recently launched an e-linking service for Datastar and for the
OnDisc range of products – and will be adding e-linking to Dialog later
this year. What was Dialog able to learn from those organisations who paved
the way – and what were the choices that had to be made? |
INTRODUCTION
The glut of free information on the web is blamed for many things -
including some things that have absolutely nothing to do with it. We’ve
seen over the past decade first a perceived devaluing of secondary information;
the theory being that the primary info will be made available free directly
to the users - under alternative publishing models perhaps. But more recently
we’ve begun to see a counter-trend. Users increasingly want single
points of entry for their information gathering process; points of entry
that offer them guaranteed reliability, content, quality, and above all
- comprehensiveness. This is leading to a re-evaluation of the traditional
secondary databases - and an increased appreciation of the important role
that they play in finding that proverbial needle in a haystack.
BUT - such secondary databases only have a valid role in the information
gathering process IF they then lead to the primary information. Traditionally,
this linking through to the primary took the form of hardcopy document
ordering and delivery - often involving dedicated staff who were (and still
are!) document delivery specialists.
However, user expectations are changing very rapidly in this respect.
Internet-literate users now expect to be able to hyper-link directly through
to the primary information referred to - and nothing less than the article
level will do.
For this reason, many of the online hosts, aggregators and publishers
have all been working hard to give their users reliable e-Linking.
Dialog is no exception - and over the past year has been actively working
on different e-Linking services. The Dialog Corporation has recently launched
an e-linking service both for Datastar and also for the OnDisc range of
products - and will be adding e-linking to Dialog later this year.
We aren’t the first - but it’s better to be “leading edge” rather than
“bleeding edge” - as they say. What was Dialog able to learn from those
organisations who paved the way - and what were the choices that we had
to make?
GOLD EDOCS
The issues and decisions for OnDisc e-Linking and Datastar e-Linking
were very similar - if not identical. Gold eDOCs adopts a “just in time”
policy - only generating the necessary URLs for e-linking on request -
whereas Datastar e-links operates a “just-in-case” policy - generating
URLs for ALL the records in the relevant set of secondary information records.
However - for the purposes of this paper, we will concentrate on “Gold
eDOCs” - the e-Linking service that is the link between your OnDisc bibliographic
database records and the electronic fulltext. Each record retrieved in
a search will have its own “eDocs” button. When you point and click, a
real-time generated URL will then take you directly to the electronic fulltext.
One of the very flexible features of Dialog’s approach is the use of
a “Remote Links Server” (RLS) . The URL that links to the fulltext at the
article level is in fact generated by a central links Server maintained
by Dialog (as opposed to having multiple static copies at customer sites).
This same RLS is used by all the different Dialog services. There are therefore
no static locally stored "links" databases at the customer sites that can
go out-of-date as soon as they are published. It is worth noting that where
aggregators have taken the approach of having the databases of links at
the customer sites, the locally stored databases can be extremely big.
As the number of e-journals available continues to increase, this problem
will only be exacerbated.
Also very important is the ability to configure Gold eDOCs so that your
users will only be presented with “live links” - those that you subscribe
to and therefore have access to. It would be very annoying and frustrating
for your users if they were to follow a link through to an e-Document -
only to find that they did not have access to it. This will not happen
with Gold eDocs. Gold eDocs will only generate URLs for your titles, via
your chosen sources. This will work equally well on agent-hosted or locally
hosted databases. This is often called “the Harvard problem” - and
we have a solution for this with Gold eDOCs.
We’ve made the decision that Gold eDOCs should be free of extra charge.
The user has already taken out a subscription to the secondary database
and to the primary journal. The link between them is software functionality
that is therefore part of their subscription.
Gold eDOCs Architecture
By using an open architecture that is centred around Dialog’s Remote
Links Server, the user has maximum flexibility in implementing and integrating
Gold eDOCs into their own network. Once the URL has been generated by DIALOG’s
RLS, the request is passed through via a second browser window to the fulltext
publisher or aggregator chosen. Under non-disclosure, the customer could
even use the Gold eDOCs XML request/receive structure to set up their own
“Remote Links Server”.
Gold eDOCs Partners
Gold eDOCs is already able to link through to most of the world’s leading
fulltext aggregators and publishers - giving access to more than 3000 e-Journals.
Discussions are underway for further linking agreements which will be announced
as they are agreed. The Gold eDOCs architecture means that new e-Linking
partners can be set at the customer site within minutes, once the central
rules have been established on the Remote Links Server.
LESSONS LEARNT IN DEVELOPING GOLD EDOCS
Scalability
If the problem is to set up links from just one secondary database,
then this relatively easy - and can be done with hard-wired links from
each record - such as happens with the NLM’s “PubMed” . If e-Linking
is just to one publisher, then this is also straightforward. “CompendexWeb
from Ei is such an example - in that it links only to the 460 journals
on Elsevier’s “Engineering Direct” .
However - the problem that a host such as Dialog faces is of an entirely
different magnitude. E-Linking now needs to work from any relevant
database to any publisher or aggregator - and it needs to take into account
all the vagaries of incomplete information and inconsistent formats.
In the Listserv “NEWJOUR” (http://gort.ucsd.edu/newjour/ ) there
are 8616 e-Journals listed as of April 10th 2000 - with approximately 150
new e-Journals being added every month. Interestingly, looking at total
of all e-Journals, the addition of new e-Journals has been approximately
linear over the past five years. We see a similar pattern if we analyse
the e-Journals available from one particular host (OCLC in this case).
This linear increase is in sharp contrast too the usual expectation that
everything on the internet grows exponentially.
It is clear that the number of e-Journals is still only a small percentage
of the total number of journals. When you consider that a big database
such as InSpec could easily cover more than 10,000 serials, then it is
likely that only a small proportion of these will produce valid e-links
at present. Similarly, if we compare the NEWJOUR number of 8000+ e-Journals
with the total number of ISSNs issued to end 1998 (nearly 900,000) then
this also shows that the number of serials available over the internet
is at present only a single digit percentage. Nonetheless - this tiny proportion
does already include most of the top serials subscribed to. It is
an interesting question as to how the remaining 90%+ ejournals will go
electronic - and how quickly. The inevitable conclusion is that existing
systems using locally held databases of links at customer sites are not
sustainable.
Each e-Journal may be available via a number of different routes - including
the original publisher (such as Springer Verlag) and the top e-Journals
aggregators (such as SwetsNet and Rowecom IQ) . Dialog’s e-Linkiing
therefore neds to be able to link from 400+ databases through to articles
hosted by many dozens of different aggregators and publishers.
Design Issues
Mutual Customers or Pay-Per-View? Both is best. However - the
linking to mutual customer subscriptions must come first. The first release
of Gold eDOCs is able to link to existing customer subscriptions. The authentication
and access to the fulltext is not in any way handled by Gold eDOCs.
However - Pay-Per-View e-Linking with credit card billing is actively
being developed for all Dialog’s services (OnDisc and Online) - and clearly
offers an important route to fulltext journals that are not used so heavily.
Reliability - or Single Click? It’s superficially tempting to
imagine that a single-click link to the fulltext is the ideal. However
- primarily because of the vagaries of inconsistent and incomplete information
- the ACTUAL ideal is to give the user the most reliable link possible
under the circumstances.
Dialog does this by first populating a “Submit” form - which can
then be checked by the user before being submitted to the Remote Links
Server for the URL to be generated. The Submit Form will be available to
the user even if the information is incomplete . However - a “traffic light”
system indicated to the user whether a valid link is likely. Green for
Go, Amber for Caution, and Red for No . This system ensures maximum likelihood
that the user will be able to link through to the primary article.
Flexibility - or Ease of Use? Another balancing act. Both are
clearly important to the user. The design of Gold eDOCs ensures that there
is no compromise on all-important flexibility in trying to make it as easy
to use as possible. However, system administrators must of course first
set up teir holdings information if “dead links” are to be avoided.
Linking to All Content Types: Linking at present is primarily with STM
journals. However - Dialog’s databases cover every sector - and clearly
the e-linking technologies implemented by Dialog need to handle every type
of information. In addition, Primary information comes in a rainbow of
varieties; from traditional reports and articles, through to multi-media
presentations and audio-visual materials. The ultimate aim is to provide
multi-directional linking from and to any type of information.
Future Proofing: Dialog is actively monitoring the DOI and CrossRef
initiatives. Gold eDOCs has been designed to be full compatible with the
these initiatives - and others.
Publisher Attitudes
Gold eDOCs is able to link through to nearly one dozen aggregators or
publishers - with more imminent. It is interesting to note that not
all potential partners are equally enthusiastic!
Reactions are of six different types:
YES Please - we want as many links as possible!
WAIT we need a corporate strategic review of this
WAIT we don’t have out technology ready yet
NO - you can link to our home page but not to our articles (this is
NOT what the users want!)
NO Go away! You’re a competitor
ZZZZzzzz if we ignore you, perhaps you will go away….
CONCLUSIONS
It’s early days for e-linking. We’re going to be seeing increasing demand
for multi-directional links that loop back on themselves and perform creative
contortions. Everything has to link to everything. It’s clear that there
is a need for a “server” that provides all of these links. The concept
is relatively straighforward - but the implementation is not.
Most of the problems stem from unstructured or incomplete data - and
solutions such as Gold eDOCs can only be so clever.
It may be time to reconsider some of the sacred cows of the information
industry.
It is wishful thinking to imagine that everything will have a DOI/CrossRef
- and this will only happen slowly. But - why not a standard bibliographic
reference format?
With the help of appropriate rulebases, this would then allow you generate
the relevant URL for the fulltext hosts in question. But - why not a standard
URL format?