Thanks to Santy and Les for the pertinent comments on OJAX federated
search service software and on OAI-PMH harvesting in general - and
apologies for the delay in replying.
On 9 Jan 2007, at 17:54, Santy Chumbe wrote:
> I am not totally convinced on the ultimate effectiveness of a
metadata-centric search service created and sustained using mainly OAI-PMH.
Your article on OAI-PMH harvesting
(http://eprints.rclis.org/archive/00006394) raises some critical issues
about metadata quality which we'd agree with wholeheartedly. As the old
maxim says, GIGO (Garbage-In-Garbage-Out). Metadata quality is vital
whether you're harvesting it or not and whether you're using OAI-PMH or
an alternative standard.
>> How easy is the aggregated data by OJAX to maintain?
This hasn't been our focus - our main focus in developing OJAX has been
the GUI. We aim to provide the user tools to extract the most from
whatever metadata does exist. Also, because OJAX has been developed
using loosely coupled Web Services, it should be relatively easy to mix
and match the GUI with alternative back-end solutions.
>> Is the OAI harvesting of metadata completely automated and free of
human intervention? Is the normalization (metadata quality, metadata
augmentation, etc.) a complete automatic process?
Yes and yes. Of course, the higher the quality of metadata, the better
the results. CELESTIAL (mentioned in Les' reply) sounds like something
we should explore further.
>> Which other issues should a developer/implementator considering OJAX
take into account?
Federated search (ie searching harvested data), such as facilitated by
OJAX, has advantages over metasearch, (searching remote databases):
end-user searching is local so bandwidth is kept low - and searching is
unaffected by the availability of the remote datastores.
On 10 Jan 2007, Leslie Carr wrote:
> Can Judith comment on how the Web 2.0 UI facilities enhance the
'information seeking' activities of the end users? Do they really help?
And if so, which ones have the biggest win? Autocompletion? Or dynamic
content presentation?
We are about to start further usability testing but informal user
feedback so far suggests that auto-completion can help with
difficult-to-spell names and is also useful in providing immediate
feedback as to whether a term is present. It also shortens search time,
particularly when long search terms are used, and can increase user
confidence in the search.
Dynamic content presentation can be very helpful, once the users get
used to this alternative paradigm. It can shorted search time as users
can find that fewer search terms are needed than initially anticipated.
We'd be happy to advise/collaborate if anyone has any application
ideas/issues.
Regards
Judith Wusteman
>
> Date: Wed, 10 Jan 2007 13:18:37 +0000
> From: Leslie Carr <[log in to unmask]>
> Subject: Re: OJAX federated search service now in Beta
>
> On 9 Jan 2007, at 17:54, Santy Chumbe wrote:
>
>> However, after a couple of years working with the PerX project
>> (http://www.icbl.hw.ac.uk/perx), which has developed a pilot
>> service using similar technology to what OJAX uses, such as OAI-PMH
>> and Lucene; personally I am not totally convinced on the ultimate
>> effectiveness of a metadata-centric search service created and
>> sustained using mainly OAI-PMH. We have learned that in despite of
>> its relative simplicity, an OAI-PMH service can be harder to
>> implement and maintain than expected. We have spent a lot of
>> effort harvesting, normalising and maintaining metadata obtained
>> from OAI data providers. In particular the issue of metadata
>> quality is an important factor here. A summary of our experiences
>> dealing with OAI-PMH can be found at http://eprints.rclis.org/
>> archive/00006394
>
> Your paper raises many issues that are relevant to managing OAI-
> services: managing the harvesting process, interpreting (and
> correcting) OAI responses and interpreting OAI DC. All of these are
> problematic!
>
> CELESTIAL (celestial.eprints.org) offers some help with the first and
> second of those problems by acting as a proxy OAI harvester and
> normaliser. You should be able to reharvest corrected records from
> CELESTIAL that were obtained from "any" OAI data provider (currently
> 812 sources are monitored). CELESTIAL is used as the basis of the
> CITEBASE citation analysis and ranking service and the PRESERV
> project preservation profile service.
>
> The last of these problems (interpreting OAI DC) is addressed by the
> Eprints Application Profile (http://www.ukoln.ac.uk/repositories/
> digirep/index/Eprints_Application_Profile). Although it will be a
> while before all repositories take on board its recommendations, it
> does provide clear and well-thought-out advice about how to map
> repository metadata into an OAI container.
>
> I'm interested in the OJAX implementation from the point of the
> user's experience. Can Judith comm ent on how the Web 2.0 UI
> facilities enhance the 'information seeking' activities of the end
> users? Do they really help? And if so, which ones have the biggest
> win? Autocompletion? Or dynamic content presentation?
> --
> Les Carr
--
Dr Judith Wusteman
UCD School of Information and Library Studies
University College Dublin
Belfield
Dublin 4
Ireland
Tel: +353 1 716 7612
Fax: +353 1 716 1161
URL: http://www.ucd.ie/wusteman/
|