From: Gerry McKiernan <gerrymck_at_iastate.edu>
_ Agent-based Cooperative Collection Development_
In response to my recent posting _Content Analysis of Academic
Departmental Homepages_[COLLDV-L no. 1774], a respondee raised one of the
major issues relating to Web resources of the day -- Quality. In
considering a means of identifying Quality of resources that might be
presented for
consideration by the Model presented in this posting, it occurred to me that
would could adopt/adapt the technology employed in Recommendation Agents,
i.e., agents that present a user with candidate resources based upon
criteria established from an initial questionnaire _and_ subsequent
interaction with recommendations made by the Recommendation Service.
For example,_Firefly_, one of the oldest and well-known Web-based services
that employs Recommendation Agents, is a personal software agent that
understands people's tastes and interests, and allows users to access reviews
and receive personalized recommendations on movies and music.
In the context of Collection Development, I see a parallel service that
would initially generate a set of candidates resources based _not_ upon a
completed questionnaire, but on a Research Interest Profile (RIP) that would
automatically be derived from individual faculty members and departmental
homepages from a Content Agent [No doubt, we would also want to consider
integrating a Conceptual Agent here at some point such that the
'text-expressed' content would be massaged with appropriate synonyms, BT,
RT, etc. for the subject domain to represent the homepage _not_ as a
collection of terms _but_ of concepts. This massaged RIP would become the
basis for searching Internet accessible resoures using an associated Search
Agent. The Search Agent would submit the RIP to such sites as
electronic bookstores (e.g., Amazon.com) to identify candidate monographs
works, or to individual
publisher sites (e.g. Elsevier) to identify possibly relevant print or
electronic journals for the 'local' collection', professional society
clearinghouse (i.e.,Scholary Societies Project
(http://www.lib.uwaterloo.ca/society/overview.html),
as well as significant Web resources (e.g., OMNI for Medical resources)
The results of these submissions to these sources would be sent based to
the Review Agent that in turn would collate all the results from a Hard
Night's Day [:->] and present teh results based upon the relative importance
of an individual's RIP or a collective departmental RIP, using an
appropriate algorithm. Based upon the selection of items of these
retrieved results by a selector, a profile established for a Selector or
Bibliographer would be modified to reflect their new and ongoing choices,
subsequently affect the dynamics of teh algorithm and the ranked
presentation of subsequently retrieved items from future Search Agent
searches [One could consider integrating this feedback activity into a
Duo RIP that would be the synthesis of a faculty member RIP and that
generated from the selecting activity of a bibliographer].
Likewise, as noted in the original post, as a subscribed member of The
Service, the activity of a individual faculty member in his/her selection of
material could be feedback to his/her RIP by a Feedback Agent such that the
RIP exists in Real Time and at a high level of specificity.
One could also envision bibliographers who have similar subject
responsibility be subscribers to the Service, such that _their_ selections for
_their_ cleintele as well as _their_ faculty and departmental RIPs are
integrated into a larger System.
It is here through the Combined Group behavior within The System and
Service that we can _indirectly_ determine Quality. To more formally
establish Quality, we would need other agents, notably an Acquisitions
Agent and a Citation Agent. The Acquistions Agent that would provide
feedback to
The System and The Service to identify a work as formally Acquired [Here we
believe that formal Acquisition is a manifestation of a Quality judgment]
[The Acquisitins Agents of course would need to be integrated within the
institution's Acquisition system]. The Citation Agent would monitor the
publications of a researchers to determine if he/she cites/sites works made
known by The Service.
Whew!
In preparation for an article I will be writing in July on Agent-based
Cooperative Collection Development, I would much appreciate any reactions to
this scenario. As Always, Any and All contributions, questions, concerns,
comments, citations, or queries are Most Welcome!
Gerry McKiernan
Curator, CyberStacks(sm)
Iowa State University
Ames IA 50011
gerrymck_at_iastate.edu
http://www.public.iastate.edu/~CYBERSTACKS/
Received on Fri Jul 17 1998 - 09:03:04 EDT