An "Entry Vocabulary Modules" provides associations between the user's ordinary
language to domain-specific technical metadata vocabulary with which the user
would begin a search. The process of creating an entry vocabulary module is one
of Bayesian inference, wherein sufficient training data (consisting of document texts)
are downloaded (using the Z39.50 protocol for efficiency) from a document database to
provide a probabilistic matching between ordinary language terms and the specific
metadata vocabulary which have been used to organize the data. Developing the entry
vocabulary utilizes both natural language processing modules [Kim and Norgard 1998]
as well as statistical techniques to identify noun phrases (e.g. `color laser printer')
and individual words to map to specialized classification.
[Top]
1. Autonomy: agents work on their own without direct interventions of humans or others.
2. Social ability: agents interact with each other using an agent communication language.
3. Reactivity: agents perceive their environment and respond in a timely manner.
4. Pro-activeness: agents are goal-oriented.
For other definitions of intelligent agents, see Franklin and Grasser (1996) and
Haverkamp and Gauch (1998).
To reduce communication cost as well as system workload, the EVMs system design utilizes
a federated system
[see internal architecture] consisting of multiple interacting agents which
"divide and conquer" the entry vocabulary technology tasks. They range from builder agents
which create association from training records downloaded by data retrieval agents and
prepared by cleaner agents. Desktop agents and domain agents help the user to define domain
of interest and deploy the created association dictionaries. Proposed planner/optimizer agents
schedule and re-schedule record downloading tasks in an optimal way.
[Top]
What is the architecture of EVMs system?
Running on a UNIX platform, the EVMs system is implemented in the conventional TCP/IP
client-server environment. The user accesses the EVMs system from Java/Javascript-enabled
or standard Web browsers such a Netscape navigator and Microsoft Internet Explorer from
URL: http://www.sims.berkeley.edu/research/metadat/,
and the Web server initiates
a session (channel) between the user and the EVMs system
via a Common Gateway Interface
(CGI). The external architecture of the EVMs system
design is shown in [external architecture]
and the internal architecture of the EVMs system is shown in [internal architecture].
[Top]
How to use the EVMs system?
For a first-time user, the EVMs system will invites you to create your personal profile
that includes your email address, your preferred PIN (personal identification number),
and your operational preference such as time-out settings and operation rules.
[first-time user]
Once you create your profile, the EVMs system will send you a confirmation letter
by email. Since then, you are upgraded to a returned user, and you are requested
to enter you PIN next time you come back to the EVMs system.
[returned user]
When you go to the main page either as a first-time user or a returned user, the EVMs
system will prompt you six options. They are: