Panel Session 9
*Intellectual Access to Information Using Expert Systems as the
Filter in User Interfaces*
Sunday, March 8, 1998

Pat Young, Head Vocabulary at the Getty Information Institute, moderated
the panel session, which was attended by approximately 140
participants.  She introduced the three panel members, Karen Neimanis,
Howard Smith, and Murtha Baca, filling in for Joseph Busch.  Pat
mentioned that the panel session resulted  from Liz O'Keefe's proposal
for a session on expert systems and their use in accessing information.

Murtha Baca, Project Manager at The Getty Information Institute, was the
first speaker.  Her presentation, entitled *The Evolution of the Getty
Vocabularies: from Authority Files to *Knowledge Navigators,'* included
1) a history of the vocabularies, 2) a discussion of the importance and
structure of the vocabularies, 3) a demonstration of AAT, TGN, ULAN,
and 4) a look at the future of expert systems and the concept of *hubs.*

Originally, the Getty vocabularies were released as authority files, but
with the explosion of the Internet, they are being used as access tools
for the Web environment.  In order to emphasize the importance of
vocabularies as access tools to cultural heritage information,  Murtha
demonstrated these knowledge bases that represent concepts of
people, places and things, beginning with the Union List of Artists
Names (ULAN).  ULAN's 200,00 names are structured by equivalence
and include relationships between any variants of those names.  Murtha
pointed out that a name could be found under several variants and that
these variants were all valid name forms .  The Getty Thesaurus of
Geographic Place Names (TGN) and The Art & Architecture Thesaurus
(AAT) were then demonstrated, again emphasizing that variant names
and spelling can exist for any one concept. 

Murtha then focused on the topic of improving access to data and
demonstrated a.k.a., a project that uses the Getty vocabularies to search
bibliographic databases on the Web.  When a user knows a concept, but
cannot identify it by a specific name, he or she can first search the AAT
to find the term and then apply it in there search.  In conclusion, Murtha
discussed how access could be improved.  In the Web environment,
user interfaces can be problematic, contextual relationships are often
lost, and issues of economics and a lack of content standards on the
Web also exist.  The domain expertise *hub* concept, an organized
collection of harvested information systems, was posed as a possible
solution for mediating between information providers and end-users in a
Web environment .

In reference to a.k.a, Maryly Snow asked Murtha to repeat a step in
searching, so she could better understand how the concept was found. 
Murtha explained that the image of the term she was looking for  was
supplied by our connection to AltaVista.  At the conclusion of Murtha's
presentation, Pat addressed the subject of hubs bringing scholarly
databases together, and the challenges that presents.

The second panelist was Karen Neimanis from the Canadian Heritage
Information Network (CHIN).  CHIN is a special agency within the federal
government of Canada, which works with museum members and other
partners to provide public access to cultural heritage information through
the World Wide Web.  Karen's presentation focused on a series CHIN
projects aimed at improving access to information on their Web Site. 

In order to better understand their audience, CHIN researched how
people behave while searching for information, how to engage them and
allow them control over the information's presentation style and the
current status of Web searching, what is working and what is not.  It
was determined that, in the search for that *nugget* of information,
something was needed to avoid wading through hundred of hits. 
Therefore, CHIN's goal is to encourage the use of tools to improve
access to information, such as vocabularies, subject specific and
multilingual thesauri.  Their first attempt at addressing these issues is *the
Integrator,* a tool that allows the user to select which resources he or
she wishes to search with, using single or multiple search criteria.

Karen then demonstrated *the Integrator,* starting with a broad topic to
illustrate the problem of how the context of a search can be lost with too
many hits.  Karen discussed CHIN's focus on The National Inventory
Databases, which holds information on collections held in Canadian
museums.  The three components of the database focus on the
humanities, natural sciences, and archeological sites.  These resources
were compiled over the last 25 years by individual museums using a
common field structure, but without the benefit of terminology standards. 
Their goal is to improve access to this information, encourage the use of
vocabulary resources, provide tools to assist with terminology, reduce
extraneous hits, and enable the user to select a context when
searching.  Karen then discussed how CHIN can make this large
resource available to both the professional and public audience in a more
meaningful way.

By turning to a multilingual thesaurus, based on the AAT, CHIN's retrieval
system was greatly improved. Karen showed the basic structure of the
thesaurus, how terms are related and then demonstrated a search in
their Humanities database.  Both the full record for the desired term , and
its hierarchical position, were displayed.  Karen then concluded her
presentation with a summary of CHIN's accomplishments and future
goals, which include integrating other resources such as the ULAN,
TGN, etc., into their database. Claudia Hill from Colombia asked if the
studies CHIN conducted regarding the searching behavior of users was
published.  She replied that the study was internal was not published. 
Ted Goodman from Avery asked if CHIN planned to enter more
conservation-related terms.  Pat responded that contributors to the
Conservation Information Network, managed by CHIN, had recently
adopted the AAT and TGN as authorities.  Pat hopes that this group will
also be contributors to the vocabularies.  Pat also reported on a study
done on the CHIN humanities database that showed as much as 99
percent of lay conservation terms were found in the AAT. 

The last speaker was Howard Smith, Manager of the Architecture and
Fine Arts Library and Image Collections at the University of Southern
California.  He presented Information System for Los Angeles (ISLA), a
database of information on the city of Los Angeles.  When introducing
Howard, Pat mentioned that his project was unique in that it introduced a
new indexing criteria: time and space.  Howard demonstrated USC's
new ISLA project, which is a prototype of a larger proposed project,
Integrated Digital Access (IDA).  ISLA contains multiple information
formats such as photographs of Los Angeles, maps, census reports and
information from all historical periods of the city's history, and uses a
time/space retrieval system.  IDA has the ability to search and link three-
dimensional information and an archive can use ISLA through distributed
network links and remain autonomous. 

Howard then discussed the nature of database design and that often
they end up being creations for systems designers rather than
end-users.  Often, the interface is the last element to be created and
input from the user is secondary.  ISLA and IDA did the opposite; the
database was supervised by a team of three representatives from a
users' group and three Library and Information Science faculty members.


ISLA is made up of two important components: 1) a library of digital
materials in various formats and 2) a specific search/retrieval method
involving space/time/full-text and format indexing.  All materials are
indexed by spatial and temporal, as well as conventional indices. 
Atypical formats such as texts and photographs, are spatially indexed. 
ISLA was intended to maximize access to a mass of interdisciplinary
data about Los Angeles for the purpose of teaching, research and public
access.  Users are allowed to define searches by using historically
appropriate spatial layers or they can overlay data layers from different
time periods. Howard pointed out, however, that the user is not required
to use the spatial/temporal element, and can simply use the full text line
for conventional searches.

The first phase of the ISLA project was to develop a working prototype
of a map server, a content server  (a source of access to texts,
photographs and other materials), an attribute server, and a search
engine.  As a result, a small prototype focusing on downtown Los
Angeles, has been designed to highlight the desired functions of the
system.

Howard then further explained the aspects and issues of space/time
indexing. While bounding box searches that use City of Los Angeles
parcels and Thomas Brothers maps work well for most items, some data
elements fall out of this spectrum.  Howard gave the example of a doctor
who both owned a popular hotel and attended USC. His portrait was
linked to both the hotel building and USC. Howard then discussed the
software used.  Since most commercial forms are too slow, files were
converted into flat text forms so it could be searched via USC Library's
search engine.  Howard then demonstrated a search.  Users can click
on maps and time lines to find information, there is a pan/zoom feature
that identifies locations and GIS Aerial photographs can be accessed as
well.  The design allows search results to be displayed as hits on the
map and on a timeline.  Users may go back and forth from results to
search screens, modifying searches as needed.  Options include
displaying the full catalog record or item itself on the screen, and
importing information to a user's workstation.  ISLA will be available on
the USC Library web site in April.

Howard then took questions from the audience.  An audience member
asked about how the system searches for date ranges.  Howard
responded that the system would pick up the first date entered.  Other
questions dealt with levels of accuracy and how the coordinates of
structures were determined.  Howard responded that they were in the
process of dealing with issues of accuracy and that coordinates of
structures are determined by addresses.  USGS maps were used
because other maps resulted in distortion. Howard also mentioned that
outside contributions were welcome.  Another audience member asked if
social issue information would be a part of the data. Howard responded
that once the core data is entered, social information is a consideration
for the future.

Following the three presentations there was some discussion.  The
moderator, Pat Young, concluded the session by discussing some of the
challenges facing us in the future; to match the user's concept with that
of the author or creator of the digital resources with tools such as
thesauri and knowledge bases, to provide access to interdisciplinary
resources that have vocabularies specific to their discipline, to provide
multilingual/multi cultural points of view, to move concepts of place from
the level of gazetteer to thesaurus, to create resources for time beyond
a timeline, to represent search results in terms of time and place, to
authenticate cultural heritage information on the Web, and to structure
our own Web resources they can be retrieved in context.  Pat concluded
the session by thanking the participating panelists and to the audience
for joining in and for surpassing her expectations of attendance.

Robin Johnson, Getty Information Institute

****************************************************************************************