Position Paper for Workshop on Rule Languages for Interoperability
by Aubrey Jackson, Waymaker, Inc.
and David Scott, Haplos, Inc.
1 Introduction
We believe that a successful general framework for web interoperability
will not be based on a language for sharing rules but a rule language
for defining data collectives and the topologies into which they can be
arranged. The rule language will need to describe a collective's
membership policies, internal security policies, data standards,
functional concerns, members, and rules for exposure to other
collectives. Depending on the rules for participation in the
collective, members might not be required to implement rule languages
or sophisticated data models themselves. By plugging into a collective,
a member will inherit the formal representation and translation
resources of the collective.
We believe that solutions which depend upon the widespread adoption of
semantic web technologies or the automated generation and translation
of ontologies will have a narrow range of application. Our proposed
solution encourages the pooling of resources to reduce the cost of
integration. Our own recent work concerns data sharing between loosely
related verticals in the commercial real estate industry -- eg.,
between realtors, appraisers, building managers, investment trusts,
title companies, insurers, banks, contruction contractors, and others.
These verticals already have their own data standards which they have
taken years to develop (although the adoption of these standards is
very uneven.) The goal is to encourage the further development and
adoption of the existing standards and to come up with a framework for
the exchange of data between the different verticals. [Note: So far,
the use of metadata in this space (as defined by the Real Estate
Transaction Standard) has been a failure.]
2 Background
In the Call for Participation, it was noted that current work on rule
languages and ontology by computer scientists has its roots in prior
work in formal logic. The optimistic forecasts regarding the outcome of
this current research echoes the optimistic predictions of the
logicians, linguists, and philosophers of the 1920s, '30s, and '40s
regarding their own early work. Carnap's Aufbau (The Logical Structure of the World)
seemed to be the beginning of a program in which human knowledge would
be formally explicated in a universal, unambiguous language. Here and
there, critics pointed to areas which, they claimed, would resist
formal analysis. However, the logical positivists believed that most of
these areas would yield to analysis, in whole or part, in the
incremental progress of their project. During the 1950s, '60s, and
'70s, the logical positivists began to question their own project --
but not so much because of the discovery of areas that resisted
formalization.
As their studies matured, the logical positivists encountered
difficulties regarding the resolution of formal ontologies. This
touched off an internal critique concerning the limits of the formal
program. Quine, one of the leading lights of the analytic school,
helped kill the program with his related theses of Ontological
Relativity, the Indeterminacy of Translation, and the
Underdetermination of Theories (translation being, itself, a kind of
theory.) Formal analysis, using the tools and techniques of the logical
positivists, continues, but not in the service of a grand unifying
program to formalize all of human knowledge.
Similarly, recent blog chatter reveals skepticism regarding the promise
of the Semantic Web. The discussion generated by Clay Shirky's
pessimistic prognosis for its success helped call attention to some of
the hyperbole surrounding the Semantic Web and generated some frank
evaluations of what its development will yield by people currently
working with this technology. A view repeated by many participants in
the discussion was that what has and will emerge is not a global
ontology or a universal framework for resolving ontologies, but a patchwork of local ontologies.
Rule languages and core data types will lessen the burden of
integrating these ontologies; but in many instances, the task of
integrating local ontologies will be done by committee, not by computer.
The primary focus of these technologists is in using rule languages and
sophisticated ontologies to increase the expressive power of their data
models. They expect that this greater expressive power will lessen but
not solve the problem of data integration. It is our belief that a
successful framework for interoperation on the web will take for
granted the patchwork of local ontologies which cannot be resolved
through formal means alone. Furthermore, if it is to be widely adopted,
it must facilitate the integration of resources which do not themselves
support rule languages or formal ontologies. (The high cost of
implementation and the shortage of qualified ontologists will slow the
adoption of sophisticated data modeling.)
3 Proposal
The solution we are investigating is a framework for linking together
disparate database sources into a collaborative network. The network is
subdivided into groups which are defined by their membership policies,
services, supported data standards, and connections to other groups in
the network. In general, a participant selects a group which supports
the standard to which the participant can most easily map its existing
resources. However, a participant can join as many groups as necessary,
offering differing views of its resources to each group. Each group is
governed by a local ontology and rule set which may remain internally
implicit. That ontology is made explicit only to the exent necesssary
to expose a group's resources to other groups. Translation services are
provided by the group; so, depending on the network's topology, the
participant potentially only needs to map his resources once (ie., to
his selected group) in order to expose them to the network as a whole.
Our intent is to encourage the flourishing of communities around local
ontologies, but to allow them to interact with other communities
through black-box translation services that are a well-identified,
integral part of the community's network.