Conor Shankey

Visual Knowledge

Vancouver, B.C., Canada

http://www.visualknowledge.com/

Conor Shankey is the founder and CEO of Visual Knowledge. Visual Knowledge is an enterprise class ontology life cycle management platform and inference engine that enables the development of Lego like agent based systems. Visual Knowledge servers consists of active semantic agents that represent active ontological elements such as concepts, properties, axioms and transformations. The meta, meta layers of Visual Knowledge enable the platform to support different ontological and meta standards and the system is also fully compliant with OWL and RDF.

Key capabilities of Visual Knowledge are:

1. The ability federate ontologies and knowledge bases
2. Semantic change management and versioning of ontological concepts and ontologies
3. Configurable micro inference models
4. Integrated transaction and multi-threading
5. Micro kernel that mitigates mustering of agents up into Java or upper languages

History

In 1990 through 1996 Visual Knowledge was commissioned to develop a new kind of enterprise platform for a power utility to replace key mainframe systems that did work management, maintenance scheduling, management accounting, etc.. The first production version of Visual Knowledge (version 3) was a multi-user, transactional frame system, similar to Protégé except designed to let domain experts collaboratively build ontologies and axioms for resource scheduling and to serve as the underlying infrastructure for the enterprise. The production system managed an ontology with approximately 15,000 classes representing physical and resource systems and procedures, 3,000 rule sets and a few gigabytes of "instances". The system used ontologies of procedural classes models that could assemble work orders and identify possible resources and was benchmarked to support up to 1200 concurrent users. The system was implemented to interface with mainframe systems built in IMS and DB2 during phase out and integration. The production system was rolled out to one region but the internal IT organization was not comfortable in supporting the platform and scaling up it's use.

From 1996 through 2000 Visual Knowledge was advanced to become a platform in the aviation world and was utilized to develop trip plans, schedule crews and aircraft, automate profit and loss estimates on charters and schedule aircraft maintenance. Ontological modeling evolved from using rules based semantics to an agent based axiom toolkit for higher order knowledge representation and inferencing. The system was deployed to several charter companies and private operators.

From 1999 through 2001 Visual Knowledge was utilized to develop several B2B systems including a foreign currency exchange, semantic matching engine and payment fulfillment system. Unfortunately with the .com crash many of our industrial solution partners faded away.

From 2001 to present a biological system known as BioCAD was developed on top of Visual Knowledge to enable biologist to model gene, proteins, interactions, organisms, signaling pathways and experimental protocols. The scale of Visual Knowledge was pushed to support models with around 30 million concepts and several hundred million assertions on lower end servers. A transactional experimental system called KiNET was developed as a production system on top of BioCAD to enable the analysis of thousands of experiments in multi-dimensional manner.

From 2003 to present the fifth generation of Visual Knowledge was developed and the kernel or engine was migrated from an active object database down into our own native semantic persistence and transaction engine. This now enables us to achieve a level of scale required for the very large scale of knowledge bases that occur in even conventional technology today. We (finally) successfully developed and stabilized a packaging and version management system so that ontologies and knowledge based could be rationally changed in a federated environment of many servers. In the past year we built an OWL and RDF compliant capability on top of Visual Knowledge so that the system could readily absorb and manage ontologies represented in OWL, knowledge asserted in RDF and provide OWL DL and OWL Full compliant inferencing.

We are interested in the idea of building an experimental platform for playing with and testing more than one KR language and once with many upper and lower ontologies.