ConferenceCall 2006 07 13

This is the tenth event in a series of talks and discussions the revolve around the topic: "Ontologizing the Ontolog Body of Knowledge" during which this community will explore the "what's" and "how's" to the development of a semantically interoperable application, using the improved access to the content of Ontolog as a case in point. For more information on this series, see: OntologizingOntolog    (OD4)

[picture of Dr. Patrick Cassidy]    (O8P)

Abstract: (by PatCassidy)    (O8Q)

Defining Vocabularies: Ontological and Linguistic    (O8R)

For over ten years, study groups have been exploring the potential for broad agreement on a common upper ontology as a tool for semantic interoperability. The vision of how the upper ontology will enable semantic interoperability is usually stated in terms of unifying the semantics of multiple databases, or mapping multiple domain ontologies through a common upper ontology or interlingua. This talk will describe how utility of the upper ontology for semantic integration goes beyond these scenarios, and will propose the use of an upper ontology in a way that may immediately affect the task of ontologizing the Ontolog, or similar bodies of knowledge.    (O8S)

It has been observed previously that, when attempting to create computer applications that mimic some significant fraction of the human ability to process information accurately, while taking goals and contexts into consideration, the need to use multiple modes of reasoning appears to be necessary. This has led to several proposals for generic architectures for general cognition which allow a top-level task controller to call on any of multiple reasoning methods to solve part or all of the problems which is the current goal. This general notion was called by Minsky the "Society of Mind". But the method for passing information among these reasoning modules has sometimes been an afterthought to the work on the different modes of reasoning.    (O8T)

With the increasing sophistication of ontology technology it is now possible to visualize the role of the upper ontology as the most flexible and expressive type of data representation for message-passing among multiple reasoning modules. The use of a common upper ontology can permit any specialized reasoning method to be used in conjunction with any other, provided that both can accept requests and return answers using the common conceptual language used by the other modules. For this application, the upper ontology can be seen as the form of knowledge representation that can encode knowledge that is used by any reasoning method. This means that, if a specialized form of knowledge such as probabilistic information or spatial-graphical information needs to be passed, the upper ontology must have the ability to properly encode such information.    (O8U)

One issue that has inhibited the wide use of existing upper ontologies such as OpenCyc, SUMO, and DOLCE is the complexity of such knowledge representations. Without compelling examples of applications that motivate the effort required to use those ontologies effectively, many individuals and groups apparently conclude that building their own ontology is the easier route to serve their purposes. What is lost is both the valuable experience of prior work on ontologies and the potential to communicate knowledge accurately across multiple groups of users. To address this problem, more convenient means need to be developed to make such upper ontologies easier to use. One possibility is the development of an effective natural-language interface for knowledge entry and retrieval.    (O8V)

Developing natural-language interface to databases has been studied for decades. Likewise, controlled English formalisms have been developed that provide the ability to encode logical statements in an English-like syntax. Such interfaces, however, have not integrated the English with a formal ontology sufficiently large to permit the creation of English-language definitions of specialized concepts in unrestricted domains. In addition, the ability to handle ambiguity has been limited when it is available. More concentrated effort is needed to develop English-language interfaces to ontologies so as to make ontologies much easier to build and use than they are at present.    (O8W)

What I will propose in this talk is that the development of the Common Semantic Model (COSMO), an upper ontology being developed within the Ontology and Taxonomy Coordinating Working Group (ONTACWG), should be pursued in parallel with development of a natural-language interface integrating the COSMO with an English defining vocabulary. A starting version of the English defining vocabulary is available as the Longmans' defining vocabulary, used in the LDOCE dictionary. The COSMO should then be developed so as to support creation of the logical specifications corresponding to any term definitions created using the English defining vocabulary. The minimum size of the English and ontological-conceptual vocabularies required to support the flexible specification of meanings of terms in any domain is at present unknown, though based on the Longman's experience, it seems likely to be in the range of four to ten thousand distinguishable concepts. The actual required size will be discovered by the efforts to develop the ontological and linguistic defining vocabularies, and to apply them to specifying the meanings of specialized terms in many domains.    (O8X)

The English defining vocabulary can therefore be viewed as the linguistic form of the COSMO – though the mappings have yet to be developed. On the assumption that a mapping interpreter from the English defining vocabulary to the COSMO can and will be created, we can envision direct benefit from the immediate use of the defining vocabulary in creating definitions for terms in the Ontolog terminology or ontology. In the event that the complexity of the problem greatly delays creating the proper interpreter from English definitions to the logical forms, we would still be left with definitions of terms and concepts expressed in the basic English language that will make such terms easier for people to understand, regardless of their specialized interests.    (O8Y)

Dr. Patrick Cassidy received his Ph.D. in Chemistry from Harvard University in 1971, and started his professional career as a biochemist in the research labs at Merck pharmaceuticals. He became increasingly intrigued by the potential of computer intelligence, and took an early retirement from Merck in 1993 to devote full time to research in natural language understanding and knowledge representation. He has been involved in building ontologies since 1995. Since late 2004 he has been working at MITRE as a Senior Artificial Intelligence Engineer. Prior to that, he was an ontologist at VerticalNet, one of the pioneers in applying ontological engineering to eBusiness solutions. Dr. Cassidy has been playing an active leadership role in the ontology community, which include his founding and chairing of the ONTACWG (Ontology and Taxonomy Coordinating Working Group) of SICoP; as well as his key contribution in organizing the recent UpperOntologySummit. He considers it his mission to help at least one upper ontology become sufficiently widely adopted to attract vendors and academic groups, to use it and demonstrate the networking benefits of a common conceptual language for knowledge-based programs.    (O96)