The Role of Reversible Grammars in
Translating Between Representation Languages
Jeffrey Van Baalen
University of Wyoming
Richard E. Fikes
Acquiring and representing knowledge is the key to building large and powerful AI systems. Unfortunately, knowledge base construction is difficult and time consuming. The development of most systems requires a new knowledge base to be constructed from scratch. As a result, most systems remain small to medium in size. The cost of this duplication of effort has been high and will become prohibitive as attempts are made to build larger systems. A promising approach to removing this barrier to the building of large scale AI systems is to develop techniques for encoding knowledge in a reusable form so that large portions of a knowledge base for a given application can be assembled from knowledge repositories and other systems.
For encoded knowledge to be incorporated into a system's knowledge base or interchanged among interoperating systems, the knowledge must either be represented in the receiving system's representation language or be translatable in some practical way into that language. Since an important means of achieving efficiency in application systems is to use specialized representation languages that directly support the knowledge processing requirements of the application, we cannot expect a standard knowledge representation language to emerge that would be used generally in application systems. Thus, we are confronted with a heterogeneous language problem whose solution requires a capability for translating encoded knowledge among specialized representation languages.
We are addressing the heterogeneous language problem by developing a translation technology for knowledge representation languages based on the use of an interlingua for communicating knowledge among systems. Given such an interlingua, a sending system would translate knowledge from its application-specific representation into the interlingua for communication purposes and a receiving system would translate knowledge from the interlingua into its application-specific representation before use. In addition, the interlingua could be the language in which libraries would provide reusable knowledge bases. An interlingua eases the translation problem in that to communicate knowledge to and from N languages without an interlingua, one must write (N-1)2
translators into and out of the languages. With an interlingua, one need only write 2*N translators into and out of the interlingua.
We consider in this paper the problem of translating declarative knowledge among representation languages using an interlingua with the following properties:
? A formally defined declarative semantics;
? Sufficient expressive power to represent any theory that is representable in the languages for which translators are to be built.
In practice, one cannot expect any given interlingua to have sufficient expressive power to support usable representations of any theory that is representable in any language.