Knowledge Interchange Format (KIF)

Tabled Issues


Tabled issues are issues that the ANSI Ad Hoc Group has discussed and decided not to consider further until some specific time in the future.

First-Order Semantics.

There has been extensive discussion of the expressiveness of KIF. Some communities advocate higher-order expressiveness. Others think that even first-order expressiveness is too great.

The Interlingua Committee on three separate occasions addressed this issue and three times voted for first-order expressiveness. The X3T2 Ad Hoc Group on KIF also addressed this issue and voted for first-order expressiveness. In a departure from the conclusions of the Interlingua Committee, the Ad Hoc Group group also recommended (1) that a study be made of ways in which the language might eventually be extended to higher order expressiveness and (2) that the design of the first-order language be made in a way so as to facilitate such extension in the future. On preliminary examination, it was decided that a single type of variable would suffice and that support for second-order would take the form of fixed meaning for appropriate type relations. The Ad Hoc Group tabled further discussion of this issue until a committee draft of the first-order version is completed and approved.


Nonmonotonicity

In its meeting in March 1995, the Ad Hoc Group voted to remove all support for nonmonotonicity from the language.

The argument in favor of preserving nonmonotonicity is the value of such constructs in realistic knowledge bases.

The argument against nonmonotonicity is that there does not appear to be any single widely accepted approach.

The issue may be re-opened when there is evidence of substantial convergence of opinion on the handling of nonmonotonic knowledge in the research community.


Prefix Syntax vs. Infix Syntax

Several people have suggested that the syntax of KIF be changed to more traditional infix form. The argument is that people are more familiar with math notation and/or C-like notation than they are with Lisp-like notation.

The current syntax is, in part, attributable to the preferences of the early designers of KIF, who were by and large Lisp programmers. However, there is also a technical reason for this choice, viz. the ease of writing parsers for this syntax. There is also an answer to concerns over readbility -- namely that KIF is intended for inter-computer communication, not human consumption. It has been argued that humans should use appropriate front-ends for editing information, programs that convert from KIF to math notation, natural language, graphical notations, and so forth.

One concession that has been made in this regard is the design of a standard infix format for presenting KIF and the development of tools to support this format and its translation to and from KIF. The use of this format in discussions and presentations does not, however, change the syntax used by computers. Issue tabled until everyone gets a chance to see the infix version and until there is substantial support for reconsideration.


Michael R. Genesereth , Stanford University, genesereth@cs.stanford.edu