Knowledge Compilation

Starting with Lipski's ideas about reasoning with deductive databases, there has been a growing trend aimed at addressing the inherent computational complexity of many reasoning problems by compiling knowledge bases and/or queries into restricted forms where query-answering is easier than in the original language. In this way, computational effort at compile time (presumably at the system's leisure) can be traded for quick response at query time. In general, these approaches also involve some form(s) of approximation, because the target language for the compilation is less expressive than the original language of the knowledge base.

CIRL

Etherington (CIRL) and Dalal (i2/Columbia U.) have developed a general framework for approximating logical deduction that admits the development of well behaved, tractable approximations. Theories and queries are compiled into theories in limited vocabularies for which the deduction problem may be easier. The resulting theories may be stronger or weaker than the original, providing bounds on answers to queries. A range of mechanisms is provided, allowing tradeoffs between computational effort and accuracy, as well as control of the direction of error (i.e., conclusions that are too strong or too weak). By allowing for the possibility of hierarchies of target vocabularies, we provide for the possibility of anytime approximations, in the sense that additional effort can be expended at query time (as time is available) to provide better answers.

Pointers

Hierarchical Knowledge Bases and Efficient Disjunctive Reasoning
This article describes a method for doing approximate reasoning efficiently on restricted types of information. Written by Alex Borgida and David Etherington, appeared in Proc. KR in 1989. Compressed postscript document.
Tractable Approximate Deduction Using Limited Vocabularies
This article presents a comprehensive theoretical framework for approximate reasoning. Written by Mukesh Dalal and David Etherington, appeared in Proc. of the Ninth Canadian Conference on AI, 1992. Compressed postscript document.
List of papers by Henry Kautz and Bart Selman
The authors, working at AT&T Labs, have written a number of papers on knowledge compilation.

Parent areas:
Tractable Reasoning

Subareas:
Polytime Simplification

Back to the CIRL Overview Page.