Specifically, it is concerned with control strategies governing the formation and refinement of partial hypotheses about the identity of an utterance that can guarantee the discovery of the best possible interpretation. We assume a system that contains the following components: a) A Lexical Retrieval component that can find the k best matching words in any region of an utterance subject to certain constraints and can be recalled to continue enumerating word matches in decreasing order of goodness (where possible constraints include anchoring the left or right end of the word to particular points in the utterance or to particular adjacent word matches).

The selection of what to do next is often the hardest part of resource-limited problem solving. In planning problems, there are typically many goals to be achieved in some order. The goals interact with each other in ways which depend both on the order in which they are achieved and on the particular operators which are used to achieve them. A planning program needs to keep its options open because decisions about one part of a plan are likely to have consequences for another part. This paper describes an approach to planning which integrates and extends two strategies termed the least-commitment and the heuristic strategies.

We have introduced the notion of the closed world assumption for deductive question-answering. This says, in effect, "Every positive statement that you don't know to be true may be assumed false". We have then shown how query evaluation under the closed world assumption reduces to the usual first order proof theoretic approach to query evaluation as applied to atomic queries. Finally, we have shown that consistent Horn data bases remain consistent under the closed world assumption and that definite data bases are consistent with the closed world assumption. ACKNOWLEDGMENT This paper was written with the financial support of the National Research Council of Canada under grant A7642. Much of this research was done while the author was visiting at Bolt, Beranek and Newman, Inc., Cambridge, Mass. I wish to thank Craig Bishop for his careful criticism of an earlier draft of this paper.

Presented at the NATO Symposium Human and Artificial Intelligence, Lyon, France, October, 1981. CMU Technical Report CMU-CS-82-104. We examine computer games in order to develop concepts of the relative roles of knowledge and search. The paper concentrates on the relation between knowledge applied at leaf nodes of a search and the depth of the search that is being conducted. Each knowledge of an advantage has a projection ability (time to convert to a more permanent advantage) associated with it. The best programs appear to have the longest projection ability knowledge in them. If the application of knowledge forces a single view of a terminal situation, this may at times be very wrong. We consider the advantages of knowledge delivering a range as its output, a method for which some theory exists, but which is as yet unproven.

Technical report CMU-CS-81-135, Computer Science Department, Carnegie-Mellon University. "This is a combination introductory and reference manual for OPS5, a programming language for production systems. OPS5 is used primarily for applications in the areas of artificial intelligence, cognitive psychology, and expert systems. OPS5 interpreters have been implemented in LISP and BLISS."

A four-volume collection of articles on all the major topics of AI at that time, with an extensive bibliography. Vol I (Avron Barr and Edward A. Feigenbaum, 1981) (https://books.google.com/books?isbn=1483214370). Vol II (Avron Barr, Edward A. Feigenbaum, Paul R. Cohen, 1982) (https://books.google.com/books?isbn=1483214389). Vol III (Paul R. Cohen and Edward A. Feigenbaum, 1982) (https://books.google.com/books?isbn=1483214397). Vol IV (Avron Barr and Paul R. Cohen, 1989) (https://books.google.com/books?isbn=1483214370). Reading, Mass.: Addison-Wesley.

"Practice, and the performance improvement that it engenders, has long been a major topic in psychology. In this paper, both experimental and theoretical approaches are employed in an investigation of the mechanisms underlying this improvement On the experimental side, it is argued that a single law, the power law of practice, adequately describes all of the practice data. On the theoretical side, a model of practice rooted in modern cognitive psychology, the chunking theory of learning, is formulated. The paper consists of (1) the presentation of a set of empirical practice curves; (2) mathematical investigations into the nature of power law functions; (3) evaluations of the ability of three different classes of functions to adequately model the empirical curves; (4) a discussion of the existing models of practice; (5) a presentation of the chunking theory of learning." In J. R. Anderson (Ed.). Cognitive Skills and their Acquisition (pp. 1-55). Hillsdale, NJ: Erlbaum.

GENERALIZATIONS BASED ON EXPLANATIONS Gerald DeJong Coordinated Science Laboratory University of Illinois 1101 West Springfield Avenue Urbana, IL 61801 This paper describes a new project at the University of Illinois in computer learning. The phenomenon under study is a kind of "insight learning" of procedural schemata. The system described here is designed to grasp some principle underlying a natural language input. Once acquired, the schema serves the same purpose as the other schemata in the system: it aids in processing future natural language inputs. The neutral term "schema" rather than "frame" (Minsky (1975), Charniak (1976)) or "script" (Schank A Abelson (1977)) is used to refer to knowledge chunks because a frame (which is used to describe static objects as well as progressions of world situations) is too general a notion, and the notion behind a script is-too specific.

A framework for inductive inference in logic is presented: a Model Inference Problem is defined, and it is shown that problems of machine learning and program synthesis from examples can be formulated naturally as model inference problems. A general, incremental inductive inference algorithm for solving model inference problems is developed. This algorithm is based on Popper's methodology of conjectures and refutations [II]. The algorithm can be shown to identify in the limit [3] any model in a family of complexity classes of models, is most powerful of its kind, and is flexible enough to have been successfully implemented for several concrete domains. The Model Inference System is a Prolog implementation of this algorithm, specialized to infer theories in Horn form.

Massachusetts Institute Technology Press