EBL (explanation based learning), a framework for learning based on generalization from examples, has favorable properties for knowledge refinement in knowledge based problem solving systems. In the existing EBL, learning experiences have no coherence because each learning session is separated from the others. The purpose of knowledge refinement is to make problem solving process more efficient. Thus, in order to apply EBL to knowledge refinement, it is necessary to augment its framework to consider plural examples simultaneously and introduce appropriate operationality criteria. In this paper, the framework of EBL is augmented on plural examples, and considering relations between generalizations and operationality, a learning method is proposed to generate operational generalizations incrementally. The conceptualframework of augmented EBL is based on the generate-and-test paradigm with the generalization as a generator and the operationality as a test. This is formalized on the logic program. A generalization on one explanation structure is defined by an uninstantiation and an unresolution. This process is a generalization of the Mitchell's goal regression method. A macro table is generated from a generalization of a set of explanation structures for given examples. Operationality criteria are defined by two measures of the maximization of the usage degree and the minimization of the backtracking number. These measures reflect what is a useful macro table in pure-Prolog. The usage degree increases monotonically in the generalization space, but the backtracking number increases monotonically only in the n-usage subspace, which consists of generalizations that have usage degree of n. Therefore, minimal generalizations of the n-usage subspaces are useful to find operational generalizations. A concept of least EBG is introduced as such a generalization. There exists the least EBG for any set or explanation structures. It can be obtained by computing the least EBG of two explanation structures incrementally. They are minimal generalizations of n-usage subspaces. Thus, least EBG is usefull to find operational generalizations incrementally. A learning system with an incremental least EBG generator has been implemented. Its usefulness is demonstrated in the field of indefinite integration. It can generate complicated generalizations from a few examples, for which a huge number of examples are required in the existing SBL system. The macro tables obtained by this system are more compact and effective than those or learning systems that merely collect macros step by step.

To design intelligent and flexible computer interfaces the present paper develops models of a user which aim at guessing intention behind actions of the user. There are two types of user models : rule-based models and connectionist models. The author chooses the latter here, because they have various advantages over traditional rule-based models. Basic mechanism of the connectionist models being multiple constraint satisfaction, they go well with the requirement for guessing the intention of the user based on fragmentary information from various sources. Another advantage of the connectionist models is their flexibillty in learning and generalization. Developed in the paper are two kinds of connectionist user models. One is a model for typing error corrections of UNIX commands. The other is a group of models for extracting features from a sequence of UNIX commands. Various kinds of typing errors can be categorized into two : temporal and spatial errors. Representing these two kinds of typing errors on a network is the key to modelling the error correction problem. To represent the temporal ambiguities in typing a coarse coded layer and cross links between this layer and an input layer are employed. To represent the spatial ambiguities in typing special coding scheme is used for representing characters in a command : Hamming distance between a pair of characters be the same as the spatial distance between the corresponding pair of keys on a keyboard. Many pairs of erroneous and corrected commands are given to an input and output layers of the network as training patterns. It turns out that error correction capability of this model is not so good for short commands due to insufficient clue to restoration. To improve its performance this model assimilates other sources of information. The first one is the contextual information : history of commands and statistical features of the sequence. This, however, does not work satisfactorily due to large variation of statistical features even within the same user. The second source of information is the typing error characteristics of the user : frequency of each kind of typing errors. This information contributes to the restoration of erroneous typed commands. For extracting features from a sequence of commands, a model for obtaining command occurrence probabilities and models for command transition probabilities are developed using the so-called buffer model. These models are expected to suggest users candidates of useful macro commands. Much remains for further study, but the performance of the proposed connectionist models demonstrates their potential effectiveness in the user interfaces.

An interviewer has two kinds of knowledge. One is about a domain under consideration and the other is knowledge for interview itself which makes the interviewer an expert of interview. The later seems to be independent of the domain and to consist of several interview knowledge primitives. SIS, a Shell for Interview Systems is developed based on this idea. It has seven question strategy primitives as the primitive knowledge for interviewing. An interview system developer can construct several interview systems by describing the interview knowledge using the built-in primitives. In this paper, generality and effectiveness of SIS are shown through two implementaton examples of interview systems, I^2S-LD and MORE.

This paper describes a support system for database construction and manipulation employing dynamic analysis for acquisition of domain knowledge. The system consists of three parts, I^2S-LD (Intelligent Interview System for Logical Design), a knowledge base for database construction and an interface to database management systems. It interviews a user to design logical structure of the target database and constructs its prototype. Then, the system gets the user to use the prototype trying to detect problems of it. If some problems are found, the system redesigns and reconstructs the database. By employing the dynamic analysis, the system can support a user constructing a database efficiently.

Recently, many expert systems have been developed as the expert system building tools have come into the market. However, most tools are designed for the knowledge engineers and/or the programmers. This paper presents four environments to build expert systems based on the analysis of the expert system building process and its participants : (1) Experiment environment which allows experts to make a small knowledge base, (2) Development environment which allows knowledge engineers and programmers to design and implement the expert systems for practical use, (3) Operation environment which allows end-users to execute the expert systems, and (4) Maintenance environment which allows experts to refine the knowledge base. The development environment among them provides basic functions to build expert systems, which includes three primitives : OBJECT which represents knowledge, K-SET (Knowledge SET) which stores knowledge, and STAGE which is a working area to verify knowledge. OBJECTs interact with each other on STAGE by sending and receiving message, each of them is so modular that one can modify any OBJECT, respectively. STAGE is designed to shorten the turn around time for verifying the knowledge base. K-SET clarifies the source of knowledge and is responsible for security. Three kinds of view for knowledge base, defined by the knowledge engineers, are presented for other environments. In the operation environment, the knowledge operation view allows one OBJECT to join in more than one expert system. Knowledge maintenance view presents application specific user interface in the maintenance environment to refine knowledge base. Standard knowledge maintenance view (experiment view) acts in the experiment environment as if it were domain specific expert system building tool. The presented architecture supports whole life cycle to build expert systems. Once a new expert system is built by knowledge engineers, its knowledge maintenance views would be also used as the standard knowledge views in experiment environment for similar expert systems. If an expert system generated from the experiment environment is not appropriate for a new domain, then it would be modified in the development environment. The presented environments facilitate the enlargement of the computerization field in the business applications.

A user interface management system is presented with a new mechanism which reports keyin data completed to event handlers automatically. A problem has been that the user interface often costs much more man power to develop than the rest of the expert system. A user interface management system called UI Builder/UI Frame is developed to overcome this problem. UI Builder is a user interface development tool and UI Frame is a run-time support tool for the user interfaces developed by UI Builder. UI Frames are frames representing the user interface parts. UI Frame employes the mixed-initiative interaction mode, which consists of the program initiative mode and the user initiative mode. In the user initiative mode, user operations to the presented user interface are translated into messages to UI Frames, which invoke methods described to express corresponding actions to the operations. This translation makes user operation handling programs declarative, such as "when X is picked, execute Y." A new mechanism automatically sends messages to report input data in keyboard input fields at the time of switching from the user initiative mode to the program initiative mode. The mechanism enables highly modular program to handle key-in data and not to manage key-in field to read.

It is difficult even for an expert to fully understand the behavior of a complex system. Human being can move around the different level of abstraction based on the functional structure of the system, and focus on the phenomena of interest at the right level. Furthermore, s/he tries to identify the role of a part in a system to reduce ambiguity of possible behaviors in understanding a working mechanism of a system. S/he then identifies a set of elements as a meaningful composite element and regards it as a single entity. The way we understand a complex system is hierarchical. This paper discusses a hierarchical qualitative reasoning method which enables acquisition of heuristics and simultaneous utilization of knowledge expressed in multiple levels with different abstraction. The method makes use of equations which express the physical principles as the deep knowledge, and compiles them into the shallow knowledge expressing a representative behavior of a particular element or composite element. The shallow knowledge expresses the way an human expert understand a complex composite element, and thus, is heuristics. The method serves as a framework to formalize various heuristics which have so far been unable to handle without an experiential method. In carrying out these process, a new generalization method called an implicit-explanation based learning (IEBL) method was introduced and found useful. IEBL generates shallow knowledge from a set of given deep knowledge and an extended explanation. Use of this latter knowledge in addition to that used in the explicit explanation suppresses over-generalization which may otherwise be caused by the necessary approximations. The envisioning processes of a Schmitt trigger and a radio circuits are explained as examples to exhibit how the proposed method actually works.

Negation As Failure is the rule for the implementation method of negation which is mostly used in present logic programming languages such as Prolog. However, one of the fatal defects is that it can not bind variables appearing in a negative literal to other terms. Therefore, how to bind variables occuring in a negative literal also is an important problem for the implementation of negation. Negation Technique proposed by Dr. Sato is an effective method to deal with this problem. But this method is not applicable to the programs containing internal variables. In this paper, we discuss the possibility of eliminating internal variables appearing in logic programs by using program transformations. Firstly, we give a dynamic necessary condition and a static necessary condition for eliminationing internal variables and state their relation. Then, we prove two sufficient conditions for this purpose. We also illustrate several cases together with examples which satisfy these sufficient conditions. As a significant result, we find out that possessing folding correspondence is the necessary and sufficient condition for the purpose of transforming into non-internal variable logic programs. The minimal model of the non-internal variable program transformed is the same as the original one.