Artifical intelligence delivery system

Abstract

Disclosed is a system that provides an expert system's equivalent consultation without containing either the original knowledge base or the inferencing and logic shell of an expert system. System and process implementations provide a transformation and record of results from an original expert system's knowledge base which has been exercised through a full series of all possible consultations utilizing the expert system's own shell and knowledge base. The resulting transformation of a knowledge base is captured in the form of tabular histories of inquiries, responses and results and made available to a user via a table driver access mechanism. Results identical to those produced on the original expert system are obtained in a system containing neither the knowledge base nor the inferencing and logic mechanisms. The system provides an identical dialog of interaction with a human user and produces the same conclusions as the original expert system but requires only a small fraction of the computational and memory resources needed for running the original expert system.

Claims

What is claimed is:

1. A method of providing expert system response to user inquiries comprising steps on a computer system of:

accessing a recorded data base containing al original expert system user inquiries and user and system responses in their order of occurrence for all logically valid user inquiries, said recorded data base being generated on said original expert system using the said original expert system's logic and inferencing shell; and presenting in response to new user inquiries, when said new user inquiries correspond to any of said recorded inquiries, said recorded original expert system responses corresponding to said inquiries in their original sequence of occurrence.

2. The method of claim 1, further comprising a first step at an original existing expert system of:

transforming an original expert system's knowledge base into a transformed knowledge base by sequentially consulting said original expert system through all logically possible user inquiry and user response sequences and recording said inquiries and user responses and the original expert system's consultation responses thereto in their order of occurrence.

3. A method as described in claim 2 for modifying an original expert system's knowledge base records on a computer system, when records contain indicators defining the external sources from which expert system-requested input will be required and when said records also contain indicators defining the receivers for said expert system's output responses, further comprising steps of:

electronically reading said knowledge base records to discover said recorded indicators therein; and

recording in said knowledge base records new indicators for each said indicator discovered, said new indicators controlling said transforming step to access, for any defined external source of expert system-requested input, an internal process in said computer system and controlling said transforming step to provide any expert system output responses to another internal process in said computer system as the receiver of any said expert system output responses.

4. An expert knowledge delivery system having a system logic and inference shell and means for providing output information in response to a user's inputs and inquiries for information regarding a subject for which said expert knowledge delivery system contains electronically recorded and electronically readable knowledge, said delivery system comprising:

an electronically recorded and electronically readable knowledge base, said knowledge base containing a sequentially arranged electronic record of all logically possible valid user inquiries, inputs and the original expert system's responses thereto, said recorded user inquiries, inputs and original expert system's responses being recorded and arranged in said knowledge base in the sequence generated by exercising said original expert system's logic and

electronic access and presentation means for accessing said electronic knowledge base and presenting, as expert knowledge delivery system information output to a user, portions of said recorded responses.

Description

FIELD OF THE INVENTION

This invention relates generally to artificial intelligence expert systems which accept inquiries for information and provide responses thereto. More specifically, the invention relates to a method and apparatus for providing, in a system having neither the knowledge base nor the inferencing and access driving mechanism of an actual expert system, responses equivalent to those provided by a full expert system.

BACKGROUND OF THE INVENTION

Prior Art

In the field of artificial intelligence, expert systems have come to be well known. These are systems that simulate a human expert's abilities. One type of such a system conducts an interactive consultation with a human client to address the client's needs, concerns or problems. The system and its program provide the assistance and advice that would be provided by a human expert in response to the same consultation questions and inquiries. Systems of this type are being increasingly used in many different applications including complex system controls, network management, hardware and software problem determination, computer system configuring, medical diagnosis, mineral exploration and other "knowledge domains" where human expertise is generally required.

These systems normally utilize a controlling program for analyzing a client's inquiries and providing access to the system's knowledge base that contains responsive queries and various answers and data. The controlling program provides access to the knowledge base utilizing inferential capability and logic. The inferencing and logic means is generally contained in a "shell" to which the knowledge base for the user's domain of interest is added.

Expert system shells access the domain's knowledge base to provide responses to a client's inquiry either in the form of a question to be answered by the client in order to elicit enough information to provide an answer, or eventually provide an answer or data that will suit the client's request. The controlling shells are designed to work with the knowledge bases for the various knowledge domains that they are to access and utilize. The development of such shells is a highly time consuming and costly task. As a consequence, generalized shells that operate on more generalized knowledge bases have been developed. These general control system shells have come to be known in association with various areas of expertise. Several commercially available shells exist and can be purchased or licensed for use in a variety of knowledge base domains. Some examples are EMYCIN which was developed by Stanford University and which has been useful with knowledge bases in the field of bacterial disease. Another commercially available shell is M.1 developed and licensed by the Teknowledge Corporation. Yet another expert system shell is known as the Expert System Environment developed by the IBM Corporation as program number 5798-XXX. Expert System Environment is available in a number of versions depending upon the computer operating system on which it is desired to run. A general information manual GH20-9597-1 is available and a reference manual SH20-9609-2 and user's guide SH20-9608-1 exist and can be purchased separately from the IBM Corporation for a greater understanding of its existing shell.

The operation of each of the "shells" is somewhat similar but differs with respect to the particular computer or operating system in which it is intended to run and with respect to the type of knowledge that it can most conveniently process.

An expert system shell operates on a knowledge base. The development of a knowledge base or a given domain of expertise is a tedious, costly and time consuming task. The knowledge base itself is structured according to the operating details of the specific shell with which it is intended for use. Human expertise in the particular domain of knowledge is required to construct a knowledge base. A human "domain expert", whose data may be augmented by human textual material that may be managed or made available by human beings or by computer retrieval systems or the like, is employed to help create the knowledge base. The domain expert, of course, needs to have the details and/or instructions of how to build the specific knowledge base for the particular "shell" that is to be used. The shell vendors provide the instructions and also provide sample knowledge bases to facilitate construction of specific knowledge bases by domain experts who intend to use the vendor's shells.

For example, MYCIN is the commercial name of an expert system that provides consultations on bacterial disease. MYCIN was developed at Stanford University and is described in the book entitled, "Rule Based Expert Systems" by Buchanan and Shortliffe.

A problem exists in this field of which is due to the complexity of expert systems and the amount of storage required. An expert system itself is the combination of the inferencing and access logic means, generally referred to as the "shell", and the knowledge base with which the shell is to be used. Such systems require substantial amounts of active computer memory for processing and generally require a high speed logic system for their execution and/or development. This is true partially because each shell contains an inferencing mechanism that requires substantial portions of active memory for storage, analysis and formulation of further logical alternatives and inquiry paths in response to specific user inquiries. The large storage requirements are also due to the massive contents of any specific knowledge base.

This problem becomes acute and raises substantially the cost of using replicated expert systems, a type of usage that is often desired since access to the available expertise of the system is often needed at a variety of locations. Thus, expert systems are usually not available at all without access to large computer system resources. Potential users of the expertise within a given knowledge domain managed by an expert system of this type often have dispersed computer or data processing operations and do not have at each location the amount of computer system resource that is required. Each individual location at which access to consultational execution of centrally developed expert systems might either be necessary or desired may not, therefore, be able to obtain such services. Even the users that do have access to such major system resources may desire to conduct their consultation at some other more remote location where the computer resource is not available. Additionally, some computer environments differ from those on which the expert system may have originally been developed and thus the expert system cannot be used at all.

It might be useful, therefore, to attempt to provide an expert system that is available for clients without regard to the type of computer system on which the expert system will be operated. This would require that the users have a computer system that can run the "shell" used for implementing and developing the expert system. This system need not necessarily be the specific computer system for which the shell that they intend to use was developed. However, such "universal" expert systems are not available due to the aforementioned problem of a great investment in system size and resource that is necessary.

It is also apparent that a great deal of expense and time will be utilized in developing specific knowledge bases and shells. The resulting cost of delivering a full expert system to a human client is therefore prohibitively large in many instances. It would be desirable if the full benefit of expert system consultation could be made available to end users without exposing either the knowledge base or the expensive inferencing mechanism and logic system in the "shell" portion. It is the shell that combines the known pieces of information, decides what other information to seek and which uses all of those resources available to produce a valid consultation result in the original expert system. Thus, the inferencing methodology of a KB/shell combination is a highly valuable piece of system control programming of which exposure should desirably be limited or prevented, if possible.

In the foregoing cases, it would be most desirable if a "delivery vehicle" that does not expose the interactions of the knowledge base and the shell mechanism of the original expert system could be provided. Such a "delivery vehicle" would be any means or technique that permits dissemination of or delivery of or access to the available expertise in the expert system without exposing either the expert system shell or all of the specific knowledge base. For example, it would be most desirable to provide the capability of running consultations against the expertise of an expert system on small personal computers without exposing either the original knowledge base or the inferencing shell mechanism to duplication and/or discovery by the users.

OBJECTS OF THE INVENTION

In light of the foregoing known difficulties with the development of expert system delivery mechanisms and inquiry systems, it is an object of this invention to provide an improved means of supplying expert system consultation services without exposing the knowledge bases or the inferencing shell mechanisms to the user.

Yet another object of the invention is to provide an improved expert system equivalent that can be run in small computer systems but which does not contain either the original knowledge base or the inferencing or logic (known as the shell) that provides access to and interpretation of the contents in the knowledge base.

BRIEF SUMMARY

The present invention addresses the above requirements and constraints by providing a transformation and capture of the knowledge base. This may be captured in the form of tables or otherwise. The transformation process is applied to the original expert system's full series of all possible consultations utilizing the expert system's knowledge base. The resulting transformation and capturing of the consultations in the form of tabular histories is made available to a user by providing a table driver or access means. The tables of the transformed expert system consultations and knowledge base may be provided in a diskette or other similarly portable media for operation in a small computer system. The table driver constitutes an access means or program that can access the tables of information that have been transformed from their original expert system. As such, the small computer, together with the transformed and captured results of the full expert system, constitutes an "equivalent" expert system.

The invention provides the equivalent to the full expert system responses for consultations. Results identical to those produced by the original expert system are obtained in a system that does not contain either the knowledge base or the inferencing and logic mechanisms. The equivalent expert system provides an identical dialog of interaction with the human user and produces the same conclusions as the original expert system. In our invention, all logically possible consultations have been previously executed and captured in the form of tables during the transformation process conducted on the original expert system. In our invention, the transformation is accomplished utilizing transformation programs that interact with the knowledge base and with temporary data files constructed for intermediate storage. The tables or files which result may then be delivered as the transformed knowledge to be used in a user's system where they may be accessed by a table driver program. In this way, only a small fraction of the computational and active memory resources needed for running the original expert system are required for a user to perform an equivalent consultation on a much smaller system.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention is further described with reference to several figures depicting aspects of a preferred embodiment thereof in which:

FIG. 1A illustrates the generally known configuration in the prior art of an expert system.

FIG. 1B illustrates in schematic form the equivalent expert system of a preferred embodiment of the invention.

FIG. 2A illustrates schematically the knowledge base modification process and apparatus.

FIG. 2B illustrates schematically a modified expert system resulting from the use of a modified knowledge base by the existing expert system shell.

FIG. 3 illustrates schematically the arrangement for providing a transformation process utilizing transformation programs to convert and to capture all logically possible consultations and their results utilizing the modified expert system.

FIG. 4 schematically illustrates the interaction during the transformation process that results in creating the transformed knowledge base for use in the present invention.

FIG. 5 illustrates in a schematic fashion the program or processing steps utilized in performing the transformation process.

FIG. 6 schematically illustrates details of a portion of the transformation process from FIG. 5.

FIG. 7 schematically illustrates details of another portion of the transformation process from FIG. 5.

FIG. 8 schematically illustrates further details of the transformation process from FIG. 5.

FIG. 9 illustrates schematically an equivalent expert system according to a preferred embodiment of the invention utilizing the transformed knowledge base together with a table driver program which is schematically shown also. The table driver program runs on the user's system.

FIG. 10 illustrates an example of a path's table resulting from the transformation process and used by a table driver program.

DETAILED SPECIFICATION

As outlined briefly in the summary above, our invention includes novel methods and mechanisms for modifying an existing expert system's knowledge base and for transforming the modified knowledge base by exercising the system through all logically possible sequences of inquiry and response and capturing the resulting responses. This results in a sequential history of inquiry and response patterns or findings that do not themselves contain either the knowledge base or the original expert system's inferencing and logic shell means. The resulting captured output of responses can be accessed by a table driver access means or program. The access program can find an individual inquiry and its associated response or responses that would have been generated by the original expert system utilizing its inferencing and logic shell means and its original knowledge base. From the user's viewpoint, the "system" that is accessed is an "equivalent expert system" since it provides equivalent results, but it does not contain either the original knowledge base or the original expert system's shell. These highly expensive knowledge base and shell components of an expert system are thus not exposed to duplication or discovery of their unique contents. However, the effective result of using these expensive components in any given consultation can be achieved for the user without exposing either of these costly elements.

Our invention includes novel means and methods for achieving the transformation of the original expert system's knowledge base. These are included in subsections below generally describing the preparatory modification process and then the actual transformation process. Together, the modification and transformation process use the implicit logic tree contained in the pattern of legitimate logic decision points made during real consultations on their original expert system to create results in a form that contains the real logic tree. The real logic tree is embodied as the logical and sequential record of all legitimate inquiries and responses as found in the original expert system. Our process captures the real logic tree and renders it in a usable form without exposing either the original knowledge base or the original expert system's shell. Our transformation method and apparatus effectively drive an original expert system sequentially through all possible logically valid consultations and capture the expert system's data requests, valid responses and the sequences of these requests and responses as they lead to a final consultation which itself is captured as well. The overall result may be called a "transformed expert system" which is captured in a storage medium such as a diskette and used with a table driver access program to provide a "virtual expert system" delivery vehicle for a given user. The user employs a table driver or access means to gain access to the tables of results utilizing a small computer as his "equivalent" or "virtual" expert system delivery method and apparatus.

A table driver program is used to access the captured tables of the transformed expert system and to properly interpret and deliver the results of the data found therein. Such a table driver presents the expert system's data requests to the client in the sequence that would normally occur if a client were using the actual original expert system. The client's responses are used by the table driver program to properly access the next indicated result or action and to make an information presentation to the user. The information presented may be a conclusion, intermediate advice, or a request for more data. The consultation continues and finally completes on an equivalent expert system just as it would have on the original expert system: It provides a recommended answer or solution for the client in response to his inquiry.

Of particular value of our invention is the ability to generate table driver programs for delivering in any target computer system, the captured result tables from the transformed knowledge base. This constitutes for the user an equivalent expert system that may be consulted utilizing his own small computer.

Our transformation process includes a step of driving the original expert system through all logically possible consultations without the need for human interaction. Our process first modifies the knowledge base. It alters the contents and controls of the original knowledge base during the modification step so that the knowledge base can interact with automatically executed computer program inputs rather than with a human being. In our system, after the modification of the knowledge base has been completed, a transformation process is carried out to generate and capture the necessary history of a completely exhaustive consultation. In our preferred embodiment of the invention, the information is captured in the form of tables. The process that we utilize controls an exhaustive execution of all of the logically possible consultations of the modified knowledge base of the original expert system, including deciding how to proceed and recognizing when the process has exhausted all of the consultations that are possible.

The modification process used on the knowledge base of the original expert system will now be described.

Modification of Knowledge Base

A typical rule-based prior art expert system, such as that shown in FIG. 1A interacts with a human client within a particular domain of knowledge. During the transformation process in our invention, it is desired to exhaustively execute all logically possible consultations of the original expert system within its knowledge domain without the need for a human being to interact with it. A preliminary change, prior to the actual transformation of the knowledge base is necessary to allow removal of the human interaction. In our invention, this preliminary change is achieved by a process which we have called a modification process. The preliminary change is necessary to automate the execution of all possible consultations during the transformation process.

The modification process changes the embedded controls of the expert system knowledge base so that they no longer require interaction, i.e., responses from a human being. Instead, the modified knowledge base enables interaction to occur with automatically accessed computer files.

The changed or modified knowledge base, in combination with the appropriate original expert system inferencing and logic "shell" is referred to herein as a modified expert system. The modified expert system is shown schematically in FIG. 2B. In order for the modified expert system to operate in the same fashion as the original expert system, it is required that the knowledge content and the way in which that knowledge is processed be unchanged in any way from the original. This may be accomplished without exposing either the original knowledge base or the original inferencing shell as will be shown in the following example.

As an example of modifications done using part of a knowledge base in a domain of knowledge that can be consulted using IBM's previously noted shell, we will consider the following. It will be understood, for example, that the process is identical in many rule-based expert system utilizing its related knowledge bases and shell.

Some definitions are in order at this point. For example, a knowledge base is composed of a "domain of knowledge" of facts and reasoning documented in the form of "parameters" and "rules". These constitute instructions on how to process the knowledge and instructions that specify with whom or where an expert system is to interact to obtain its inputs and to provide its outputs. For the purposes of this discussion, "parameters" are defined as variables of the knowledge domain to which values need to be assigned. During a consultation by a human client, a parameter and its corresponding assigned value are together referred to as a "fact". "Rules" are a relation between "facts". Rules and parameters together are referred to as "entities" in this description.

The instructions for processing the knowledge in the knowledge base and for controlling the system's interaction with a user and for indicating where to obtain inputs and to provide outputs are all contained within the "entities" of the knowledge base. The entities determine how the expert system is to request needed information and/or to provide advice and other information to the client. For example, requested information might be provided by a client through a terminal keyboard and correspondingly, advice or answers might be provided to the client through a terminal display.

A knowledge base modifier 3, as shown in FIG. 2A, conducts the modification process. It generates a modified knowledge base 4 and a series of captured input request sequences in the form of an Input Request Data Table 5 by operating upon the original knowledge base 2. The required actions of the modification process are as follows: first, it is necessary to change the knowledge base at all points where sources of requested information are recorded so that no human input is required by the system during the subsequent transformation process. Second, it is necessary to change to a single output all of the indicated receivers of output, advice or other information as contained in the instructions in the knowledge base 2 and to thereby bypass the need for human interaction. An intermediate data file must also be built to facilitate execution of the transformation process. Several of these intermediate files are built during the modification process and include an Input Data run time file and an Input Request Data Table 5, the latter of which is shown in FIG. 2A. The run time file is included generally as any one of several files contained within box 6 in FIG. 3. These several functions and file constructions will now be described in greater detail.

Change All KB-Designated External Sources of Requested Information

During a consultation of the original expert system, the system will attempt to assign values to parameters to be used in combination for reaching a conclusion. How the parameters are assigned values depends upon the source of requested information for each parameter as specified by the knowledge base. The source for the parameter is defined in the knowledge base 2 in FIGS. 1A or 2A. Types of sources might be a human client, a computer program communicating with the expert system, a default value defined in the knowledge base itself and/or assigned by the expert system. Alternatively, the expert system itself may determine a value by access to and utilization of rules that can be processed or which may be inferred from the logical processes involved. Any parameter whose value is to be assigned by the expert system's operation itself is considered to have an "internal" source. A parameter whose value is not to be assigned by the expert system is considered to have an "external" source. Our modification process changes the knowledge base entries that specify sources so that all external sources are changed to be the same defined source, namely a program of a given name. The named program will be invoked during our transformation process. Parameters with internal sources are not changed in any way in our preferred embodiment.

As an example, consider the following. A parameter entity entry in a knowledge base before modification might look like this:

    ______________________________________
    Name       Direction
    ______________________________________
    Prompt     In what general direction are you heading?
    Legal Values
               North, South, East, West
    Source     Client will input from terminal.
    ______________________________________

    ______________________________________
    Name        Direction
    ______________________________________
    Prompt      In what general directions are you heading?
    Legal Values
                North, South, East, West
    Source      External data program
    Program Name
                EX IN
    Program ARGS
                Direction
    ______________________________________

    ______________________________________
    Fall-- and Season =
    Then SHOW -You should be able to see the Big Dipper in the sky
           45.degree. above the horizon-
    ______________________________________

    ______________________________________
    Fall-- and Season =
    Then PROCESS -You should be able to see the Big Dipper
    in the sky
           45.degree. above the horizon-
    USING EX-OUT
    ______________________________________

    ______________________________________
    Name       Direction
    ______________________________________
    Prompt     In what general direction are you heading?
    Legal Values
               North, South, East, West
    Source     Client will input from terminal.
    ______________________________________

    ______________________________________
    Name       Direction
    ______________________________________
    Prompt     In what general direction are you heading?
    Legal Values
               North, South, East, West
    Source     Client will input from terminal.
    ______________________________________

    ______________________________________
    /1.1            /35.1  /36.2
    ______________________________________

• Inventors
  Gordon, O. Michael; Hubbell, John R.; Woodland, Norman J.;
• Assignee
  International Business Machines Corp. (Armonk, NY)
• Published
  Dec-28-1993
• Current US Classes:
  706/45
  706/50
  706/59
  707/3
• Application #
  584404
• International Classes
  G06F 015/40
• Field of Search
  364/513 364/200 395/600 395/700 395/50 395/54 395/75 395/60
• Examiner
  Kriess; Kevin A.
• Agent
  Duffield; Edward H.
• US Patent References:
  4636974
  4675829
  4704695
  4763277
  4809219
  4815005
  4835690
  4841441
  4866634
  4866635
  4884218
  5021992