System for revelance criteria management of actions and values in a rete network

Abstract

A method for programming a computer, and software products for implementing the method, are disclosed. According to the method, objects are created which, upon execution of the system, may initiate actions and may acquire values. A specification of relevance criteria is stored as an attribute of each object. Relevance criteria, which may refer to the values of other objects, describe all conditions necessary and sufficient to make the binary decision for whether the object will be considered relevant to the program's operation during execution. Action criteria may also be specified to control the actions of a relevant object. The method is particularly well suited for programming expert systems, but may be used for programming any type of computer program.

Claims

I claim:

1. A computing system including a processor and a data set, the data set comprising a plurality of objects, each object having:

(a) a data field which may contain a value, the value of the object, which value may be set to one of a plurality of possibilities by the processor,

(b) a relevance criterion which may be evaluated by the processor to yield a logical true or a logical false, and

(c) an action, in the form of processor instructions or a pointer to processor instructions, which will be executed by the processor if evaluation of the relevance criterion by the processor produces a logical true; and

(d) in one or more of the objects, the action of the object instructs the processor to set the value of the object to one of the plurality of possibilities, and,

(e) in one or more of the objects, the relevance criterion includes a reference to the value of another object such that the relevance criterion will or will not be satisfied depending on the value of the other object.

2. The computing system of claim 1 in which the action of an object comprises making a change to another object.

3. The computing system of claim 2 in which the change to another object is a change in the relevance criterion of the other object.

4. The computing system of claim 3 wherein the data set has a specified sequence of the objects beginning with a first object and ending with a last object, in which sequence the relevance criterion of each object makes no reference to the value of a later object in the sequence.

5. The computing system of claim 4 in which the change to another object changes the sequence of the objects.

6. A computer method for performing a computation, by use of a processor, on a data set comprising a plurality of objects, each object having:

(a) a data field which may contain a value, the value of the object, which value may be set to one of a plurality of possibilities by the processor,

(b) a relevance criterion which may be evaluated by the processor to yield a logical true or a logical false, and

(c) an action, in the form of processor instructions or a pointer to processor instructions, which will be executed by the processor if evaluation of the relevance criterion by the processor produces a logical true;

the method having steps comprising:

(d) determining the value of a first object,

(e) evaluating the relevance criterion of a second object, which relevance criterion includes a dependency on the value of the first object, and

(f) if the relevance criterion of the second object is satisfied, executing the action specified by the object and setting the value of the object.

7. The computer method of claim 6 in which the action of an object comprises making a change to another object.

8. The computer method of claim 7 in which the change to another object is a change in the relevance criterion of the other object.

9. The computer method of claim 8 wherein:

the data set has a specified sequence of the objects beginning with a first object and ending with a last object, in which sequence the relevance criterion of each object makes no reference to the value of a later object in the sequence.

10. The computer method of claim 9 in which the change to another object changes the sequence of the objects.

11. A computing system including means for creating or changing a data set which may be processed on a computing system with a processor, the data set comprising a plurality of objects, each object having:

(a) a data field which may contain a value, the value of the object, which value may be set to one of a plurality of possibilities by the processor,

(b) a relevance criterion which may be evaluated by the processor to yield a logical true or a logical false, and

(c) an action, in the form of processor instructions or a pointer to processor instructions, which will be executed by the processor if evaluation of the relevance criterion by the processor produces a logical true; and

(d) in one or more of the objects, the action of the object instructs the processor to set the value of the object to one of the plurality of possibilities, and,

(e) in one or more of the objects, the relevance criterion includes a reference to the value of another object such that the relevance criterion will or will not be satisfied depending on the value of the other object.

12. The computing system of claim 11 in which the action of an object comprises changing the relevance criterion of another object.

13. The computing system of claims 12 wherein the data set has a specified sequence of the objects beginning with a first object and ending with a last object, in which sequence the relevance criterion of each object makes no reference to the value of a later object in the sequence and the action of one or more objects changes the sequence of the objects.

14. A computing system for generating validation tables from a data set which may be processed on a computing system with a processor, the data set comprising a plurality of objects, each object having:

(a) a data field which may contain a value, the value of the object, which value may be set to one of a plurality of possibilities by the processor,

(b) a relevance criterion which may be evaluated by the processor to yield a logical true or a logical false, and

(c) an action, in the form of processor instructions or a pointer to processor instructions, which will be executed by the processor if evaluation of the relevance criterion by the processor produces a logical true; and

(d) in one or more of the objects, the action of the object instructs the processor to set the value of the object to one of the plurality of possibilities, and,

(e) in one or more of the objects, the relevance criterion includes a reference to the value of another object such that the relevance criterion will or will not be satisfied depending on the value of the other object;

the computing system comprising:

(f) instructions for reading the contents of an object, and

(g) instructions for generating a validation table from the contents of the object.

15. The computing system of claim 14 in which the action of an object comprises changing the relevance criterion of another object.

16. The computing system of claim 15 wherein the data set has a specified sequence of the objects beginning with a first object and ending with a last object, in which sequence the relevance criterion of each object makes no reference to the value of a later object in the sequence and the action of one or more objects changes the sequence.

17. A computing system for interpreting into natural language the contents of a data set which may be processed on a computing system with a processor, the data set comprising a plurality of objects, each object having:

(a) a data field which may contain a value, the value of the object, which value may be set to one of a plurality of possibilities by the processor,

(b) a relevance criterion which may be evaluated by the processor to yield a logical true or a logical false, and

(c) an action, in the form of processor instructions or a pointer to processor instructions, which will be executed by the processor if evaluation of the relevance criterion by the processor produces a logical true; and

(d) in one or more of the objects, the action of the object instructs the processor to set the value of the object to one of the plurality of possibilities, and,

(e) in one or more of the objects, the relevance criterion includes a reference to the value of another object such that the relevance criterion will or will not be satisfied depending on the value of the other object;

the computing system comprising:

(f) instructions for reading the relevance criterion and action of one of the objects, and

(g) instructions for interpreting the relevance criterion and action into natural human language, and

(h) instructions for adding one or more natural human language words to make one or more complete natural human language sentences.

18. The computing system of claim 17 in which the action of an object comprises changing the relevance criterion of another object.

19. The computing system of claim 18 wherein the data set has a specified sequence of the objects beginning with a first object and ending with a last object, in which sequence the relevance criterion of each object makes no reference to the value of a later object in the sequence and the action of one or more objects changes the sequence.

20. A computing system for examining a data set which may be processed on a computing system with a processor, the data set comprising a plurality of objects, each object having:

(a) a data field which may contain a value, the value of the object, which value may be set to one of a plurality of possibilities by the processor,

(b) a relevance criterion which may be evaluated by the processor to yield a logical true or a logical false, and

(c) an action, in the form of processor instructions or a pointer to processor instructions, which will be executed by the processor if evaluation of the relevance criterion by the processor produces a logical true; and

(d) in one or more of the objects, the action of the object instructs the processor to set the value of the object to one of the plurality of possibilities, and,

(e) in one or more of the objects, the relevance criterion includes a reference to the value of another object such that the relevance criterion will or will not be satisfied depending on the value of the other object;

the computing system comprising:

(f) instructions for reading information from the data set, and

(g) instructions for determining what steps a computer would take upon execution of the actions specified by one or more of the objects.

21. The computing system of claim 20 in which the action of an object comprises changing the relevance criterion of another object.

22. The computing system of claim 21 wherein the data set has a specified sequence of the objects beginning with a first object and ending with a last object, in which sequence the relevance criterion of each object makes no reference to the value of a later object in the sequence and the action of one or more objects changes the sequence.

Description

BACKGROUND

The invention pertains to methods for programming computers and the software tools for implementing these methods.

Early computer programming methods were limited to procedure based languages in which the actions of the computer were controlled by a long sequence from beginning to end. During execution of the sequence, control might be passed from one program to another which was specified by the first program and then control might be returned to the original program or might continue on to a third. In any event, control followed a sequence established by the programmer.

In order to model the reasoning processes of human experts, expert computer systems were devised which included long lists of facts and long lists of rules stating inferences that might be drawn if and when certain facts exist. When a rule is satisfied, consequences will typically cause the specification or updating of some of the facts. These changed facts will then cause other rules to become satisfied and the choice must again be made of which rule to apply next. In such rule-based systems, there are typically many different rules which are satisfied by any given condition of the set of facts and the application of one rule ahead of another will frequently cause a difference in the result or the speed at which it is achieved.

For this reason, and also because the numbers of rules are so large, applying every rule which is satisfied in each iteration of the system would be prohibitively slow, considerable research and experimentation has been devoted to developing "control strategies" and "inferencing methods" for determining which of the rules should be applied when. For a given set of facts and a given set of rules, different control strategies will produce different results or will reach the results with different efficiencies.

The creation of rule-based expert systems is a complex task, typically requiring a "knowledge engineer" to work with an expert to translate the expertise into long lists of potentially relevant facts and long lists of rules. Once the system is built, it is difficult to verify that the system will produce correct results in each situation and it is difficult to analyze the exact steps taken by the system to achieve each result.

Another architecture for modeling human expertise in computers uses "frames" or "objects" to represent items or classes of items in the real world, and then allows "children" of those frames or objects to be defined which inherit characteristics of their "parents" but are further differentiated from their "siblings" with additional information. Such a system is quite effective for representing taxonomic knowledge such as the method of organizing animal life into kingdoms, phyla, classes, orders, families, genera, and species.

SUMMARY OF THE INVENTION

The invention is a novel method of programming a computer system and the various software tools and components which implement this method. Although the method was developed to meet the need for improved expert systems, it has been discovered that the method is also useful for programming many other types of computer systems and is not limited to use for expert systems.

A fundamental concept of the invention is that of the "object": a necessary fact, a calculated result, a conclusion, or an identifiable element of the expert analysis that is being modeled. Experts do not use the term "objects", but they work with such objects all the time. When human experts use their expertise to solve a problem in the domain being modeled, they think by manipulating such objects. They ask questions to acquire necessary facts for their analysis; they calculate results; they come to intermediate conclusions, which affect their later reasoning; and they calculate or conclude final answers or recommendations. What makes them experts at solving such problems is that they have a long-standing familiarity with these elements of their analysis and they know the relationships among them. They know what questions to ask under what circumstances, and what consequences flow from which facts.

Although this description of the invention employs the word "objects", the meaning of this word has little in common with the "objects" of prior art expert systems.

The invention is designed to eliminate the "knowledge engineer", who serves in conventional expert systems as an intermediary between the domain expert and the system being built. The expert is typically not a computer programmer, and cannot be expected to know how to model his expertise in software, using conventional languages or expert system shells. The knowledge engineer, in the building of a typical expert system, forms this crucial link between the expertise to be modeled and the computer programming that is usually required in order to model it. The knowledge engineer interviews the expert to elicit the expertise and knowledge, and then applies his or her own computer programming expertise to represent that knowledge in the software. Unfortunately, there is enormous potential for error in this process, and it can be frustrating and expensive for all concerned.

The invention allows an expert to model expertise directly in the software. The object paradigm is intuitive enough that most experts take to it relatively easily, and can design their own systems first-hand. To create a knowledge base, one simply creates objects and then specifies the relationships among them in order to perform the analysis. The analysis happens in a specific sequence specified by the expert. This exactly mirrors one of the ways experts solve problems in the real world: they start by acquiring relevant facts, and then "reason forward" in some way from those facts to arrive at the consequences that flow from them. Each fact, consequence and all other factors which are useful in the course of the analysis become objects in the knowledge base.

During execution of a consultation, as facts are acquired from the user, some objects will turn out to be important, while others will be found to be irrelevant to the analysis and therefore will be ignored. In this way the system responds flexibly to different inputs and thereby effectively replicates in software the expertise of the designing expert.

Distinguishing features of the knowledge representation method are that:

(1) Action in the system takes the form of evaluating and firing objects, rather than firing rules as in prior art systems. Objects and their appropriate behavior, rather than rules and search strategies for applying them, become the central focus of system design.

(2) When an object fires, among other possible actions, it can set a value for itself.

(3) Knowledge is represented by objects and a description of what factors can affect the relevance and appropriate actions of objects. For most of the objects, the relevance criteria include references to values set for themselves by previously fired objects.

(4) The description of all such factors that could affect an object's relevance and appropriate actions are collected in only one place in the form of a set of evaluatable statements associated with the object, the object's behavior criteria.

A distinguishing label for this method is "relevance-actions-value based programming".

The knowledge being represented in the system thus takes the form of an enumeration of the instructions necessary to make a binary decision: whether to take action with respect to an object. The significance of that decision for the analysis being modeled lies in the inherent meaning of the object itself, as intended by the system designer, and in the inter-object dependency relationships that are created by these relevance criteria instructions.

There are important distinctions between the relevance criteria of the present invention and rules of the prior art. Prior art rules are organized in a long list, sometimes with groupings for ease of management, separately from the database of "facts" or "objects". The central question is, given the set of facts, which of many rules that could be applied should be applied? Different search strategies for choosing the rule to be applied can produce different results. New rules can be added at any time without changing existing rules.

In contrast, in the present invention, there is no search for applicable rules. Each object is considered in its turn and its relevance criteria are evaluated. Although the relevance criteria for a particular object can be modified, rules cannot be added to the system.

To translate a rule-based system of the prior art into the relevance-actions-value based system of the present invention, a description of relevant objects must be developed, which will not match one for one with the set of facts or objects in the rule based system, and then each of the rules must be examined to determine whether a portion of the rule should be reflected with an appropriate expression in the behavior criteria for one or more of the objects. With such a translation, some possible results of the original system would not be achievable.

Prior art systems devote processing to navigating a decision tree and reducing the system search space in order to make the system behave more the way we perceive human experts to work, and to achieve greater processing efficiency at run time. But the processing necessary to search rule conditions, navigate the tree, and reduce the search space, is work that the system must perform, and therefore consumes time and system resources. In this invention, a tradeoff is made by committing to examine all objects in the knowledge base in return for avoiding the navigational processing of conventional systems. Yet the same rich network of logic, knowledge, heuristics, and analytical dependencies are represented in the behavior criteria which determine the firing of objects. Performance considerations then focus on ensuring that objects are processed efficiently, particularly those objects that do not fire and are therefore bypassed.

The invention achieves numerous advantages over prior art expert system programming methods. Avoiding the problem of search strategies allows the system to run faster. The use of objects which have behavior criteria and which can be given a value for reference by other objects allows expert systems to be more easily programmed and debugged. The sequence the program will follow and the results that it will achieve are more predictable. The invention allows a consultation session to be easily restarted at any point. As discussed in the detailed description below, the invention allows validation tables to be constructed which allow for easy proof that the system has been correctly constructed.

Although the programming method was invented to meet the need for improved methods for building expert systems, it can be used for programming many other types of computer applications as well. It is particularly well suited to problems which involve multiple decision tree branches with interdependencies and where ease of building, debugging, and validating the system are relatively important compared to the processing speed of the resulting system.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 shows a sequence of objects containing the essential elements of objects in the invention.

FIG. 2 shows a simple form of object processing program which may be used with a data set comprised of objects according to this invention.

FIG. 3 shows the various computer programs, along with their inputs and outputs, which make up this invention.

FIG. 4 shows the various computer files which the object processing program of the invention uses to perform a consultation.

FIG. 5 shows the file structure of the knowledge base.

FIG. 6 shows the file structure of the generated data file.

FIG. 7 shows the data structure of the control strings.

FIG. 8 shows the data structure of the queue arrays.

FIG. 9 shows the data structure of the object values file.

FIG. 10 shows an overview flow chart of the object processing program in the preferred embodiment.

FIGS. 11a and 11b show a detailed flow chart of the object processing program in the preferred embodiment.

FIG. 12 shows a sample validation table as presented by the validation table generator component of the invention.

DETAILED DRSCRIPTION OF PREFERRED EMBODIMENT

Definition of terms

Application--An application is a set of one or more related expert-system modules that collectively form a conceptual whole and are designed to work together to represent the entire problem domain modeled by the expert system. A given application will have as its centerpiece a database file (the Object Data File 40) containing information about all objects used in the application, and information about each of the modules as well.

Module--A module is a subset of the defined objects of the application that forms a stand-alone expert system in its own right. Each module handles one conceptually distinct aspect or area of the expertise needed to be proficient in the overall application domain. In the preferred embodiment, modules are able to call other modules as subroutines in the course of their own processing.

Knowledge base 10--A knowledge base is the collection of data files in the invention that contains all of the data about the application. These files consist of the Object Data File 40, the Statement Data File 42, and the Expression Data File 44, together with their associated index files.

Object--Objects are the basic building blocks of a knowledge base. Each object represents a fact, question, conclusion, calculation, recommendation, or other material which can, under appropriate circumstances, be asserted as applicable in the reasoning process being modelled. If a piece of data needs to be captured during the course of a consultation, it will be represented by an object.

Attribute--Properties of objects are called attributes. According to the invention, the attributes of every object must include "relevance criteria" 1 and a specification of actions 2 to be taken if, upon evaluation, the relevance criteria are satisfied, one of which may be the setting of a value 3 for the object. In the preferred embodiment, every object also has a descriptive name, which is developer-supplied, and a unique identifying number (the "identifier"), which is system-supplied, as well as an object type designation and an attribute indicating that the object is a member of at least one module. Attributes for a particular object may vary from module to module, among the modules of which it is a member. For example, in one module an object may be an explicit question to be asked of the user, while in another module of the same application the same object might be an internal conclusion which concludes a value 3 based on the values 3 of other objects in the module.

Modules also have attributes, as does the application as a whole. These application-level attributes are stored in the Object Data File in application records, which have an object type of "A".

Object Value 3--Each object can be given a value which can then be referenced and used by other objects. Object values are stored in a file (see FIG. 9) created during a consultation.

Consultation 15--The process in which a user interacts with the developed expert system application is called a consultation. During a consultation, the system processes the objects entered into the knowledge base. Various objects ask for information from the user where necessary, and other objects draw conclusions based on the entered information. After all necessary data has been collected, the consultation typically continues by processing more objects and asserting conclusions or recommendations that are found to be applicable, based on the data collected.

Developer & User--The developer is the person (or group of persons) who creates the knowledge base in order to model the desired expertise. Usually this is the domain expert, the person who has the expertise being modeled. The resulting expert system application is distributed to multiple non-expert users, who benefit from the expertise by using the system.

Generated Data File 14--Having developed the knowledge base by creating all necessary objects and specifying all of their attributes, the developer generates the Generated Data File from the data in the knowledge base with the Generator Program 8. The Generated Data File is distributed to users together with the Object Processing Program 9, which uses it to run the analysis. The knowledge base itself remains with the developer and is not distributed. Modifications to the system are accomplished by altering the knowledge base, generating an updated Generated Data File, and distributing that new file to users.

Composition of an Application

The centerpiece of an application is the collection of files known as the knowledge base 10. The Generated Data File 14 is generated from these files, and the Object Processing Program 9 uses the Generated Data File in order to run the resulting system.

In the preferred embodiment, the invention is implemented in FoxPro, a microcomputer relational database management system. FoxPro is a variant of the "X-base" family of languages which originated with dbase. The fact that X-base is the de facto industry standard database language for micros means that the invention allows significant connectivity with other systems. This is enhanced by the invention's ability to call custom subroutines from within a consultation, to exchange data with standard external files, and to be called as a module from external programs, such as menu systems.

A consultation can be initiated from other programs by passing parameters which indicate the application number and module number for the consultation. For example, an overall menuing program might have a menu choice for each module in an application, other choices for modules of a different application, and still other choices for subroutines that are not consultations.

Several other types of files are usually needed in order to constitute the overall application in its final form. Typically there are other data files that are created by the developer to hold data for a historical record of prior analyses. These can be crucial to the correct performance of the system, since the invention allows values to be imported from such external files and assigned to objects in the course of processing a consultation. The invention also can take a consultation 15 (partial or complete) in its entirety, including answers supplied by the user, and save it in compressed form to a memo field in a particular record in such an external file. This allows consultations to be replayed at a later date.

In addition to external data tables, a finished application will usually include a file of custom subroutines which may be called by objects during the course of processing a consultation. The invention can open a channel to such a file in order to allow rapid access to such subroutines by opening it as a FoxPro procedure file at the start of the consultation 15.

Contents of the Knowledge Base

The knowledge base 10 consists of the objects within an application and their attribute data. The knowledge base holds all information about the objects, their characteristics, their interrelationships, the external files they interact with, and the customized subroutines that they may execute.

The knowledge base is contained primarily in the Object Data File 40. Each record in this file holds an object. Each object must have a name, one that should be intuitive to the developer. It will also have a system-generated, unique identification number (the "identifier") used by the invention. It will also have various attributes that distinguish it from other objects. One fundamentally important attribute of each object is its "relevance criteria" 1: the circumstances under which the object is considered to be relevant to the analysis. (An object that is determined to be relevant is said to "fire".) Other attributes include actions that an object may take, and action criteria which describe the conditions under which the actions should be executed.

Other files support the Object Data File 40 to form collectively the entire knowledge base. These are:

the Statement Data File 42, which holds unique statements used in relevance and action criteria, and

the Expression Data File 44, which contains all unique values that objects may acquire, as specified in the Object Data File, and unique left-side and right-side expressions used in criteria statements.

As shown in FIG. 5, the Object Data File 40 has the following structure:

    ______________________________________
    Field      Type        Length
    ______________________________________
    ID         Character    4
    NAME       Character   10
    ATX        Character   20
    ATT        Memo        10 (a pointer to variable-
                           length text in an associated
                           file)
    ______________________________________

    ______________________________________
    Number   Operator     Meaning
    ______________________________________
    1        =            equals
    2        #            does not equal
    3        >            is greater than
    4        <            is less than
    5        >=           is greater than or equal to
    6        <=           is less than or equal to
    7        $            is contained within string
    8        !$           is not contained within string
    ______________________________________

    __________________________________________________________________________
    Logical
           Left
               Left          Connecting
                                   Right           Right
    Operator
           Paren
               Expression    Operator
                                   Express.        Paren
    __________________________________________________________________________
    a)         {OBJECT1}     =     {OBJECT2}
    b)
      AND  (   {OBJECT3}+{OBJECT4}
                             >=    50000
    c)
      OR       {OBJECT5}     #     ({OBJECT6}+{OBJECT7})/2
    d)
      AND  (   {OBJECT8}/2   <     100000
    e)
      OR       {OBJECT9}     >     0
    f)
      AND      {OBJECT10}    =     YES             ))
    __________________________________________________________________________

    ______________________________________
    Pathway   Successful statements
    ______________________________________
    1 2 3     a,b a,c,d a,c,e,f
                                ##STR1##
    ______________________________________

    ______________________________________
                <statement 1, having identifier number: 0123>
    AND   (     <statement 2, having identifier number: 0234>
    OR          <statement 3, having identifier number: 0345>
                                           )
    ______________________________________

    ______________________________________
    From the Objects Data File:
    Attribute   Value
    Object name:
                MGMTAPPROV
    Object type:
                Conclusion
    Translation:
                Required management approval
    Relevance Criteria:
                9182    .rarw. pointers to Statements
    Possible Values:
                YES 0321.sub.-- (0432.0543).vertline.NO
                                  .rarw. Data File
    From the Statements Data File:
                       (encoded form, with pointers to the
    Statement ID
             Statement Expressions Data File)
    9182     975310987
    0321     065430765 .rarw. (left pointer: 0654
    0432     087610987 (operator: 3
    0543     123454321 (right pointer: 0765
    From the Expressions Data File:
    Expression ID
              Expression
    9753      {8163}    .rarw. object references are in curly
    0654      {0246}    brackets and use object ID
    0765      25000     numbers as pointers to the
    0876      {0357}    Objects Data File
    0987      YES
    1234      {0468} + {0680}/{5432}
    4321      {6543}
    From the Objects Data File:
    Object ID:
             8163
    Object name:
             STDTYPE
    Translation:
             Contract type is "Standard"
    Object ID:
             0246
    Object name:
             CONTVALUE
    Translation:
             Contract value
    Object ID:
             0357
    Object name:
             WARRANTINV
    Translation:
             Warranties are involved
    Object ID:
             0468
    Object name:
             HISTPAID
    Translation:
             Historical payments made
    Object ID:
             0680
    Object name:
             PENDORDER
    Translation:
             Value of pending orders including the
             current one
    Object ID:
             5432
    Object name:
             YRSHIST
    Translation:
             Number of years of this relationship
    Object ID:
             6543
    Object name:
             AVEANNUAL
    Translation:
             Average annual order value for all
             customers
    Standard operator translations:
    Number Operator Translation
    1      =        equal to
    2      #        not equal to
    3      >        greater than
    4      <        less than
    5      >=       greater than or equal to
    6      <=       less than or equal to
    7      $        contained in
    8      !$       not contained in
    The resulting natural language interpretation 12:
    The Conclusion object for required management approval
    will fire if:
    Contract type is "Standard" is equal to YES.
    A value of "YES" will be assigned to this object if:
    Contract value is greater than 25000
    AND     either warranties are involved is equal to YES,
    OR      the sum of historical payments made and the value of
            pending orders including the current one, when
            divided by the number of years of this relationship, is
            greater than or equal to the average annual order
            value for all customers.
    A value of "NO" will be assigned in all remaining cases.
    ______________________________________