JPH0877014A - Knowledge acquisition processing system - Google Patents

Abstract

PURPOSE: To provide the knowledge acquisition processing system which acquires accurate attribute data possessed by an object in the knowledge base of the object. CONSTITUTION: This system is equipped with the data base 10 which is used for intellectual activities and manages attribute data that objects have and scenes so that they are made to correspond to each other, domain by domain, common knowledge 11 which manages essential attributes and attached attributes that the objects have, a mechanism 12 which retrieves the essential attribute that a specified object at the time of an intellectual activity, a mechanism 13 which specifies an object having the retrieved essential attribute as an attached attribute among objects present at the time of the intellectual activity, a mechanism 14 which registers the acquired essential attribute corresponding to the scene that the specified object has when the scene of the intellectual activity matches the scene that the specified object has, and a mechanism 15 which generates the scene of the intellectual activity for the object and registers the acquired essential attribute corresponding to it unless the two scenes match each other.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a knowledge acquisition processing system for acquiring attribute data of an object and registering it in an object knowledge database, and more particularly to acquiring accurate attribute data and registering it in an object knowledge database. Therefore, the present invention relates to a knowledge acquisition processing method for realizing the construction of a knowledge database that can effectively support intellectual activities.

[0002]

2. Description of the Related Art Conventional databases are composed of relational databases. As is well known, this relational database employs a structure for managing data in a tabular format without bringing in a specific structure. This is because if you adopt a structure that manages data by bringing in a specific structure such as a tree structure, you may be able to easily search for specific information but not for other information. .

In this relational database, the concept of a primary key and a subordinate key is introduced as a search key, but this cannot sufficiently express the attributes of data. From now on, as shown in FIG. 38, the metadata which is the attribute data of the data is managed by the DD / DS, and the metadata is associated with the user data managed by the DBMS. It's coming.

[0004]

However, the metadata managed by this conventional technique is only management data such as who created the data and when. That is, the meaning of data is not managed.

Moreover, even the same data has different meanings in different domains (fields), and even the same domain has different meanings in different scenes. The prior art does not even consider such a thing, as can be seen from the fact that even the meaning is not managed. From now on, even if a configuration for managing the meaning of data according to this conventional technique is adopted, the meaning must be inaccurate.

[0006] Therefore, there is a problem that the conventional database cannot sufficiently support intellectual activities. Against this background, the present inventors here
We have proposed a new object-oriented data processing system. Next, the object-oriented data processing system proposed by the present inventors will be described.

In order to support system design, etc.,
How easy and faithful to model the real world is important. Therefore, the present inventors have paid attention to the fact that the human words that we use are "extended interpretations that give conceptual recognition," and O-id (extended phrasing or Introducing the object command), the object thinking design method for designing an object by this extensional name (since this design method constructs a virtual human, we dare to call it an object-oriented Called thinking).

[0008] With this, it is possible to construct an object model in the same sense as in the way we usually grasp phenomena in the real world, and to create a world that can be called a virtual human world by bringing the human world into the world of objects. became.

In this object-thinking design method, the object in the real world is divided into the world of extensional nouns and the world of information-hiding comprehension. The world of this extensional name is called the object world. When the object world is classified in terms of schema, it is roughly classified into three: "world of static model", "world of dynamic model", and "world of causal relationship model". The static model in the object world indicates a relationship that exists at the same time even if the modeling target time is stopped, and the dynamic model in the object world exists by moving the modeling target time. The causal relationship model in the object world represents a temporal relationship in a dynamic model developed from a static model.

As described above, the static world, the dynamic world, and the causal world composed of extensional nominatives are so-called "composite worlds by nominatives" in which nominatives are related by pointers or messages. Has become.

In general, human thought is used by giving more properties or names to these "complexes". Once defined, we can then construct the object world by its nomenclature. At that time, the constituents and definitions of the complex do not need to be deeply recognized in the object world, that is, the virtual human world, and are performed by extensional interpretation.

On the other hand, the information hiding model in the world of information hiding is a definition resulting from the static world, the dynamic world, and the causal world in the above-mentioned world of extensional nomenclature,
Atomic object (which is the basic unit of an object) is a model of a substance or the entire content at the level. The only interface between this information hiding model and the object world (virtual human world) is taken as an extensional name.

The object thinking data processing system newly proposed by the present inventors based on such a technical idea is as follows.
Since it is based on a completely different way of thinking from conventional software technology, adding a new function makes it possible to construct a knowledge database that can effectively support intellectual activities.

The present invention has been made in view of the above circumstances, and it is possible to construct a knowledge database capable of effectively supporting intellectual activities by acquiring accurate attribute data and registering it in the object knowledge database. The purpose is to provide a new knowledge acquisition processing method to be realized.

[0015]

FIG. 1 shows the principle configuration of the present invention. In the figure, reference numeral 1 denotes an object knowledge database provided with the present invention, which manages knowledge objects. The object expanded in the object knowledge database 1 has an extensional name (object command) derived from a combination of the symbol compression body of the object attribute and the iD information indicating the substance of the object as the internal iD. Reference numeral 2 is a terminal connected to the object knowledge database 1.

The object knowledge database 1 is a domain knowledge database 10 for implementing the present invention.
, Common knowledge 11, search mechanism 12, and identification mechanism 13
And a first registration mechanism 14 and a second registration mechanism 15.

The domain knowledge database 10 is used for intellectual activities defined by a domain, and manages attribute data of an object for each domain in association with a scene in the domain. The common knowledge 11 is prepared as a knowledge source of the objects registered in the domain knowledge database 10, and manages the attributes of the knowledge object without distinguishing the domain while distinguishing the essential attribute and the auxiliary attribute. To do.

The retrieval mechanism 12 retrieves the essential attribute of the object designated during the intellectual activity. The identifying mechanism 13 identifies, among the objects existing during the intellectual activity, one having the essential attribute retrieved by the retrieval mechanism 12 as an ancillary attribute. The first and second registration mechanisms 14 and 15 are
Attribute data is registered in the objects managed by the domain knowledge database 10 while being associated with scenes in the domain.

[0019]

In the present invention, when a user operating the terminal 2 is performing an intellectual activity while displaying an object of a certain domain registered in the domain knowledge database 10 on the display screen, the user uses the certain object. If an object does not exist in the scene when some operation is performed on the scene, the search mechanism 12 searches the common knowledge 11 using the name of the object to search for the object. Search for essential attributes that you have.

For example, when the paper is about to fly, and when the paperweight is used to hold the paper, the paperweight does not exist in the scene (paper scene). The search mechanism 12 searches the common knowledge 11 to search for an essential attribute of the paperweight, which is "a thing that becomes a weight".

When the retrieval mechanism 12 retrieves the essential attributes,
Subsequently, the identifying mechanism 13 identifies, among the objects existing during the intellectual activity, one having the retrieved essential attribute as an ancillary attribute.

For example, when there are a cup, a desk, and a desk lamp as objects that exist during an intellectual activity in which the paper tends to fly, among them,
It identifies the ones that have the attribute "weight" as ancillary attributes. Thereby, for example, the cup is specified.

When the specifying mechanism 13 searches for an object, the first registration mechanism 14 judges whether or not the scene of the intellectual activity and the scene of the object specified by the specifying mechanism 13 match, and matches. When it is determined to do so, the essential attribute searched by the search mechanism 12 is added to the object managed by the domain knowledge database 10 in association with the scene.

For example, when the scene of intellectual activity is a “paper scene”, the domain knowledge database 1
The specified cup of the specific mechanism 13 managed by 0 is
When a "paper scene" is included, the attribute "weight becomes" found by the search mechanism 12 is added in association with the "paper scene".

On the other hand, when the specifying mechanism 13 searches for an object, the second registering mechanism 15 detects the scene of intellectual activity,
When it is judged whether or not the scene of the object specified by the specifying mechanism 13 matches, and when there is no match, the scene of the intellectual activity for the object managed by the domain knowledge database 10 is judged. Is generated and the essential attribute searched by the search mechanism 12 is registered in association with the generated scene.

For example, when the scene of intellectual activity is a “paper scene”, the domain knowledge database 1
The specified cup of the specific mechanism 13 managed by 0 is
When you have a "water scene" without having a "paper scene", generate a "paper scene" and
The attribute “weight becomes” that is searched for by the search mechanism 12 is registered in association with.

In this way, according to the present invention, in consideration of the fact that the attribute data of an object has different meanings in different scenes even in the same domain, the attribute data of the object is associated with the scene. By adopting the configuration of acquiring the object and registering it in the object knowledge database while corresponding to the scene, accurate object attribute data can be registered in the object knowledge database. It will be possible to build a knowledge database that can effectively support intellectual activities.

[0028]

EXAMPLES The present invention will be described in detail below with reference to examples. First, the object thinking data processing system proposed by the present inventors (daringly written as object thinking)
Will be described.

In the object thinking data processing system proposed by the present inventors, as shown in FIG. 2, the real world is interpreted by extension / comprehension and a name is attached to it. Then, the name and the extension are combined to define an extensional name that is a word, and the deep meaning of the word, the intensification, is the world of information hiding. This extensional name, or word, is made up of a name and extension, so it is necessary to easily understand the meaning of the word just by looking at it.

Therefore, considering an explanation sentence (extension + comprehension) representing this extensional name, the explanation sentence (metadata) is symbolically compressed, and the object i which is the actual data of this word.
Together with D, define it as an object command.

That is, an object is made up of real data and metadata as shown in FIG. 3, and this real data is directly represented by an object iD, and the metadata is a signature as a compressed body of attributes. It is represented by Ardata. This “object iD + signature” is an object command, and this object command means “word / tacit knowledge”. Here, as shown in FIG. 4, a word is a knowledge whose external iD is determined like a “dog”, and tacit knowledge is knowledge that is not a word, and an external iD is nothing. This is undecided knowledge. If you decipher this signature, you will understand the entire structure of the object.

The object thinking data processing system proposed by the present inventors adopts a configuration in which data is executed by skipping a message consisting of a list of the object commands, and an actor object (the present inventor In the case of an object that realizes the object thinking data processing system proposed by et al., And an object that realizes a virtual human world described later) does not have an entity, as shown in FIG. The configuration is such that the substance is learned by being transferred from the object. In other words, the idea of learning can be said to be a function that automatically copies the substance as if the cells were growing, by skipping the minimum necessary information called object commands. Here, what is subject to learning includes not only data but also programs.

According to the message communication method and the learning function, when the system is composed of a plurality of workstations, the program of the terminal A can handle the data of the terminal A and the data of the terminal B, for example. That is, since each workstation functions as both a client and a server, it becomes possible to integrate the client and the server. Since each workstation can regard all the workstations other than itself as a server, it is possible to realize ideal distributed processing.

A characteristic of the object thinking world considered by the present inventors is that the object command is interpreted as a word and is formed, and thus the word can be used to create a virtual human world.

That is, as shown in FIG. 6, in the object thinking world, the world of information hiding becomes knowledge, and the object world, that is, the virtual human world, is a model of a virtual society.
Looks like a virtual human model. The content of this knowledge of the information hiding world is the source (including the program, the screen, and the definition). The contents of the virtual society model are made up of the static world, the dynamic world, and the causal relationship world described above. The contents of the virtual human model are virtual humans living in the virtual social model.

This virtual human model is made up of programs and screens corresponding to the physical body and knowledge corresponding to the spirit. The virtual human having this physical body and spirit is self-recognition / self It exerts the agent function of evaluation / self-organization / distributed cooperation.

That is, a human being has a body made up of genes, but has completely different appearances, faces and shapes, and has a spiritual thing called knowledge in a completely different form from this physical body. According to such a human being, in the actor object, as shown in FIG. 7, the signature is associated with the gene,
The method (program) corresponds to the body, the view (screen as a representative) corresponds to the figure, face, and shape, and the definition of words corresponds to knowledge.

The structural element of the method, which is the physical aspect, is a sequence of the order in which the programs that are atomic are executed, and the structural element of the view, which is the physical aspect, is the atomic screen pattern. It is the coordinates of where to display the piece, and the structural element of the word, which is a mental aspect, is each element of the word expression. That is, the signatures corresponding to genes are roughly divided into a structural signature regarding methods and views, which are physical aspects, and a mental signature regarding words, which is a mental aspect.

Here, self-recognition means that an object has a self-awareness, and thus has knowledge of metadata. Self-evaluation means that an object has a goal, and instinctively continues to grow to a goal without a goal according to an independent function, or achieves a goal with a goal indicated by a user according to an autonomous function. Self-organization is self-reliance / autonomy in the light of instinct, and this instinct is acquisition of knowledge, learning, understanding and thinking. Distributed cooperation is cooperation with other independent objects and humans one by one.

On the other hand, the substance of knowledge is a compound object that is a combination of an atomic object that is the basis of all knowledge and an atomic object, and this atomic object includes an atomic program, an atomic screen, and atomic words. Yes, compound objects include compound programs,
There are compound screens and compound words.

The metadata of these objects include
There are dictionary data and directory data, and this dictionary data includes object structure, management data, semantic data, and the like, while directory data indicates the storage address of each data. A component attribute file is prepared to store all these data.

Next, the structural network of the object thinking world will be described. The information hiding world is where knowledge is stored and is made up of atomic objects and compound objects. This atomic object is created from a software IC group, is internally expanded into an EXE program (execution program), and is an actual method.
Therefore, it is also called atomic class. On the other hand, the compound object is represented by the object command attached to this atom class. Therefore, the compound object is made up of the definition program (signature) by the genetic algorithm, not the EXE program, and the information is hidden.

Since the atom object is reused as a component, it has complete independence of encapsulation, has orthogonality that allows it to be combined with other components, and has normality without redundancy. And has an abstraction that allows free exchange of data / procedures, and has a simple series structure.

As one means for imparting this independence, a configuration is adopted in which the causal relationship processing is delegated to another program. That is, even if a relationship that requires exclusive processing is assumed between the atomic objects, the primitive objects are obtained by disregarding the relationship, and the causal relationship processing is performed separately from the primitive objects. It adopts a configuration in which it is dealt with by the processing of the causality processing program.

As one means for providing this orthogonality, each primitive object is provided with a self-PR function, and the self-object is of any type,
It is configured to have a function of notifying the user what data is needed. Then, it adopts a configuration including a mechanism that executes compatibility of arguments so that specified combinations can be combined.

Further, as one means for providing this normality, a structure is provided in which primitive objects are put together by using is-a and part-of relationships established between objects. For example, while a primitive object is being generated, a primitive object P, a primitive object Q, and a primitive object R are present, and most of them are the same, and each P portion is a peculiar portion. , Q portion, and R portion exist, a primitive object A corresponding to the superclass is generated corresponding to the common portion, and a peculiar P portion is left for the primitive object P, and the superclass is the primitive object. A is specified, a peculiar Q part is left for the primitive object Q, its superclass is specified as a primitive object A, and a peculiar R part is left for the primitive object R. It has a configuration instructing that the class is the primitive object A.

Further, as one means for giving this abstraction, the primitive object instructs the buffer what (WHAT) is given as input, and what (HOW)
It adopts a configuration having a unified WHD structure in which the buffer is instructed to perform processing and the processing result (DO) is instructed to the buffer.

As one means for providing this simplicity, a control component such as if is prepared as a primitive object. For example, when there is an object having a branch, the part corresponding to this branch is cut out as a control part, and the other parts are cut out as non-control parts. If it is cut out in this way, each non-control component will have a series logic without branching. Then, it is configured to realize simplicity by deciding whether or not to leave those non-control parts as a primitive object.

In the space world of the virtual human world created by the definition body (signature) by the genetic algorithm, from the inclusion side to the extension side, the structure class, sequence class, and instance class are experienced as shown in FIG. Defined as a rule, session instances are created in parallel as a program that actually operates in the time world. As shown in FIG. 9, there is causal synchronization between those session instances.

This structure class is created by grouping atomic objects, the sequence class is created by defining connections in the structure class, and the instance class is created by fixing the parameters in the sequence class. The session instance will be generated by combining a plurality of instance classes.

Here, the compounding of the object is executed by a mechanism called hyper-language processing by checking the compatibility of the argument matching. That is, a new compound object is created by combining atomic objects, a new compound object is created by combining an atomic object and an existing compound object, and a new compound object is created by combining existing compound objects. However, as shown in FIG. 10, the hyperlanguage processing refers to the past combination result and interacts with the user as necessary, and as shown in FIG. Complex objects will be created by performing checks.

On the other hand, the composite object is also modified by a genetic operation called a reverse link technique to create a new object. That is, a gene that describes the structure of a complex object is visually displayed, and when the user performs a gene operation such as deletion or addition on the gene, the metadata and the actual data pointed to by the gene are changed. In the configuration, the new object is associated and created. For example, the association of objects is executed by replacing some of the genes in the structure of connection of multiple objects, and the creation of objects is executed by adding new genes to the genes.

Such a composite object has a composite class and a composite instance. As described above, the composite class includes an atomic class, a structure class,
There are sequence class and instance class, and composite instance has session instance.

When this composite object is activated, as shown in FIG. 11, the main program called the object management mechanism follows the definition of the composite object and synchronizes the causal relationship with the atoms of the EXE program. By calling the object, you will execute the composite object.

In the conventional object-oriented system, once instantiated, an object cannot be created below it, but in the object thinking proposed by the present inventors, the object is made persistent even if instantiated. And use that persistent instance as a scaffold,
There is a reverse superclass / reverse subclass mechanism that creates a class underneath it and creates a completely new instance of that class. For example, as shown in FIG.
The value of is nullified once to classify it, and a new instance can be created by applying another value to X again.

Further, in the conventional object orientation, the relation between the objects is extremely poor. Therefore, in the object thinking proposed by the present inventors, as shown in FIG. Is-a, part-of, a-kind-of, and has-a that specify relationships, and relational-of and networking-of that specify network-like relationships.
, 2-term-of are available. These relationships will be used primarily for object retrieval.

This is-a is, for example, "a dog is an animal".
In this case, the relative position of the dog is expressed by layering the relationship between the lower "dog" and the upper "animal".
The part-of expresses a structural relationship in which, for example, “Kyushu” constitutes a part of “Mainland Japan”. a-kind-of is almost the same as is-a, but is-a
Is used to group all the sub-items, while all are included. For example, "dog, cat, monkey" and "tiger, panther, lion" are considered equivalent in is-a and classified as mammals. At this time, by using a-kind-of,
"Pet" for "dog, cat, monkey" with the is-a relationship
In addition, "tiger, panther, lion" can be grouped into "predators".

The has-a is used when the part-of is satisfied only when all of the child classes attached to the parent class are complete, whereas when only part of the child class is satisfied. Relational-of is used when a class is dependent on a class. Use the networking-of when it is appropriate to associate parent / child links within a group. 2-term-of expresses a binary relation.

In this is-a relationship, as shown in FIG. 14, for example, the class X ₁ and the class X ₂ are in the is-a relationship, and the class X ₁ is the method attribute Pa ₁ / data attribute A. ₁ / behavior P ₁ and when class X ₂ is represented by method attribute Pa ₂ / data attribute A ₂ / behavior P ₂ , when class X ₂ is fired, a composite class Y ₁₂ An for attributes and data attributes
It will be processed in d, and the behavior will be processed in or. Further, also in relation to part-of, for example, when the class X ₃ and class X ₄ is in the part-of relationship class X _34, the composite class Y ₃₄ can when the class X ₃₄ is ignited , The method attribute and the data attribute are processed by and, and the behavior is processed by or.

FIG. 15 shows a frame representation of the actor object constructed in this way. That is, a conventional object is composed of a name, an attribute, and an operation, whereas an actor object is
It consists of knowledge, structure level, operations, and views. Here, this structure level represents the type of structure network such as a primitive object or a structure object, and the operation parameter represents the sequence of the object to be executed, and the view parameter. Represents the display coordinates on a screen.

Recently, there has been proposed a concept known as a language, which should be called "a fluid body of bubbling" meaning "energy that is verbalized in various degrees." In this language Araya, it is said that a portion of the "meaning" energy flowing deeply is activated individually or in combination in response to the demand of time and occasion, and in the light of daylight of surface consciousness. The knowledge is explained by appearing in. In other words, this linguistic knowledge explains that knowledge changes dramatically depending on the situation.

Knowledge is replaced by a word that can be understood, but, as is the case with the same word, the meaning varies with different domains, corresponding to the explanation in the concept of language araya. Therefore, knowledge is a word that can be understood in a certain domain, and the object command, which is a word, must also take the domain into consideration.

In this way, in the object thinking data processing system proposed by the present inventors, an object command which is an extensional name is defined as an internal iD, which is used as a word in a certain domain, and the word is used as a word. The meaning or definition of is expressed by atomic parts / composite parts in the world of information hiding. Then, using words with this meaning, we imitate a virtual human in the object world. The virtual human being expressed by this word lives in a static world that represents a spatial relationship, a dynamic world that represents a temporal relationship, and a causal relationship world that represents a synchronous relationship that exists between them.

According to this configuration, the object thinking data processing system proposed by the present inventors can easily and faithfully model the real world. Next, the present invention will be described in detail according to an example applied to a domain builder / system builder.

FIG. 16 shows the overall configuration of a domain builder / system builder constructed by the object thinking data processing system proposed by the present inventors. In the figure, 20 is a domain builder (sometimes referred to as domain / parts builder in the figure) that constitutes a knowledge acquisition area, 21 is a system builder that constitutes a knowledge use area, and 22 is a domain commonly prepared for the system. Common knowledge that does not depend on the reference numeral 23 is a knowledge storage area for storing knowledge for each domain constructed by the domain builder 20. Although not shown, the knowledge storage area 23 may store in advance information prepared by the system.

The domain builder 20 is a mechanism for setting the domain and using the common knowledge 22 to acquire the knowledge of the domain, that is, the empirical rule and store it in the knowledge storage area 23. Specifically, the domain builder 20 interacts with an expert who is familiar with the domain to be constructed and, in that domain, object parts (program parts, view parts, language parts) as knowledge.
There are different types). That is,
Learning is done. At this time, this newly acquired knowledge is reflected in the common knowledge 22, and the common knowledge 22 grows accordingly.

FIG. 17 shows the execution process of the domain builder 20. As shown in this figure, an expert who is a user of the domain builder 20 firstly
Recognize domains. If there is no domain, a new domain is created, and if the domain already exists, that domain is selected. Then, knowledge is generated in the domain. This knowledge is the program part,
It is a view part and is a word part that is valid in its domain and is used for explaining the structure and the like of these parts.

To specifically explain this knowledge generation method, the expert who is the user first selects and copies the model knowledge from the common knowledge 22. At this time, all the common knowledge 22 may be copied, or only the knowledge that the user likes may be copied. Then, using that knowledge as a starting point, transform that knowledge into knowledge of your domain,
Create new knowledge. The task is to change the definition of knowledge or create a new definition of knowledge. Therefore, it is necessary to match whether or not the definition is suitable for the knowledge. Explaining the compatibility of program parts, the process of matching the arguments of the primitive objects making up the program part is performed.

In this way, the domain builder 20
Builds the knowledge, that is, the words, that can be used by its friends in its own domain and stores it in the knowledge storage area 23.
The common knowledge 22 used by the domain builder 20 has a definition of knowledge that can be considered in any domain without weight, and can be said to correspond to the above-mentioned language araya knowledge. This common knowledge 22 should be prepared considering all domains, but since it is impossible to prepare from the beginning, as shown in FIG. 16, the newly acquired knowledge is common. It has a structure that grows as it is reflected in the knowledge 22.

On the other hand, the system builder 21 is a mechanism for constructing a system by using the knowledge created by the domain builder 20. Specifically, the system builder 21 builds the system desired by the user while providing the knowledge created by the domain builder 20 to the user who builds the system. That is, intellectual activities are performed. At this time, Domain Builder 2
When there is a lack of knowledge created by 0, a request to acquire the knowledge of the lack will be issued to the domain builder 20.

FIG. 18 shows an execution process of the system builder 21. As shown in this figure, the system designer who is the user of the system builder 21 first recognizes the domain and prepares a world that can be expressed in the words of his own world. Next, grasp the entire specification while having a meeting with the system user. This grasp is made by collecting and organizing the minutes of meetings.

Then, the vocabulary shown in the minutes is decomposed, the vocabulary is found from the meaning shown in the minutes, and the knowledge storage area 23 is searched by using the vocabulary. The required program parts and view parts are retrieved from the similarity search (for example, by evaluating the similarity of the signature). It is the structural definition (structural signature) of a part that is used to directly find a program part or a view part by performing lexical decomposition, while semantically, the program part or view part is semantically defined. It is the mental definition of the part (spiritual signature) that is used to find out.

Then, the vocabularies are put together into meaningful groups in the domain to group the parts, and the connection of the parts is defined to form a scenario. Then, by inserting concrete data into the parts that are made into scenarios, the program parts and view parts necessary for the system are identified, and the system is completed by automatically programming them by automatic compatibility and checking the operation. . That is, the system builder 21 completes the system according to the procedure as shown in FIG.

In this way, the system builder 21
First stepped through the specification knowledge, then
The system is constructed by searching for the parts by the specification keyword, then combining the parts by an empirical rule, and then creating the specifications, programs, and screens.

As shown in FIG. 20, the system to be constructed, which represents the real world, can be regarded as knowledge having a hierarchical structure such as a table of contents, chapters, and terms, as shown in FIG. In terms of the design procedure, the system construction by the system builder 21 firstly performs a conceptual design that determines a rough format, and then a detailed design such as instance design, synchronization design, link design, etc. Then, it means to take the procedure of concrete design of instantiation. Figure 21
20 shows a specific example of this FIG.

In this way, the domain builder 20 uses the common knowledge 22 while setting the domain so that the knowledge of the domain, that is, the rule of thumb is acquired and stored in the knowledge storage area 23. System Builder 21
The knowledge created by the domain builder 20 is used to perform intellectual activities and build a system. Figure 2
2 and FIG. 23 schematically show the processing of the domain builder 20 / system builder 21. The Q / A function 24 in the figure is a mechanism for issuing a knowledge acquisition request to the domain builder 20 when the knowledge required by the system builder 21 is not stored in the knowledge storage area 23. Is. Further, here, the knowledge storage area 23
The common knowledge 22 is developed in the above.

FIG. 24 shows the detailed structure of the knowledge storage area 23 for expanding the knowledge acquired by the domain builder 20. In the figure, 30 is a recognition module file, 31 is single recognition data expanded in the recognition module file 30,
Reference numeral 32 is semantic network recognition data expanded in the recognition module file 30, 33 is an actual data file, and 34 is a real data file.
Is a recognition index file, 35 is a single recognition index constituting the recognition index file 34, and 36 is a navigation recognition index constituting the recognition index file 34.

The recognition module file 30 develops the attributes of an object to be knowledge by an object command, and in addition to this, for example, "a cup is made of glass, made of glass, and the water is a fluid. A module that recognizes the semantic network of properties and structures that are established between the attributes of objects such as "is" or "apples and pears are fruits" (object attributes)
Expand every time. The single recognition data 31 corresponds to the attribute information of this object, and the semantic network recognition data 32
Corresponds to this semantic network. The semantic network recognition data 32 will be automatically updated according to the update of the simplex recognition data 31.

The actual data file 33 stores the substance of the object and the substance of the object attribute. For example, the substance of the first recognition module is stored in the file of the relational database structure, and the substance of the second recognition module is stored. The entity is stored in the file of the network database structure, the entity of the third recognition module is stored in the file of the time series database structure, and so on. Store.

The single recognition index 35 of the recognition index file 34 constitutes an index of the single recognition data 31 and is composed of indexes for each recognition module such as a name index, a domain index, a scene index and a semantic index.

On the other hand, the navigation recognition index 36 of the recognition index file 34 constitutes an index of the semantic network recognition data 32, and is recognized for each recognition module such as a name navigation index, a domain navigation index and a structure navigation index. It consists of an index. This navigation recognition index 36 is basically a super class,
It manages the relationship with the classes under it,
It also manages the relationship between the super class like and the classes under it, such as what happened on one day and what happened on that day. In addition, this single recognition index 35
And the navigation recognition index 36 will be automatically updated according to the update of the single recognition data 31.

FIG. 25 shows the data structure of the unit recognition data 31 expanded in the recognition module file 30. Here, the substance of the single recognition data 31 is stored in the actual data file 33.

In this embodiment, the unit recognition data 31 is
It is configured to have an entity reference recognition module, a structure recognition module, a name recognition module, an idea support recognition module, a maintenance / management recognition module, a domain recognition module, a meaning recognition module, and a growth recognition module. .

This entity reference recognition module is a recognition module prepared for recognition of the entity of the object, and manages pointer information to the actual data of the object. By referring to this entity reference recognition module, the entity of the object can be recognized. That is, the storage mechanism used in the conventional computer technology of storing the entity of the object as it is can be used as it is.

The structure recognition module is a recognition module prepared for recognition of the structure of an object,
A part type indicating whether the object is a program part / screen part / definition part, a structure level indicating whether the object is a primitive object / structure object / sequence object / instance object / session object, and an object It manages genetic information about the structure of. This structural gene information basically represents the types of objects that make up the object and the link data between those objects.

The structure of the object can be recognized by referring to this structure recognition module. That is, if it is a program part, it can recognize the type of the program on which it is based, its link data, its parameter type, and its parameter value, and if it is a view part, it is the basis. You can recognize the type of view and its link data, and if it is a word part,
It is possible to recognize the type of words on which it is based and the link data.

Is-a managed by this structure recognition module
The structural network of the object is extracted from the structural data of the object such as a part-of or the like, and is expanded as the semantic network in the semantic network recognition data 32 provided for the structure recognition module. It is mainly used in the reuse mechanism of parts such as creating another object by making a part of it different, or changing it by changing the role name without changing the structure itself. There is.

Here, the gene information of the object structure managed by this structure recognition module also represents the context of the object. That is, FIG.
As shown in 6, "inu" and independent "i" and "nu"
Is an entirely different object, and "inu ga ga" is completely different from "inu" and "walking", so an object that includes context is completely different from the object part that constitutes it. Be different. Regarding the context of the object, which is such extremely important structural information, the gene information of the object structure represented by the signature is structurally expressed.

Incidentally, when the object is a program part, as shown in FIG. 27, the object command changes as the program progresses. This is also one of the contexts of the object. Yes, it indicates that the context categories change.

As described above, when the object command is skipped, the side receiving the object command learns it if it has no substance.
Therefore, using the property of this object command, as shown in FIG. 28, the workstation of the data processing execution source skips the object command “100-08”, so that the object command “100-08” Program processing can be delegated to another workstation. In other words, by using the property of this object command, you can control the load distribution freely and
Functions can be easily distributed by changing the specifications in one place.

On the other hand, the name recognition module is a recognition module prepared for recognizing the name of an object, etc., and the name is the external iD of the object, the object command is the internal iD of the object, and the table keyword of the object. And manage. Here, the table keyword is a word that characterizes an object similar to a name. Further, this object command is used as an internal iD and is also used when evaluating the degree of similarity between two objects because it is symbolized data of object attributes. By referring to this name recognition module, the name of the object, the object command, and the keyword of the object can be recognized.

The idea generation support recognition module is a recognition module prepared for supporting the idea generation of an object, and manages the back keyword. This back keyword is defined by the paired data of the name of the object that is the source of modification and the domain name to which it belongs, and the user who uses the object registered in the recognition module file 30 has the same When a new object is created by changing the name, table keyword, meaning data of the meaning recognition module, etc. (when the modification source object is left as it is), the user who registered the modification source object , Is registered so that the contents of the modification can be known.

In other words, according to this back keyword, the object of another domain created from the own object can be searched, so by recognizing the difference between this searched object and the own object, the name is changed. It is possible to know what kind of table keyword has been changed, what kind of meaning data has been changed, and what kind of meaning data has been changed. Therefore, by using this back keyword, the user who has registered the object can know another way to look at another domain, and self-develop. For example, if the user who registered the table keyword thought that he would only eat apples,
If another user modifies it as a paperweight in another domain, it can be known that it can also be used as a paperweight by this back keyword.

The maintenance / management recognition module is a recognition module prepared for maintenance / management of an object, and management data such as an administrator and management date / time, management data of metabolism of an object, time management and process management. Manage the management data such as. That is, it manages when, when, who, for what purpose, created, changed, added, and deleted the object. Also, the metabolism management data of this object manages the difference information of the updated object, and the object can be restored using this.

The domain recognition module is a recognition module prepared for harmony with different domains, and manages the domain name to which the own object belongs and the difference information between the different domains. When the domain to which your object belongs is different from the domain of the other party, you can use the management data of this domain recognition module to dynamically adjust your knowledge to the knowledge of the other domain. Will be realized. In addition, the domain name is hierarchized such as which device of which industry.

The meaning recognition module is a recognition module prepared for recognition of the meaning of an object.
It manages semantic data such as properties, behaviors and meanings of objects, but even within the same domain (the same device in the same industry), apples become food for humans, apples for paper, and papers for paper. For a painting, since the meaning changes when the partner, that is, the scene (field), such as the apple becomes the shape and color, the meaning data is managed for each scene. This scene is the scene that the user is about to use.

As the semantic data managed by the meaning recognition module, for example, as shown in FIG.
There are 11 types of meaning level / purpose / work / property / role / entity / similarity / antagonism / prefix / suffix / explanation. Here, the meaning level indicates which of the three modes the object is in, the purpose (What) represents what the object is used in the scene, and the function (Ho).
w) represents how the object is used in the scene, and the property (State) represents the state of the object in the scene, and the role (Role).
Is the object's original role in the scene, entity is what the object's actual machine is in the scene, and similarity is a similar term for the object in the scene. Means what the object is, the antonym means what the object's antonym is in the scene, the prefix is what the object's prefix is in the scene, and the suffix is Represents what the suffix of the object is in the scene,
The description represents a comment on the object in the scene. An example of this semantic data is illustrated in FIG.

Of this meaning data, meaning level / purpose /
The work / property / role can be said to be hierarchical meaning data based on the idea of horizontal thinking, and from that point, it constitutes a hierarchical meaning recognition module. Further, among the meaning data, since the entity is the actual machine name, it constitutes the entity name recognition module. Further, among the meaning data, the similarity / antagonism / prefix / suffix can also be said to be associative meaning data, and the associative memory module is configured from this point. The meaning recognition module and the domain recognition module together form a situation judgment recognition module.

Relationships such as is-a and part-of that are established between the semantic data managed by the meaning recognition module and is-a and part-established between the management data managed by the maintenance / management recognition module. is-a or part-of that is established between the relationship such as of and the management data managed by the name recognition module
The property network of the object is extracted from the relationship such as, and is expanded as a semantic network in the semantic network recognition data 32 provided for each recognition module. Since it is used for etc., it is mainly used for the thinking mechanism.

The growth recognition module is a module prepared for automatically adding a schema when new recognition occurs. That is, this growth recognition module is prepared as a mechanism for the individual recognition data 31 expanded in the recognition module file 30 to grow.

In this way, the knowledge storage area 23 stores the single recognition data 31 having such a recognition module structure.
Is expanded into the recognition module file 30,
The objects learned by the domain builder 20 will be memorized.

FIG. 30 shows a processing flow of object storage processing in the knowledge storage area 23 executed by the domain builder 20. That is, the domain builder 20
As shown in this processing flow, when a new object part is created by combining the arguments of objects according to the above-mentioned hyper language processing, the actual data of the object is stored in the actual data file 33, and then the object is stored. Object command is created, and then the single recognition data 31 of the object is created.

Then, the single recognition index 35 is created by creating an index of the single recognition data 31 for each recognition module. And for each recognition module,
It searches whether or not the relation such as is-a is included, and if it is included, the superclass and the subordinate classes are extracted, and the semantic network recognition data 32 is created from them. In addition, the navigation recognition index 36 is associated with the meaning network recognition data 32.
Is created.

In registering an object in the knowledge storage area 23, the domain builder 20 creates a recognition module category suitable for the object to be registered, each time according to the human storage mechanism. The configuration is such that the registration process is executed according to the category of the created recognition module. Then, when the object to be registered includes a context, the category of the context is created each time, and the registration process is executed in a form including the created context. There is.

That is, for a person, for example, in the case of a portrait picture for certification, only the person's name is the object of recognition, whereas in the case of a portrait picture of a travel, where, when, and with whom It is not a constant recognition module that is always recognized, but instead of always using a fixed recognition module, you can freely select a recognition module according to the recognition purpose and create a new one if necessary and memorize the real world. It has a memory mechanism to do. In accordance with this human memory mechanism, the domain builder 20 creates a category of recognition module suitable for the object to be registered each time and registers the object in the knowledge storage area 23. It should be noted that the recognition modules that are the target of the category also include a hierarchical recognition module group that is developed inside the above-described recognition module group.

When the object to be registered includes a context, the object is registered in the knowledge storage area 23 in a form including the context in accordance with the rule of the context.

Next, the reference processing of the object stored in the knowledge storage area 23 executed by the system builder 21 will be described. When the system builder 21 uses the objects stored in the knowledge storage area 23 to perform an intellectual activity for system construction, the system builder 21 uses a recognition module that characterizes objects necessary for the intellectual activity in order to realize the purpose recognition. Object recognition is realized by searching for objects stored in the knowledge storage area 23. For example, when acquiring what happened at a specific date and time as knowledge, the maintenance / management recognition module is used to search for objects.

This purpose recognition usually starts by recognizing the domain while being aware of the name, and therefore it is started using the domain recognition module and the name recognition module, and thereafter, for example, , The structure recognition module is used to recognize the structure, and then the meaning recognition module is used to recognize the meaning.

As mentioned above, the domain builder 20
Has a structure in which objects are stored in the knowledge storage area 23 in accordance with the human storage mechanism. On the other hand, the system builder 21 uses the knowledge storage area 2
The object stored in 3 is used to realize the re-recognition and recall mechanism of the human memory mechanism.

FIG. 31 shows a system builder 2 for realizing the recall mechanism of the human memory mechanism.
1 illustrates a processing flow executed by 1. That is, as shown in this processing flow, the system builder 21 detects in step 1 that one or more names, table keywords, etc. are designated by the user to issue a target recognition instruction, and the subsequent step is executed. In step 2, the single recognition index 35 provided in association with the specified recognition module is searched, and in the subsequent step 3, whether the result of the search, the name of the search instruction, or the like is registered in the recognition module file 30. To judge.

When it is determined by this determination processing that the search method is registered in the recognition module file 30, the process proceeds to step 4, and it is determined whether the search method designated by the user is the navigation search or the single search. do it,
When it is determined that the simple substance search is designated, the process proceeds to step 5, the object command included in the simple substance recognition data 31 searched by the simple substance recognition index 35 is obtained, and in the following step 6, from the actual data file 33, The actual data of the object pointed to by the object command is retrieved, and the process ends. When the actual data file 33 has a structure in which the actual data is stored according to the multiple file structure as shown in FIG. 24, the search process of the actual data file 33 is executed in parallel.

On the other hand, when it is determined in step 4 that the navigation search is designated, the process proceeds to step 7 to search the navigation recognition index 36 provided in association with the designated recognition module, and continues. In step 8, as a result of the search, the name of the search instruction is the meaning of the recognition module file 30. Network recognition data 3
It is judged whether or not it is registered in 2.

When it is determined in step 8 that the name of the search instruction is registered in the meaning network recognition data 32, the process proceeds to step 9, and the name of the search instruction is used as a starting point to recognize the meaning network. Data 3
By performing a navigation process such as navigating 2 from the super class to a subordinate class, a series of simple substance recognition data 31 related to the name of the search instruction is searched, and in step 10 that follows, The object command included in the retrieved single recognition data 31 is obtained.

Subsequently, in step 11, the actual data of the object pointed to by the object command is retrieved from the actual data file 33, and in step 12, the actual data and the attribute information of the object pointed to by the object command are displayed. Then, in the following step 13, by specializing a part of the retrieved single recognition data 31 in accordance with an instruction from the user for the display information, an object having a retrieval request is created and the process is terminated.

On the other hand, when it is determined in step 8 that the name of the search instruction is not registered in the meaning network recognition data 32, the process proceeds to step 14 to notify the user to that effect, and in the following step 15, The process according to the instruction from the user is executed and the process ends.

In this way, the system builder 21
Is equipped with a navigation search function, and in accordance with this navigation search function, it is possible to search for objects stored in the recognition module file 30 while navigating from general categories to individual categories. Become. This makes it possible to retrieve a set of objects related to an object, such as an event that occurred on a certain day.

This means that the recall mechanism possessed by the human memory mechanism is realized without any correction. In the processing flow of FIG. 31, it is disclosed that the system builder 21 realizes the recall mechanism of the human memory mechanism.
Can also realize the re-recognition mechanism of the human memory mechanism.

In order to realize this re-recognition mechanism, the system builder 21 issues a re-recognition mechanism search request when a user designates one or more names, table keywords, etc. A signature is created from the single recognition data 31 of the recognition module file 30, and one or more single recognition data 31 having a signature similar to the signature.
To search.

Then, according to an instruction from the user, a part of the retrieved single-unit recognition data 31 is changed, or the meaning of the retrieved single-unit recognition data 31 is changed, whereby a search request is made. Create the object and finish the process. Here, when searching for similar items, it is common to search for items that have a high degree of similarity, but in order to achieve human flutter, etc., items that are very similar are searched for. It is also necessary.

In this way, the system builder 21
Is equipped with a similarity search function, and according to this similarity search function, the object stored in the recognition module file 30 is searched for an object similar to the object to be recognized, and the object to be recognized is It becomes possible to recognize the object while comparing it with the stored object.

With this, when the user requests such a program, a registered program close to it is presented so that the user can re-recognize what the user's requested program is. Meanwhile, the requested program can be completed by changing some of the attributes and the structure of the presented program according to the user's instruction. For example, when the program requested by the user is an equality comparison program that determines whether or not the input data is equal to a specified value, a magnitude comparison program that determines the magnitude comparison of two input data is presented, By changing the role of one of the input data as a specified value, the program requested by the user is completed.

This means that the re-recognition mechanism possessed by the human memory mechanism is realized without any correction. In this way, the system builder 21
As shown in FIG. 32, the object stored in the knowledge storage area 23 is used to realize the re-recognition / recall mechanism of the human memory mechanism.

Here, as also shown in the figure, the re-recognition mechanism and the recall mechanism realized by the system builder 21 are intricately entangled with each other. That is, a superclass is created from the single recognition data 31 obtained by the re-recognition mechanism, and the super class is used as a starting point to move to the recall mechanism, or the single recognition data 3 obtained by the re-recognition mechanism is used.
1 is used to search for the semantic network recognition data 32, or the similar recognition data 31 obtained by the recall mechanism is searched to move to the re-recognition mechanism, or the simplex obtained by the recall mechanism is used. It is possible to perform complicated processing such as re-recognition using the recognition data 31, but from this point as well, it can be said that the system builder 21 accurately realizes a human memory mechanism. .

Next, the knowledge acquisition process executed during the intellectual activity of the system builder 21 will be described. Figure 25
As explained in, apples become food for humans, apples for paper, paper for apples, apples for paintings become shapes and colors, even within the same domain. For example, the meaning of an object also changes. From now on, in the knowledge storage area 23, the meaning recognition module that manages the meaning of the object has a structure that manages the meaning data for each scene. To be more precise, the management data of other recognition modules also have the property of being different if the scene changes.

That is, as shown in FIG. 33, the attribute data of an object managed in the knowledge storage area 23 indicates what role (Ro) the object has in the scene.
le) should be understood by the scene / roll data structure. Here, the scene is identified by the name of the opponent object, and for example, when trying to hold down a paper that is about to fly with a paperweight, it is identified as a “paper scene”. The name of the object may be used as the first scene name.

By registering the attribute data of the object grasped by this data structure in the knowledge storage area 23, the system builder 21 executes the knowledge acquisition processing for making the attribute data of the object sophisticated. Become.

In order to realize this knowledge acquisition processing, the system builder 21 uses the knowledge stored in the common knowledge 22. As described above, the common knowledge 22 manages domain-independent knowledge commonly prepared in the system, and the attributes of objects are shown in FIG.
As shown in (1), it manages an essential attribute which is a common attribute and an auxiliary attribute which is a horizontal thinking attribute. As shown in FIG. 35, the common knowledge 22 is configured to grow by fusing the attribute data of the objects for each domain registered in the knowledge storage area 23. Here, the domain knowledge for each user held in the knowledge storage area 23 shown in the figure is the knowledge created by the domain builder 20 and the system builder 21, and the system common domain knowledge is the knowledge prepared by the system.

Next, according to the processing flow shown in FIG.
The knowledge acquisition process executed by the system builder 21 will be specifically described. When the system builder 21 enters the intellectual activity of the system construction processing, as shown in the processing flow of FIG. 36, first, in step 1, the knowledge storage area 2
The object stored in 3 is used to display a screen on the workstation, and the user is prompted to enter the name of the object in the screen required for the intellectual activity.
For example, when the paper is about to fly and the user tries to hold the paper with the paperweight, the name of the paperweight is input.

Subsequently, in step 2, it is determined whether or not the object having the inputted name exists in the screen. When it is determined that the object exists, the process proceeds to step 3 and is designated by the user. Make it work. For example,
The paperweight on the screen holds down the paper that is about to fly.

On the other hand, when it is determined in step 2 that the object having the input name does not exist in the screen, the process proceeds to step 4 and the common knowledge 22 is searched using the input name. , Get the essential attributes of the object with that name. For example, when a paperweight is input, the common attribute 22 having the management data shown in FIG. 37 (a) is searched to acquire the essential attribute of the paperweight, which is "weight." .

Then, in step 5, the auxiliary attributes of all the objects in the screen are extracted from the common knowledge 22, and the ones matching the essential attributes acquired in step 4 are searched from among them. As a result, the object in the screen having the essential attribute acquired in step 4 as an additional attribute is specified. For example, when there are a cup, a desk, and a desk lamp in the screen, an object having an attribute of "a thing that becomes a weight" as an ancillary attribute is specified. For example, as shown in FIG. 37 (b), if the cup managed by the common knowledge 22 has the additional attribute of "a thing that becomes a weight", the cup will be specified.

Subsequently, in step 6, it is determined whether or not the current scene of the intellectual activity and the scene of the object specified in step 5 (registered in the knowledge storage area 23) match. For example, the scene in which the paperweight that is about to fly is pressed by the paperweight is the "paper scene", so does the scene of the object specified in step 5 include the "paper scene"? It is decided whether or not.

If it is determined in step 6 that the current scene of the intellectual activity matches the scene of the object specified in step 5, the process proceeds to step 7 and the step of storing in the knowledge storage area 23. The essential attribute acquired in step 4 is added to the object identified in step 5 in association with the current scene of the intellectual activity possessed by the object. Then, the process proceeds to step 8 to perform the operation designated by the user in accordance with the newly registered scene / roll data.

For example, in the case where the current scene of intellectual activity is a "paper scene", and the cup identified in step 5 has a "paper scene" in which "memo writing" is a roll. As shown in FIG. 37 (c), the attribute “becomes a weight” acquired in step 4 is added in association with the “paper scene”. Then, use a cup instead of a paperweight to hold down the paper that is about to fly.

On the other hand, when it is determined in step 6 that the current scene of the intellectual activity does not match the scene of the object specified in step 5, the process proceeds to step 9 and is stored in the knowledge storage area 23. For the object identified in step 5, a new scene (that is, the current scene of the intellectual activity) having a scene name of the partner object of the intellectual activity object is generated, and the new scene is associated with the scene. Add the essential attributes acquired in 4. Then, the process proceeds to step 8 to perform the operation designated by the user in accordance with the newly registered scene / roll data.

For example, if the current scene of intellectual activity is a "paper scene", and the cup specified in step 5 has a "water scene" with the "drinking tool" as a roll, the "paper scene" 37 in the case of not having a “scene of
As shown in (d), a "paper scene" is generated in the identified cup, and the essential attribute acquired in step 4 is added in association with the scene. Then, use a cup instead of a paperweight to hold down the paper that is about to fly.

In this way, the system builder 21
36 executes the processing flow of FIG. 36 to make the attribute data of the object managed in the knowledge storage area 23 accurate according to the scene / roll data structure.

[0138]

As described above, in the present invention, considering that attribute data of an object has different meanings in different scenes even in the same domain,
Since the attribute data of the object is acquired in association with the scene and is registered in the knowledge database of the object in correspondence with the scene, the accurate attribute of the object in the knowledge database of the object is adopted. By being able to register data, it will be possible to build a knowledge database that can effectively support intellectual activities.

[Brief description of drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is an explanatory diagram of an object thinking model.

FIG. 3 is an explanatory diagram of an object command.

FIG. 4 is an explanatory diagram of words and tacit knowledge.

FIG. 5 is an explanatory diagram of learning.

FIG. 6 is an explanatory diagram of an object thinking world.

FIG. 7 is an explanatory diagram of a state of an actor object.

FIG. 8 is an explanatory diagram of a spatial world of a virtual human world.

FIG. 9 is an explanatory diagram of synchronization of causal relationships.

FIG. 10 is an explanatory diagram of compatibility matching processing.

FIG. 11 is an explanatory diagram of an object management mechanism.

FIG. 12 is an explanatory diagram of reverse superclass / reverse subclass.

FIG. 13 is an explanatory diagram of a relationship between classes.

FIG. 14 is an explanatory diagram of is-a, part-of.

FIG. 15 is an explanatory diagram of a frame representation of an actor object.

FIG. 16 is an overall configuration diagram of a domain builder / system builder.

FIG. 17 is an explanatory diagram of a domain builder execution process.

FIG. 18 is an explanatory diagram of execution processing of the system builder.

FIG. 19 is an explanatory diagram of execution processing of the system builder.

FIG. 20 is an explanatory diagram of a design procedure for system construction.

FIG. 21 is an explanatory diagram of a specific example of a design procedure.

FIG. 22 is an explanatory diagram of a domain builder.

FIG. 23 is an explanatory diagram of a system builder.

FIG. 24 is an explanatory diagram of a detailed configuration of a knowledge storage area.

FIG. 25 is an explanatory diagram of a data structure of unit recognition data.

FIG. 26 is an explanatory diagram of an object including context.

FIG. 27 is an explanatory diagram of a context in a program.

FIG. 28 is an explanatory diagram of object transfer.

FIG. 29 is an explanatory diagram of semantic data.

FIG. 30 is a processing flow executed by the domain builder.

FIG. 31 is a processing flow executed by the system builder.

FIG. 32 is an explanatory diagram of execution processing of the system builder.

FIG. 33 is an explanatory diagram of a scene / roll data structure.

FIG. 34 is an explanatory diagram of a data structure of knowledge managed by common knowledge.

FIG. 35 is an explanatory diagram of growth of common knowledge.

FIG. 36 is a processing flow executed by the system builder.

FIG. 37 is an explanatory diagram of knowledge acquisition processing.

FIG. 38 is an explanatory diagram of a conventional technique.

[Explanation of symbols]

 1 Object Knowledge Database 2 Terminal 10 Domain Knowledge Database 11 Common Knowledge 12 Search Mechanism 13 Identification Mechanism 14 First Registration Mechanism 15 Second Registration Mechanism

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Masahiko Murakawa 1015 Kamiodanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture, Fujitsu Limited (72) Inventor Masanobu 1015, Kamedotachu, Nakahara-ku, Kawasaki, Kanagawa Prefecture, Fujitsu Limited ( 72) Inventor Hiroyuki Tominaga 1-4-4 Hakataekimae, Hakata-ku, Fukuoka City, Fukuoka Prefecture Fujitsu Kyushu Communication Systems Limited (72) Inventor Fumio Horie 3-2-1 Sakado, Takatsu-ku, Kawasaki-shi, Kanagawa Fujitsu Network Engineering Stock company

Claims (5)

[Claims] 1. A domain knowledge database that is used for intellectual activities defined by a domain and manages attribute data of objects for each domain in association with scenes in the domain, and is registered in the domain knowledge database. Common knowledge that is prepared as a knowledge source of objects and manages the attributes of knowledge objects while distinguishing between essential attributes and ancillary attributes without depending on the domain. When the specified object does not exist in the scene of the intellectual activity, a search mechanism for searching the essential attribute of the object by referring to the common knowledge, and an object existing in the intellectual activity Among the above, a specific mechanism that identifies those that have the essential attributes searched by the above search mechanism as ancillary attributes, and the And when the scene of the object specified by the specifying mechanism matches, the object managed by the domain knowledge database is associated with the scene and the essential attribute searched by the search mechanism is set. When the first registration mechanism to register, the scene of the intellectual activity, and the scene of the object specified by the specific mechanism do not match, the knowledge of the object managed by the domain knowledge database is changed. And a second registration mechanism for registering the essential attribute searched by the search mechanism in association with the generated scene by generating a scene of the physical activity. 2. The knowledge acquisition processing method according to claim 1, wherein the domain knowledge database is configured to manage not only objects created by a user but also objects prepared from the beginning by the system. Knowledge acquisition processing method characterized by being performed. 3. The knowledge acquisition processing method according to claim 1, wherein the common knowledge is configured to grow by merging the domain knowledge databases that are updated. Knowledge acquisition processing method. 4. The knowledge acquisition processing method according to claim 1, 2 or 3, wherein the scene is configured to be identified by a name of an object that is a partner of the intellectual activity object. Acquisition processing method. 5. The knowledge acquisition processing method according to claim 4, wherein an object name is used as a first scene name.