JPH0582615B2 - - Google Patents

Description

[Detailed Description of the Invention] [Summary] In order to streamline processes such as searching, deleting, and updating a knowledge base that targets a large amount of knowledge, and to facilitate maintaining the consistency of knowledge, the present invention includes real-world meaning. Knowledge (data) such as information and intellectual data is transferred to the application layer and object layer.
It is managed using a three-layered knowledge base consisting of a layer and an extended RDB (relational database) layer, and a large amount of knowledge is placed in the extended RDB layer, accessed via the object layer, and cached in the object layer. Here, individual knowledge is represented by an object structure, and the object layer and extended RDB
In the layer, objects are managed in frame format, and in particular, the extended RDB layer has a structure that allows object frames to be held in tuples of non-normal relations.

[Industrial Application Field] The present invention relates to a knowledge processing system for managing and processing a large amount of knowledge using a data processing device.

[Conventional technology]

One of the trends in recent information processing systems is to enable them to directly process a variety of information (knowledge) centered on semantic relationships that humans use in the real world.

Since knowledge processing systems are always required to be more practical due to their purpose, the amount of knowledge they handle tends to become enormous. Therefore, a method for storing and managing large amounts of knowledge is required. Such processing systems are especially called knowledge processing systems or knowledge-based systems.

Furthermore, databases are required not only for routine work but also for non-routine work, and accordingly, intelligent access to data is required.

A knowledge base is a database that stores such real-world knowledge, and enables various knowledge search requests and file processing such as insertion, deletion, and updating.

In one conventional knowledge processing system, knowledge is managed in a knowledge base called a frame and expressed in the following format.

Frame slot 1 〓 〓 Slot i Facet i1 Value i1,... 〓 〓 Facet i2 Value i2,... 〓 : 〓 : 〓 Facet im Value im, ... 〓 Slot n That is, let n and m be arbitrary numbers of 1 or more, Also, when i is a value of 1 < i ≤ n, each frame is composed of n slots 1 to n, and each slot i is composed of m facets i1 to im and their values (multiple values are possible). ).

Next, FIG. 11 shows the configuration of a conventional knowledge processing system using frames. In the figure, 110 is a knowledge processing device, 111 is a main storage device, 112 is a frame processing program, 113 is a processor, 11
4 is a secondary storage device, 115 is a collection of frames, 1
16 is a procedure written in the frame.

In this way, in the conventional knowledge processing system, all frames are transferred from the secondary storage device 114 to the main storage device 111 and processed.

Furthermore, conventional databases are suitable for routine work, but not for non-routine work. This was due to poor ability to express and manipulate knowledge of the target world. Conventionally, in order to solve this problem, data in the target world was expressed using a data model called a semantic data model, and it was stored in a relational database (RDB), thereby making it possible to handle large amounts of data. , and is taking measures such as improving the interface with users.

FIG. 12 shows the configuration of a database using a conventional semantic data model. 121 in the figure is a semantic data model, 122
is a relational database. That is, traditional relational database 122
It provides an interface using a semantic data model.

[Problem that the invention seeks to solve]

In a conventional knowledge processing system using frames, a large amount of knowledge (frames) that cannot be handled by the capacities of main memory and virtual memory is stored in a secondary storage device. When the knowledge stored in the secondary storage device is needed, the entire frame must be transferred from the secondary storage device to the main storage for access processing.

For this reason, there is a problem that the number of I/Os during searches and other information operations increases, and the processing time becomes longer.

Furthermore, conventional database systems do not allow users to freely change data structure definitions (schemas), making them suitable for routine work but not suitable for non-routine work. Even database systems that use a semantic data model have various usage constraints and do not have object-oriented characteristics that integrate data and procedures, so it is difficult to achieve a practical level of performance. It was difficult.

[Means for solving problems]

In order to solve the above-mentioned problems of conventional knowledge processing systems, the present invention provides a knowledge base with an application layer that collectively manages knowledge in a specific field, an object-oriented model that represents the knowledge, and a frame that stores a large amount of objects. Extended relational database layer (hereafter extended relational database layer) stored in
RDB layer), the application layer and extensions
It consists of three layers: an object layer, which is provided between the RDB layer and caches the object frames of the extended RDB layer. Here, the extended RDB means an RDB that allows non-normal forms in which tables are set for field values.

The application layer and object layer of such a knowledge base are placed on the main memory, the extended RDB layer is placed on the secondary storage, and the extended RDB layer, which has a large amount of knowledge that cannot be managed on the main memory, is placed on the secondary storage. on the device and access the extended RDB layer knowledge frame through the object layer,
By caching at the object layer, it is possible to reduce the number of I/Os and improve the efficiency of processing such as searches. Furthermore, complex operations on knowledge frames in the secondary storage device can be easily realized by using extended relational database commands.

Furthermore, since objects have both static data and procedures, they can effectively express real-world knowledge structures and simplify various data operations.

Furthermore, the application layer collectively manages knowledge of specific application fields, allowing users to freely define data structures and
By making it possible to create and change it, it is easy to adapt it to non-routine business processing.

FIG. 1 shows a simplified basic configuration of the knowledge processing system of the present invention.

In the figure, 1 is a user, 2 is a terminal, 3 is a knowledge processing device, 4 is a main storage device, 5 is a processor, 6
is a secondary storage device, 10 is a knowledge base, 11 is an application layer, 12 is an object layer, 13 is an extended RDB layer, 14 is an individual application interface,
15 is an object interface, and 16 is an extended relational algebra interface.

As shown in the figure, the knowledge base 10 of the present invention
is application layer 11, object layer 1
2. It has a three-layer configuration consisting of an extended RDB layer 13. The application layer 11 and object layer 12 are mainly placed in the main storage device 4 (or virtual storage device), and the extended RDB layer 13 is placed in the secondary storage device 6.

In the application layer 11, knowledge in a specific field is accumulated and managed. For example, it includes a group of application programs, a graphics interface, an object editor, etc., and is activated by a user command.

In the object layer 12 of the present invention, real world knowledge (data) is expressed in an object structure (described later) using frames. Also, extended
In the RDB layer 13, frames are expressed and managed as tuples (rows of a table) (in the extended RDB, it is allowed to have tuples of non-normal relations, that is, to have a table for field values). In this way, real-world knowledge (data) can be expanded.
It is maintained and managed by the RDB layer 13.

The user 1 calls the knowledge processing device 3 through the terminal 2 and can access each layer through the individual application interface 14, object interface 15, and extended relational algebra interface 16.

The basic knowledge representation in object layer 12 is a frame. A frame takes the form described above and consists of one or more "slots." Slots can also have multiple "facets."
The facet value may also be multi-valued as well as single-valued.

The objects to be managed and operated in the object layer 12 are "objects".

Objects represent real, abstracted entities such as things, things, actions, and events in the real world. The object must represent three aspects of the entity's attributes:

(1) Properties of the entity (2) Relationship between the entity and other entities (3) Behavior of the entity These objects are expressed using frames. Make one object correspond to one frame.

The slots within the frame represent the properties, relationships, or behaviors of the entities described above. A slot representing a property (for example, a person's name) is given an actual value representing that property (for example, Hanako). In the case of a relationship, it has the name of another entity (ie, frame) as the value. And in the case of behavior,
A program (or function) is stored in the value.
(Specifically, program names and function names are stored as pointers).

[Effect]

In the configuration of the present invention shown in FIG. 1, the user 1 uses the knowledge of the application layer 11 via the individual application interface 14 to search for, insert, delete, update, etc. knowledge (data).
The format of the knowledge output at that time is the application layer 1
Based on the knowledge of 1, it is made suitable for each application field.

An instruction (or command) of the user 1 is converted into an instruction of the object layer 12, and the object layer 12 is accessed via the object interface 15.

Also, if necessary, the translated instructions can be extended
The commands are converted into commands of the RDB layer 13, and the extended RDB layer 13 is accessed via the extended relational algebra interface 16.

The result of the command for extended RDB layer 13 is
It is passed to the object layer 12 via the extended relational algebra interface 16. The results of instructions to object layer 12 are passed to application layer 11 via object interface 15 .

Instructions for the application layer 11 are passed to the user 1 via the individual application interface 14.

The knowledge base of the knowledge processing system according to the present invention also expresses knowledge of the target world using a data model called an object-oriented model, and stores it in a relational database RDB.
The system of the present invention can also be said to have a common feature with conventional systems in that knowledge of the target world is stored in a database using a secondary storage device. but,
The system of the present invention differs from the conventional system in the following three points.

First of all, in the knowledge processing system according to the present invention, an object-oriented model can be processed at high speed in main memory using frames, whereas in the above-mentioned semantic database, the data unit of the semantic data model is The problem is that there is no means to directly process entities.

Second, the three extensions described below with examples
By using RDB, version management and historical data management can be performed at high speed.

Third, the object-oriented model of the present invention:
Compared to semantic data models, it has higher knowledge expression ability and operational ability. it is,
One reason is that the object-oriented model uses objects, which are a combination of data and procedures, as the unit of knowledge.

〔Example〕

Next, details of the present invention will be explained according to an embodiment of the knowledge processing system shown in FIG.

In the figure, 1 is a user, 2 is a terminal, 3 is a knowledge processing device, 4 is a main storage device, 6 is a secondary storage device,
11 is an application layer, 12 is an object layer, 13 is an extended RDB (or XRDB) layer, 1
31 is a log XRDB that temporarily stores relatively recent access information, 132 is an individual XRDB that stores individual information, 133 is a shared XRDB that stores shared information, 31 is a UQL management section, and 32 is a
The OQL management section 33 is an XQL management section.

Note that UQL stands for user query language, OQL stands for object query language, and XQL stands for extended RDB query language.

Also, in Figures 3 and 4, each OQL management department 3
2 shows the flow of layer processing in XQL management section 2 and XQL management section 33. This will be explained below using FIGS. 2 to 4.

The user 1 inputs a command from the terminal 2 using a menu (or an icon, etc.) or directly inputs a command. Such user commands are expressed using UQL (User Query Language).

User 1 inputs a user command (UQL) from terminal 2. The input user command (UQL) is sent to the UQL management unit 3 of the knowledge processing device 3.
1 is input. The UQL management unit 31 analyzes the user command and starts individual application programs, object editors, etc. in the application layer 11. In the individual application program, data (commands) are input in the form of input/output, and desired processing is performed. At that time, the application program uses the knowledge of the individual application in the object layer to output data to the user, and uses the OQL that manages the object layer based on the input data from the user and the knowledge of the individual application. To management department 32
Send OQL (Object Query Language) commands. As an example, the OQL command to search for the names of people born in 1960 from personnel information is shown below.

(←PERSON GET NAME BIRTH 1960) The OQL management unit 32 analyzes and executes the input OQL command. When the operation target of the OQL command is in the object layer 12, if the operation result is obtained, it is transferred to the application layer 11.
send to When there is no operation target, convert OQL commands to XQL (Extended RDB Query Language) commands. This conversion uses a knowledge base to
Necessary optimizations are also performed when converting OQL to XQL. For example, change the execution order of commands that operate on the database so that the processing time is the shortest. The converted XQL command is sent to the XQL management section 33.

For example, the XQL command that corresponds to the OQL command example above is an instruction to select a tuple that matches BIRTH=1960 from the table "RERSON" and extract the field "NAME" from it. It takes shape.

PROJECTION〔NAME〕(SELECT〔BIRTH=
1960] (PERSON)) The XQL management unit 33 analyzes the input XQL command, performs optimization, and executes it targeting the extended RDB layer 13. The execution result is notified to the UQL management unit 31 through the object layer 12, and output from there to the terminal 2. Note that if the operation target does not exist, the user 1 is notified of this fact. In addition, the operation result is an object
The user can set it using an editor, or it can be written in an individual application program in advance.
Converts to the specified output format and outputs it, for example, in graph format.

User commands are provided graphically using applied knowledge and device information, and in a user-friendly format suitable for the application. Applied knowledge and device information are expressed in an object structure using frames.
stored in layers and extended relational databases (RDBs).

Next, each layer in the knowledge base will be explained.

The application layer 11 stores individual application programs, object editors, graphic interfaces, and the like.

Here, individual user commands are executed by individual application programs. A graphic interface is used for the user interface. The object editor is provided so that the user can easily define, change, and manage objects (internally, frames).

User commands include commands for starting application programs and object editors.
The user interface is a graphical interface that includes menus, icons, and commands.

FIG. 5 is an example of an object configuration graphed and displayed using the application layer's graphic interface. This represents the data model for “PERSON”.

Figure 6 similarly shows the graphical representation of the application layer.
This is an example of demographic statistics output as a graph using the interface.

The objects of object layer 12 are
Classified into three types: instances, classes, and metaclasses. An instance object is the smallest unit of data that represents something, and a class to which similar instance objects "belong" is called a class object. A class object has attributes that are common to instance objects belonging to the class and attributes that are related to itself. A class object itself also belongs to a class, and that class is called a metaclass.

The PERSON data model in Figure 5 mentioned above is
The relationship between these instances (INSTANCE), classes (CLASS), and metaclasses (META) is illustrated.

In this way, a "belonging relationship" can be provided as a relationship between objects. From the standpoint of knowledge representation, this is called ``abstraction by classification.'' In addition to this, in the object layer,
It is capable of ``abstraction by generalization,'' ``abstraction by aggregation,'' ``abstraction by evolution,'' and has a high level of knowledge expression ability.

FIG. 7 is an example of a frame representing a class object, and FIG. 8 is an example of a frame representing an instance object.

The frame representing the class object in FIG. 7 represents PERSON, that is, a data relationship (or structure) regarding information about a person.

Class, Super, Property in the frame,
Name, Address, Birth, and Age are slots, and represent the belonging class name, management frame name, attribute type, and attributes of name, address, date of birth, and age, respectively. Also value, multiple, mandatory
is a facet, value is a value, multiple is whether or not it can have multiple values (Yes/No),
Mandatory indicates whether the item is a mandatory item that must take a value (Yes/No).

The frame representing the instance object in FIG. 8 is information having attribute values of individual persons whose IDs are represented by P1 and P2.

P1 and P2 in the frame are slots, and their respective Name, Address, Birth, and Age are facets. The value of each facet gives the name, address, date of birth, and age of P1 and P2.

Next, the extended RDB layer 13 will be explained.

In the extended RDB layer 13, objects (i.e. frames) in the object layer 12
is stored in the table in the form of tuples (rows of the table).

There are two methods for storing frames: a "standard storage method" and a "transposed storage method."

The general form of relation in standard storage is:
It is expressed as follows.

META(FRAME, SFV(SLOT, FV
(FACET, VALUE))) The general form of the relation in the transposed storage method is expressed as follows.

Class i (instance-id, slot il, ..., slot in) Figure 9 is an example of a table in which the frame in Figure 7 is stored using the standard storage method, and Figure 10 is an example of a table in which the frame in Figure 7 is stored using the transposed storage method. This is an example of a table storing the frames of FIG. 8.

In each table, frames become tuples, and processing such as searching is performed on a frame-by-frame basis.

In the extended RDB, non-normal forms are allowed, and relations (tables) can be set as field values of relations (tables). Therefore, the hierarchical structure of knowledge can be expressed without redundancy.
It also has the advantage that the repeating structure of frames can be expressed concisely. FIG. 13 shows an example of a storage format of a non-normal relation using the example of instance P1 in FIG. 10. A in FIG. 13 is an example of a list format, and b in the same figure is an example of an array format. These methods can be used recursively when the internal relation hierarchy is deep.

Next, the configuration, version management, etc. of the extended RDB layer 13 will be explained. Extended RDB layer 13 consists of XRDB 131 for logs and individual
It is composed of an XRDB 132 and a shared XRDB 133.

All versions are stored in the individual XRDB and the shared XRDB.

The log XRDB stores the history (log) from the initial setting of the log XRDB until it is stored in the individual XRDB.

The shared XRDB stores only shared information,
General users are only allowed to search.

The individual XRDB stores all versions specified by the user. The user's object (frame) operations are almost identical to the data operations in the log XRDB, and when the individual XRDB version storage is specified, the latest information in the log XRDB is moved to and stored in the individual XRDB.

The log XRDB contains the necessary information in the shared XRDB.
data is transferred from the individual XRDB, and data operations are performed according to the user's object operations.

〔Effect of the invention〕

According to the present invention, by using an extended RDB layer, etc., it becomes possible to manage, maintain consistency, and share a large amount of knowledge (frames, objects) and large-scale knowledge. Furthermore, by using the object layer and extended RDB layer, processing such as searching, inserting, deleting, and updating knowledge can be speeded up.

[Brief explanation of drawings]

Figure 1 is a diagram showing the basic configuration of the knowledge processing system according to the present invention, Figure 2 is a diagram showing the configuration of an embodiment of the system according to the invention, and Figure 3 is a flow diagram of the processing of the OQL management section in the embodiment system of Figure 2. , Figure 4 is a processing flow diagram of the XQL management section, Figures 5 and 6 are graph output examples from the application layer, Figure 7 is an explanatory diagram of an example frame representing a class object, and Figure 8 is an illustration of an example of a frame representing a class object. 9 is an explanatory diagram of an example of a frame representing an instance object, FIG. 9 is an explanatory diagram of an example of a table using the standard storage method, and FIG.
The figure is an explanatory diagram of an example of a table using the transposed storage method, Figure 11 is a configuration diagram of a knowledge processing system using conventional frames, Figure 12 is a configuration diagram of a database using a semantic data model, and Figure 13. is an explanatory diagram of an example of a storage format of a non-normal relation. In Figure 1, 3: knowledge processing device, 4: main memory, 5: processor, 6: secondary storage, 10:
Knowledge base, 11 application layers, 12 object layers, 13 extended RDB layers.

Claims (1)

[Scope of Claims] 1. A knowledge processing system that includes a knowledge base and performs knowledge processing, which comprises: an application layer 11 that includes an application program group, a graphics interface, an object editor, etc., and a knowledge data model. An object layer 12 that holds a frame that represents the object structure as an object structure.
and an extended RDB that can hold frames representing individual knowledge as tuples of non-normal relations.
A knowledge processing system characterized by being configured in three layers including layer 13. 2 In item 1, the application layer 11 and object layer 12 are managed on the main storage device 4, but the extended RDB layer 13, which contains a large amount of knowledge that cannot be managed on the main storage device 4, is stored in the secondary storage device 6. The application layer is started based on an inquiry command from the user, the application layer accesses the object layer to obtain the necessary knowledge, performs processing, and outputs the processing results to the user. , a knowledge processing system characterized in that when the knowledge required by the application layer cannot be obtained from the object layer, the extended RDB layer is accessed to obtain the necessary knowledge.