JPH05150988A - Inference device - Google Patents

Abstract

PURPOSE:To improve the connectability of a relational data base and thereby, to expand an application field. CONSTITUTION:The inference engine 14 of an inference device 10 performs inference by using a knowledge base 12 to obtain the solution of a problem supplied from an input/output device 16. When such inference is performed, a data base instruction is generated by a data base instruction generating means 22 when it is required to obtain knowledge data from the relational data base 50. Also, data base output data obtained by the instruction is converted into the knowledge data usable by the inference engine 14 by a knowledge data conversion means 30. Thereby, the connectability of the relational data base can be improved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a storage means for storing a knowledge base relating to a problem to be processed, and an inference engine for inferring using knowledge data in the knowledge base to obtain a solution to a given problem. And an inference device having
The present invention relates to an inference device that improves connectivity of a relational database and can expand the field of application thereof.

[0002]

2. Description of the Related Art A general reasoning device such as an application of an expert system operates on a relatively small-scale workstation, which is independent of the core system to some extent. Also, there were relatively few that worked in cooperation with the core system.

However, in recent years, an increasing number of systems perform complex tasks close to those of the core system and inference devices that operate as subsystems of the core system in close cooperation with the core system.

In order for the core system and the inference device to cooperate closely with each other, the inference device must be able to handle the data of the core system. Therefore, various techniques have been disclosed for the reasoning device to exchange data with the core system.

The conventional information processing means such as the basic system and the relational database (hereinafter also referred to as RDB) are collectively referred to as a host hereinafter.

The expert system, which is one of the inference devices, mainly uses a knowledge base in which experience rules and knowledge of experts in the problem area to be processed are stored as knowledge data, and the application knowledge possessed by the inference engine is used. Using the selection execution function of, the solution of the given problem is sought while giving the framework for solving the given problem.

Therefore, all the information other than the information stored as the knowledge base, such as the information necessary for starting the inference and the information to be referred to in the inference process, are conventionally positioned as external data.

A technique for the inference apparatus to take in such external data from the procedural information processing means is disclosed in Japanese Patent Laid-Open No. 2-2.
27732. This Japanese Patent Laid-Open No. 2-2277
In the technology disclosed in No. 32, in an inference device designed to fetch information from a data area of procedural information processing means, the data area of the procedural information processing means is directly accessed and knowledge type data and procedural data By providing an external data operation means that performs bidirectional conversion between the two, it is possible to capture changes in the external situation as knowledge type data in real time.

Further, there is a method of transferring data between the inference apparatus and the host using a working memory.

FIG. 7 is a block diagram showing data transfer between a conventional inference device and RDB.

In FIG. 7, the inference apparatus 10 has an interface means 18a and a working memory 18b. The conventional information processing means 60 includes a database interface means 62a and a working memory 62b.
And have. The RDB 50 has a database management device 52 and a database file group 54.

Data transfer from the RDB 50 to the inference apparatus 10 is performed via the conventional information processing means 60.

That is, as indicated by the symbol P1, the conventional information processing means 60 first reads the data from the RDB 50 into the database interface means 62a inside the conventional information processing means 60. The data read into the database interface means 62a is written into the working memory 62b while being converted into a form of inference data that can be used by the inference apparatus 10 as indicated by a symbol P2. After that, the data in the knowledge data format written in the working memory 62b is the working memory 18b in the inference apparatus 10 as indicated by the reference symbol P3.
Transferred to.

According to the data transfer from the RDB 50 shown in FIG. 7 to the inference apparatus 10, even a relatively large amount of data can be efficiently transferred.

Further, in Japanese Patent Laid-Open No. 2-230331, in an inference device adapted to take in information from procedural information processing means, procedural data activated by said procedural information processing means and given from said procedural information processing means. There is disclosed a technique relating to an inference apparatus which is capable of capturing and processing a change in an external situation in real time as knowledge type data by including a data delivery means for converting the data into knowledge type data and writing the knowledge type data into the working storage area.

According to the technique disclosed in Japanese Patent Laid-Open No. 230331/1990, data can be efficiently transferred from the host such as RDB to the inference device.

[0017]

However, in the prior arts such as the technique disclosed in Japanese Patent Laid-Open No. 2-227732 described above, which is incorporated in the inference device for the predetermined external data, the predetermined data (fixed data) There is a problem that it can only be transferred. Therefore, assuming all cases, there is a problem that the delivery area increases.

Further, in the technique described in FIG. 7,
In most cases, data transfer between the RDB and the inference device must be performed by batch processing, so it is difficult to capture the external situation change in real time in the inference device.

In the above-mentioned Japanese Patent Laid-Open No. 2-230331, data can be transferred between the procedural information processing means and the inference device in real time, but the transfer timing is determined only by the procedural information processing means. There is a problem that it has been done.

Therefore, it is necessary to constantly transfer data between the procedural information processing means and the inference apparatus in order to always incorporate the latest external situation change. However, if the amount of data to be delivered is increased and the delivery time is extended, there is a problem that the real-time property of the delivered data is deteriorated.

Therefore, in any of the conventional techniques,
Data cannot be exchanged with the RDB or the procedural information processing means at the timing led by the inference device side.
Another problem is that it is difficult.

The present invention has been made to solve the above-mentioned conventional problems, and it is an object of the present invention to provide an inference apparatus capable of improving the connectivity of a relational database (RDB) and thereby expanding the field of application. To aim.

[0023]

According to the present invention, a memory means for storing a knowledge base relating to a problem to be processed and a knowledge data in the knowledge base are used for inference to obtain a solution to a given problem. In an inference apparatus including an inference engine, in order to obtain knowledge data used for inference to be executed from a relational database, a database instruction generating unit that generates an instruction of the relational database system, and the relational database according to the instruction. The above object is achieved by providing a knowledge data conversion means for converting the generated database output data into knowledge data that can be used by the inference engine, and making the use of the relational database convenient.

Further, the inference to be executed from now on comprises interface request judging means for judging whether or not the knowledge data obtained from the relational database is required, and the database instruction generating means is currently executing the inference or in the future. Since the knowledge data used for inference that is expected to be executed is obtained from the relational database, a database command generation unit that generates a command of the relational database system achieves the above-mentioned object and achieves processing efficiency. It is intended to improve.

Further, the database command generating means obtains the knowledge data used for the inference to be executed from the relational database by issuing the command of the relational database system.
A database command generation unit that generates a plurality of command groups as necessary, and further has a database command storage unit that stores the generated commands in the process of generating the command group by the database command generation unit. By doing so, the above-mentioned problems can be achieved and the processing efficiency can be improved.

[0026]

According to the present invention, in order to construct an inference apparatus more effectively in various fields and to simplify the maintenance of the data, the inference apparatus and the relational database (RDB) should be connected. This was done focusing on the points that must be improved.

By improving the connectivity between the inference device and the RDB, the application field of the inference device can be expanded.

At this time, the following technical consideration is required.

A1: Data transfer between the RDB and the inference device can be performed at the timing led by the inference device side.

A2: The knowledge data in the knowledge base related to the connection with the RDB should not be complicated by the description about the connection with the RDB.

A3: The configuration of means for connecting the RDB connected to the inference device to this inference device is reduced as much as possible.

A4: If possible, various RDBs can be connected relatively easily.

The present invention has been made in view of such technical considerations.

FIG. 1 is a block diagram showing the gist of the present invention.

As shown in FIG. 1, the inference apparatus 10 of the present invention first comprises a knowledge base 12 and an inference engine 1.
4 and an input / output device 16. Further, the inference apparatus 10 includes a database command generation means 22 and a knowledge data conversion means 30.

On the other hand, the relational database 50 connected to the inference apparatus 10 of the present invention comprises, for example, a database management device 52 and a database file group 54.

The knowledge base 12 is constructed by storing knowledge data relating to a problem to be processed on a predetermined storage means.

The inference engine 14 infers using the knowledge data in the knowledge base 12 in order to obtain a solution to a given problem.

The input / output device 16 is the inference device 10 concerned.
It has an HMI (human machine interface) function with other users, and may have a connection function with other hosts. The input / output device 16 inputs the problem inferred by the inference device 10 and outputs the solution of this problem. It also inputs and outputs the data required during the inferences made to obtain the solution to this problem.

The database command generation means 22 generates a command of the system of the relational database 50 when the knowledge data used for the inference executed by the inference engine 14 must be obtained from the relational database 50.

The present invention does not limit the input path and the form of the database command generating means 22 (the broken line portion in FIG. 1 and the like). For example, the means for determining whether or not the knowledge data used for the inference executed by the inference engine 14 has to be obtained from the relational database 50 is not limited. For example, the determination means may be the inference engine 14 itself, or may be provided with a dedicated means such as the interface request determination means of the embodiment described later.

The knowledge data converting means 30 follows the command generated by the database command generating means 22.
The database output data generated by the relational database 50 is converted into knowledge data usable by the inference engine 14.

The database output data output from the normal relational database 50 has a data format different from the inference data, such as file format data. The knowledge data conversion means 50 converts the database output data in such a format into knowledge data in a format that can be used by the inference engine 14, such as a frame format.

The database file group 54 of the relational database 50 is composed of a plurality of predetermined files. The database management device 52 adds, changes, and deletes data in the database file group 54 by performing a predetermined relation operation or the like. In addition, the database management device 52, according to an external command,
The requested data is generated from the database file group 54. When generating the requested data, a plurality of files may be combined, data may be selected, and relational operations such as projection may be performed.

Therefore, the database instruction generating means 2
The command generated by 2 is a command of the system of the relational database 50. The data input from the relational database 50 by the knowledge data conversion means 30 of the inference apparatus 10 is output as a response to the instruction transferred from the database instruction generation means 22 to the relational database 50. The data is in the format of the type database 50 system.

According to the configuration of the present invention as described above, the timing of data transfer between the inference apparatus and the relational database is led by the inference apparatus. Also, various data conversions at this time are performed on the inference device side, so that the connectivity to the relational database can be improved. As a result, the connectivity of various hosts having a relational database with the backbone system can be improved, so that the application field of the inference apparatus can be expanded. Further, in the present invention, the above-mentioned technical consideration items A1 to A4 can be almost achieved.

It should be noted that the present invention does not limit the knowledge base, the inference engine or other components, and the configuration between them. For example, the individual configurations of the knowledge base and the inference engine are not limited. That is, according to the present invention, a database command generating means for generating a command of the relational database system for obtaining knowledge data used for inference to be executed from the relational database, and a relational database is generated according to the generated command. The knowledge data conversion means for converting the database output data into knowledge data that can be used by the inference engine of the inference apparatus of the present invention.

Further, the present invention does not limit the relational database connected to the inference apparatus of the present invention, and the data contained therein can be output in a predetermined format in accordance with some predetermined command. Anything will do.

[0049]

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 2 is a block diagram of hardware used in the embodiment of the present invention.

The inference apparatus of this embodiment is constructed with a relational database connected (coordinated) with the inference apparatus.

As shown in FIG. 2, the inference apparatus of this embodiment mainly includes a CPU 72, a main memory 74, a hard disk device 76, a magnetic tape device 78, and a CR.
It is built in a workstation that is composed of the hardware of the T device 80 and the keyboard 82.

FIG. 3 is a block diagram of the inference apparatus of this embodiment.

As shown in FIG. 3, the inference apparatus 10 of the present embodiment has a knowledge base 12, an inference engine 14, an input / output device 16, as in the inference apparatus 10 of FIG.
It has a database command generation means 22 and a knowledge data conversion means 30. Further, the inference device 10 of the present embodiment is
In addition to such a configuration, it has an interface request determination means 20 and a database command storage means 24.

The knowledge base 12 shown in FIG.
The inference engine 14, the input / output device 16, the database command generating means 22, and the knowledge data converting means 30 are substantially the same as those having the same reference numerals as described above with reference to FIG.

In the interface request judging means 20 of FIG. 3 provided in this embodiment, the inference data to be executed by the inference engine 14 is the knowledge data obtained from the relational database (the relational database 50 in FIG. 1). Determine whether or not

The input (determination target) of the interface request determining means 20 is read from the knowledge base 12 on the basis of the knowledge data being executed by the inference engine 14. That is, the interface request determining means 20 needs knowledge data obtained from the relational database for each preceding inference (which will be executed in the inference engine 14) prior to the inference in the inference engine 14. It is determined whether or not.

For example, when there is a possibility that the next inference may branch due to the result of the inference executed by the inference engine 14, the inferences of the case of not branching and the case of branching are obtained from the relational database. Determine the need for knowledge data. Further, similarly, for further inference, it is determined whether the knowledge data obtained from the relational database is required.

If the result of the determination by the interface request determining means 20 performed prior to the inference by the inference engine 14 is that the knowledge data obtained from the relational database is required, the database command is issued. The generation unit 22 generates the database command associated with this, prior to the inference by the inference engine 14. The subsequent processing of the database command storage means 24, the relational database management device and the knowledge data conversion means 30 is also executed prior to the inference by the inference engine 14.

Therefore, when the interface request judging means 20 is provided as in this embodiment, the inference by the inference engine 14 and the processing for obtaining the knowledge data from the relational database can be executed in parallel, The processing efficiency of the entire inference apparatus can be improved.

The database command generating means 22 shown in FIG. 3 obtains, if necessary, a command of the relational database system for obtaining knowledge data used for inference executed by the inference engine 14 from the relational database. A plurality of instruction groups are generated. Further, the inference apparatus of this embodiment includes the database command storage means 24 as shown in FIG. The database command storage means 24 stores the commands generated in the process of generating the command group in the database command generation means 22.

As described above, in the inference apparatus of this embodiment, the database command generating means 22 can generate a plurality of commands as a command group as needed. Therefore, for example, a plurality of related commands can be collectively transmitted to the relational database, and for example, by shortening the communication time between the inference apparatus and the relational database, and improving the processing efficiency of the relational database, etc. The processing time for the reasoning device to obtain the solution to a given problem can be improved.

FIG. 4 is a flowchart of the processing of this embodiment.

When a problem to be inferred is given in the inference apparatus of this embodiment, first, as shown in step 102, it is judged whether or not the connection with the relational database is necessary.
This is to judge whether there is a possibility that the knowledge data obtained from the relational database may be needed in inferring the problem given to the inference device, and if there is a need,
It is determined whether the connection state between the inference device and the relational database is currently being connected.

If it is determined in step 102 that connection to the relational database is required, the following step 104
Then, an interface between the reasoning device and the relational database is established.

Then, in step 108, the inference of the problem given to the inference apparatus is executed.

After performing the inference process for the given problem, in step 120, it is determined whether or not the connection between the inference device and the relational database is released. The determination of disconnection is made according to whether or not the connection between the inference device and the relational database is currently connected and whether or not this connection may be used thereafter. When it is determined in this step 120 that the connection has been released, the following step 122 is executed. On the other hand, if it is determined not to cancel, all the processing shown in FIG. 4 is ended.

In step 122, the interface between the inference device and the relational database is disconnected.

FIG. 5 is a flowchart of the RDB instruction generation processing executed in this embodiment.

The process shown in FIG. 5 is executed at any time during the process shown in step 108 of FIG.
In addition, the RDB instruction generation process shown in FIG. 5 is performed in parallel with the inference of the inference engine 14 when the inference engine 14 infers using the knowledge data in the knowledge base 12 to obtain the solution of the given problem. Then executed.

In FIG. 5, first, in step 142, the interface request determining means 20 is expected to be executed by the inference engine 14 of the knowledge base 12 while referring to the current inference portion in the inference engine 14. Read inference data.

Then, in step 144, it is determined whether the inference data read in in step 142 requires knowledge data obtained from the relational database when the inference engine 14 uses the inference data. .. If it is determined that this is necessary, the process proceeds to the following step 148, and if it is determined that it is not necessary,
All the processes shown in the flowchart of FIG. 5 are ended.

In the following step 148, the above step 1
At 44, a command of the system of the relational database for generating the knowledge data required to be obtained from the relational database is generated. When a plurality of instructions are required to obtain the necessary inference data from the relational database when the instructions are generated, the plurality of instructions are continuously generated. At this time, after the generation of these plural instructions, the processing of the following step 150 is performed.

This step 150 is the same as step 14 described above.
The command of the relational database system generated in 8 is transmitted from the inference device to the relational database.

FIG. 6 is a flowchart of the knowledge data conversion processing executed in this embodiment.

The processing shown in FIG. 6 is executed by the knowledge data converting means 30 shown in FIG. This knowledge data conversion processing is to convert the RDB output data output from the relational database corresponding to the RDB instruction generation processing of FIG. 5 into knowledge data that can be used by the inference engine 14. The process shown in FIG. 6 is executed during the inference process of step 108 in FIG.

In FIG. 6, first, in step 160, the RDB output data output by the relational database is read in correspondence with the RDB instruction generation processing of FIG. In the following step 162, according to the RDB output data read in the step 160, the information necessary for converting the RDB output data into the knowledge data format is read. At step 164, the RDB output data read at step 160 is format-converted into data (knowledge data) usable by the inference engine 14 while using the information read at step 162. Step 166
Then, the format-converted data in step 164 is written into the working memory readable by the inference engine 14.

As described above, according to this embodiment,
Data can be passed to and from the relational database at the timing led by the inference device. Further, since this data transfer is performed by a database command generated by the database command generating means, the knowledge data in the knowledge base related to the connection with the relational database is not unnecessarily complicated. The relational database side does not require any special configuration for connecting with the inference device.

[0079]

As described above, according to the present invention, the excellent effect that the connectivity of the relational database is improved, and thereby the reasoning device capable of expanding the application field can be provided. You can

[Brief description of drawings]

FIG. 1 is a block diagram showing the gist of the present invention.

FIG. 2 is a block diagram of hardware used in an embodiment of the present invention.

FIG. 3 is a block diagram of the inference apparatus of the above embodiment.

FIG. 4 is a flowchart of the processing of the above embodiment.

FIG. 5 is a flowchart of an RDB instruction generation process executed in the above embodiment.

FIG. 6 is a flowchart of a knowledge data conversion process executed in the above embodiment.

FIG. 7 is a block diagram showing data transfer between a conventional inference device and an RDB.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 ... Inference device, 12 ... Knowledge base, 14 ... Inference engine, 16 ... Input / output device, 18a ... Interface means, 18b ... Working memory, 20 ... Interface request determination means, 22 ... Database command generation means, 24 ... Database command storage Means, 30 ... Knowledge data converting means, 50 ... Relational database, 52 ... Database management device, 54 ... Database file group, 60 ... Conventional information processing means, 62a ... Database interface means, 62b ... Working memory, 72 ... CPU, 74 ... Main memory, 76 ... Hard disk device, 78 ... Magnetic tape device, 80 ... CRT device, 82 ... Keyboard.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Shoichi Moriya 2-3-2 Uchisaiwaicho, Chiyoda-ku, Tokyo Kawasaki Steel Co., Ltd. Tokyo headquarters (72) Inventor Shinya Arai 2-2-1 Uchisaiwaicho, Chiyoda-ku, Tokyo No. 3 Kawasaki Steel Co., Ltd. Tokyo head office

Claims (3)

[Claims] 1. An inference device comprising storage means for storing a knowledge base relating to a problem to be processed, and an inference engine for inferring using knowledge data in the knowledge base to obtain a solution to a given problem. In order to obtain the knowledge data used for the inference to be executed from the relational database, the database command generating means for generating a command of the relational database system, and the database output data generated by the relational database according to the command, An inference apparatus, comprising: a knowledge data conversion means for converting knowledge data that can be used by an inference engine, for the convenience of using a relational database. 2. The method according to claim 1, further comprising interface request judging means for judging whether the inference to be executed from now requires knowledge data obtained from the relational database, and wherein the database command generating means is currently used. In order to obtain knowledge data used for inference under execution or inference expected to be executed in the future from a relational database, it is a database command generation means for generating a command of the relational database system. Reasoning device. 3. The instruction of the relational database system according to claim 1, wherein the database instruction generation means obtains knowledge data used for inference to be executed from the relational database, A database command generating means for generating a plurality of command groups as needed, and further comprising a database command storing means for storing the generated commands in the process of generating the command group by the database command generating means. An inference device characterized by that.