JPH0855031A - Cooperative knowledge base system - Google Patents

Abstract

PURPOSE:To write a problem which can not be solved by a self-knowledge base device in a problem notice board device and request another knowledge base device to solve the problem. CONSTITUTION:A system control means 20 lets the problem inputted through an input/output part 10 be inferred by activating an inference engine 50 from an inference control engine 30 and using a knowledge base 60. When the problem is not solved, the system control means 20 informs the problem notice board device 3 of the problem through a presenting/fetching means 120 and the problem notice board device 3 receives the problem through a reception means 140 and holds it in a notice content storage part 160 through a management means 150. The system control means 20 requests the investigation of the new problem to the problem notice board device 3 through the presenting/ fetching means 120 at every constant interval of time. When the new problem from another knowledge base device 1 is present, the problem notice board device 3 extracts the problem from the notice content storage part 160 and informs the self-knowledge base device 1.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a cooperative knowledge base system, and more particularly to a cooperative knowledge base system that cannot be solved by a single knowledge base device by using the knowledge base of all distributed knowledge base devices connected by a network. A collaborative knowledge base system applicable to various problems.

[0002]

2. Description of the Related Art Conventionally, as a technology of a collaborative knowledge base system, a problem solving device described in Japanese Patent Laid-Open No. 4-279935 (hereinafter referred to as Document 1) and Japanese Patent Laid-Open No. 2-1143
There is an inference controller and the like in a distributed knowledge base system described in JP-A-33 (hereinafter referred to as Document 2). In the problem solving device of Document 1, when a problem is given, a plurality of agents having different knowledge and dispersed create a solution plan,
These solutions are negotiated between agents to create solutions that each agent allows. In the inference control device of Document 2, the structured knowledge bases are held in a distributed manner, and they are connected using a device for connecting them,
The problem is solved by using the plurality of knowledge bases and the knowledge base selection knowledge.

[0003]

The problem retrieval device described in the above-mentioned document 1 has a problem that it cannot be used alone because it is an agent having no user interface. In addition, since the model of another agent is retained, when an agent is added, it is necessary to rewrite the model of the agent added to all of the existing agents, and the extensibility is poor. There is a problem in that the feasibility is questioned because the details are not shown, and there is no guarantee that a solution that will satisfy all the agents will be generated. Furthermore, the load of each agent is not taken into consideration, and it is impossible to solve the problem efficiently by distributing the load.

[0004] Further, the inference control device in the distributed knowledge base system described in Document 2 has a selection knowledge of which one of a plurality of knowledge bases is to be used to solve a problem. There is a problem that extensibility when a base is added is poor. Further, there is no description when a plurality of inference control devices are connected, and there is a problem that the applicable range of the problem is narrow.

Furthermore, in both the devices of Documents 1 and 2, there is no structure for utilizing the result of problem solving, and there is no means for correcting the knowledge base initially constructed, so that the experience of problem solving cannot be utilized and the problem solving There is a problem that it cannot be expected to improve the ability.

SUMMARY OF THE INVENTION In view of the above, it is an object of the present invention to write a problem not described in the knowledge base of its own knowledge base device into a problem bulletin board device, request a problem solution to another knowledge base device, and receive a problem solution result. It is an object of the present invention to provide a collaborative knowledge base system which can be presented to a user by using the system.

Another object of the present invention is to take in a problem that cannot be solved by another knowledge base device from the problem bulletin board device, and judge the possibility of solution based on the solution capability and device load of the own knowledge base device. Then, if there is a solution possibility, it is to provide a collaborative knowledge base system that tries to solve the problem and sends the result of the problem solution to the problem sending knowledge base device.

Another object of the present invention is to prevent a knowledge base device having a high load from executing inference so as to prevent an increase in the waiting time for problem solving (inference), while providing a knowledge base device on all the networks with a high load. It is another object of the present invention to provide a cooperative knowledge base system capable of avoiding a temporary increase in network load by not transmitting a problem.

Still another object of the present invention is to provide a collaborative knowledge capable of compensating for the lack of knowledge in the knowledge base of the own knowledge base device based on the problem solving result from another knowledge base device. To provide a base system.

Another object of the present invention is a collaborative type in which, when a plurality of problem solving results are received, their reliability is evaluated, sorted in order of reliability, and then presented to the user. To provide a knowledge-based system.

Further, another object of the present invention is to be used for a series of claims processing from failure diagnosis of equipment and the like, failure acceptance to failure diagnosis and troubleshooting, etc., and is described in the knowledge base of the own knowledge base apparatus. For failures that have not been
An object of the present invention is to provide a collaborative knowledge base system in which a problem is posted on a problem bulletin board device to request a solution to another knowledge base device.

[0012]

A collaborative knowledge base system according to the present invention includes a knowledge base that holds knowledge for problem solving, an inference engine that solves a problem using this knowledge base, and an own knowledge. A plurality of knowledge base devices having an inference control engine that detects the load of the base device and determines activation of the inference engine and an input / output unit that exchanges information between the user and the knowledge base device are connected via a network. In the knowledge base system, a problem bulletin board device, which is a problem posting area common to all knowledge base devices, is provided on the network so that another knowledge base device can solve a problem that cannot be solved by the own knowledge base device. The base device stores a problem storage unit that stores the problem, a result storage unit that stores the problem solution result, and the problem bulletin board device displays the problem. In addition, a bulletin board / capturing means for capturing a problem from the problem bulletin board device, a cooperative means for transmitting a problem solution result to a problem transmitting knowledge base device and receiving a problem solution result from another knowledge base device, and the above respective means as a whole. A system control means for controlling is provided.

[0013]

According to the collaborative knowledge base system of the present invention, by combining a plurality of knowledge base devices connected via a network and a problem bulletin board device, a problem not described in the knowledge base of the own knowledge base device can be solved. Requests the other knowledge base device to solve a problem, and attempts to solve a problem that cannot be solved by the other knowledge base device when there is a possibility that it can be solved. As these methods, in addition to the knowledge base device, a problem bulletin board device is placed on a network, and a problem that cannot be solved by the knowledge base device can be written here, and the problem bulletin board device is investigated at regular time intervals. Detect new problem writing. Each knowledge base device is provided with a knowledge base essence obtained by extracting only the summary of the knowledge base possessed by the knowledge base device, and it is judged whether the problem written in the problem bulletin board device can be solved or not. The load is checked and the problem is solved. If the load is not high, an attempt is made to solve the problem, and the problem bulletin board device uses its own knowledge base to inform other knowledge base devices that the problem is solved. Describe the knowledge base device name, which is the name of the device. The result of the problem solution is sent to the problem originating knowledge base device regardless of its content, and the own knowledge base device name posted on the problem bulletin board device is deleted. Upon receiving the problem solution result, the knowledge base device holds the problem solution result in the result storage section, and when the user requests the result presentation, the reliability evaluation means evaluates the reliability and sorts in order of reliability. And present it to the user. Further, the problem input through the input / output unit is first inferred using the knowledge base of the self-knowledge base device, and if the problem cannot be solved, it is passed to the problem bulletin board device through the posting / capturing means. The problem bulletin board device receives the problem and the knowledge base device name through the reception means,
It is stored in the bulletin content storage unit through the management means. afterwards,
When there is an instruction to take in a problem from another knowledge base device, the problem is extracted from the bulletin content storage unit and passed to the knowledge base device. As described above, a problem not described in the knowledge base of the own knowledge base device can be solved by posting the problem on the problem bulletin board device by using the knowledge base of another knowledge base device. In addition, knowledge-based devices with a high load do not perform inference, thereby preventing an increase in the waiting time for problem solving (inference). On the other hand, problems are not transmitted to all knowledge-base devices on the network at the same time. A significant increase can also be avoided. Furthermore, when a plurality of problem solving results are received from another knowledge base device for a problem that could not be solved by the own knowledge base device, sorting is performed in an appropriate order, so that it is possible to shorten the problem solving time. Further, the lack of knowledge in the knowledge base of the own knowledge base device can be compensated based on the result of problem solving from another knowledge base device.

[0014]

The present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a cooperative knowledge base system according to an embodiment of the present invention. The cooperative knowledge base system of the present embodiment is configured by connecting a plurality of (two in the illustration) knowledge base devices 1 and a problem bulletin board device 3 to each other through a network (not shown).

The knowledge base device 1 is used, for example, in a series of claims processing work from the acceptance of failure contents of devices and the like to failure diagnosis and troubleshooting, and input / output for exchanging information with users. And a system control unit 20 for controlling each unit of the knowledge base device 1, an inference control engine 30 for controlling the inference engine 50, and the inputted problem and the problem taken from the problem bulletin board device 3. Problem storage unit 40, a knowledge base 60 holding knowledge for solving a problem, an inference engine 50 for actually performing inference using the knowledge base 60, and an inference engine 50
An inference history storage unit 70 for recording an inference history (inference execution progress data), an extracting unit 80 for extracting an abstract of the knowledge base 60, and a knowledge base essence 90 which is a collection of abstracts extracted by the extracting unit 80. , Problem memory 4
A result storage unit 100 for storing a problem solution result for the problem stored in 0; and a coordinating means 110 for receiving a problem solution result from another knowledge base device 1 and sending a problem solution result to the problem origination knowledge base device 1. Is stored in the result storage unit 100, and a notice / acquisition means 120 for notifying the problem bulletin board device 3 of the problem and periodically inspecting the problem bulletin board device 3 and taking in the problem when the problem is posted. A reliability evaluation unit 130 that evaluates the reliability of a plurality of problem solving results for one problem and sorts them in descending order of reliability, and a question creating unit 170 that creates a question in response to the problem solution failure result from the inference control engine 30. The main part is composed of.

The problem bulletin board device 3 is provided for each knowledge base device 1.
From the receiving means 140, a receiving means 140 that receives the instructions of the posting, investigation, cancellation, execution, and failure from, and interprets the content, the posting content storage section 160 that stores the problem sent from each knowledge base device 1, and the receiving means 140. And a management unit 150 that executes the instruction to the posted content storage unit 160 in response to the instruction regarding the posted content.

FIG. 2 is a diagram showing an example of connection between the knowledge base device 1 and the problem bulletin board device 3. The devices are connected to each other via a network, the knowledge base device 1 is connected via the coordinating means 110, and the knowledge base device 1 and the problem bulletin board device 3 are connected via the posting / capturing means 120 and the receiving means 140. Exchange.

The input / output unit 10 transfers information from the user to the system control means 20 and receives information from the system control means 20 and outputs it to an output device (not shown) of the knowledge base device 1. Present to the user.

The inference control engine 30 receives a problem from the system control means 20, the number of activations 34 of the inference engine 50 is less than or equal to a preset value of the number of activations, and the load of the entire knowledge base device 1 is in advance. When it is less than the set value of the determined load, the inference engine 50 is activated to pass the inference conditions (each set value is held in the set value 27).

FIG. 3 is a diagram showing an example of the content of a problem created by the system control means 20. The problem is composed of a problem number in the knowledge base device 1 where the problem exists, a problem number (hereinafter referred to as a local number) in the problem transmission knowledge base device 1, a problem transmission knowledge base device name, and an inference condition. . The inference condition is composed of a word string such as “computer, power supply, does not operate” and an attribute name = attribute value format such as “hardware = printer”. Here, the problem number is a number assigned for convenience by the system control means 20 to manage the problem in each knowledge base device 1, and is a unique number in the knowledge base device 1.

The inference control engine 30 receives the inference result from the inference engine 50, and when the solution status is "solved", adds the inference result to the problem and passes it to the system control means 20 as a problem solution result. . When the solution status is “unsolved”, the inference control engine 30 passes the problem solution failure result to the question creating unit 170.

FIG. 4 is a diagram showing an example of a problem solving result from the inference control engine 30. The problem solution result is a problem number, a local number, a problem origination knowledge base device name, an inference condition, a solution status, a solution result, and a process leading to solution (when the solution status is "solved") or It is composed of a reason for not being resolved (when the resolution status is “unresolved”) and a reason for judgment.

Further, the inference control engine 30 holds the number of currently activated inference engines 50 as the number 34 of activations. The inference control engine 30 increases the activation number 34 by 1 each time the inference engine 50 is activated, and decreases the activation number 34 by 1 each time the inference ends.

The problem storage unit 40 is a storage region of a problem input to the knowledge base device 1 by a user or an acquisition from the problem bulletin board device 3, and receives and holds the problem from the system control means 20. Further, the question storage unit 40, when requested by the system control means 20, delivers the question of the question number designated by the system control means 20.

FIG. 5 is a diagram showing an example of the contents of a question held in the question storage section 40. The problem storage unit 40 infers the same problem as the problem number, the problem number (local number) in the problem transmission knowledge base device 1, the problem transmission knowledge base device name, the inference condition, and the inference state of the problem. It holds a problem consisting of the name of another knowledge base device that is the inference execution knowledge base device name. Specifically, the inference state is "running" when the problem is being inferred, "not executed" when it is not executed, "solved" when it is resolved, and "unresolved" when it cannot be resolved. When posted on the problem bulletin board device 3, "posting" is entered.

The inference engine 50 is the inference control engine 3
It is started from 0 and the inference condition is passed. Inference engine 5
0 performs inference using the knowledge base 60 and passes the inference result to the inference control engine 30. The inference engine 50
There may be multiple launches for each problem at the same time. Further, the inference engine 50 passes the inference history to the inference history storage unit 70.

The knowledge base 60 is knowledge used at the time of inference by the inference engine 50, and its expression form changes depending on the inference system of the inference engine 50.

The inference history storage unit 70 is a storage area for the inference history. FIG. 6 is a diagram illustrating an example of the storage contents of the inference history storage unit 70. The inference history storage unit 70 holds an inference history including a problem number in the inference history storage unit 70, an inference date and time, a problem transmission knowledge base device name, and an inference process content.

The extracting means 80 summarizes the knowledge in the knowledge base 60. For example, as shown in FIG.
f. ． ． then. ． ． If the format is
The extracting means 80 extracts the words of the condition part of the knowledge, outputs a summary, checks whether the contents of the existing knowledge base essence 90 and the summary overlap, and registers the summary in the knowledge base essence 90 if they do not overlap. I do.

The knowledge base essence 90 is a set of summaries of the knowledge base 60, and the summaries are composed of words as shown in FIG.

The result storage unit 100 is the system control means 2
It is a storage area of the problem solution result passed from 0, and holds the problem solution result organized by problem number. The items stored are the same as the items of the problem solving result shown in FIG.

The coordinating means 110 notifies the other knowledge base device 1 of the problem solution result and receives the problem solution result from the other knowledge base device 1. That is, the cooperation means 11
0 receives the problem solving result from the system control means 20 and notifies the knowledge base device 1 described in the problem transmitting knowledge base device name of the problem solving result of the problem solving result as notification data. Further, the coordinating means 110 passes the notification data received from the other knowledge base device 1 to the system control means 20 as it is.

FIG. 8 is a diagram showing an example of the content of the notification data to be notified to the other knowledge base device 1. This notification data is composed of the name of the source knowledge base device and the result of solving the problem.

The posting / capturing means 120 executes data exchange between the knowledge database device 1 and the problem bulletin board device 3. The data transmitted and received are an instruction to the problem bulletin board device 3 and a question posted on the problem bulletin board device 3.
The bulletin / take-in unit 120 receives the instruction data from the system control unit 20 and transmits it to the problem bulletin board device 3.

FIG. 9 is a diagram showing an example of the contents of the instruction data. This instruction data includes an instruction from the problem bulletin board device 3 to the knowledge base device 1, an instruction from the knowledge base device 1 to the problem bulletin board device 3, a problem number in the problem bulletin board device 3, and a problem number in the instruction transmission knowledge base device 1 ( (Local number), instruction sending knowledge base device name, and inference condition.

When the contents of the instruction data are transferred from the knowledge base device 1 to the problem bulletin board device 3,
"Instruction" when instructing to post to the bulletin board, "Investigation" when the bulletin board / capturing means 120 monitors at regular intervals, "Cancel" when the bulletin board is canceled, "Execution" when the problem is solved, and Failure to solve the problem One of the "failures" of time is selected. When the problem bulletin board device 3 is transferred to the knowledge base device 1, “cancel” when the problem solution is stopped is selected. Instruction dispatch knowledge base The device name and instruction are always filled in, the problem number is filled in except at the time of "investigation", and the inference condition is only filled in at "posting". From the problem bulletin board device 3, when the instruction is "investigation", one newly posted problem is sent, and when it is "send", the instruction sending knowledge base device name is sent. "Post"
Nothing is sent for and "cancel".

FIG. 10 is a diagram showing an example of notification data notified from the problem bulletin board device 3 to the knowledge base device 1. This notification data includes an instruction, a problem number in the knowledge base device 1 that is the notification source, a problem number (local number) in the problem bulletin board device 3, a problem transmission knowledge base device name, an inference condition, and inference execution. Knowledge base device name and

In the notice / capture means 120, the instruction is "investigation".
In the case of, the sent inference condition and the summary of the knowledge base essence 90 are compared, and if each word included in the inference condition exists above a certain preset value in the knowledge base essence 90, the problem is The problem including the reasoning condition is judged to be solvable and is passed to the system control means 20, and "execution" is given to the problem number in the problem bulletin board device 3 as the instruction. The number is entered and the problem is notified to the problem bulletin board device 3.

The reliability evaluation means 130 is provided in the result storage 10
Among a plurality of problem solving results relating to a certain problem number held at 0, a reliability which is a measure of which problem solving result is most reliable is calculated, and the reliability evaluation result is passed to the system control means 20. . The system control unit 20 passes the problem number to the result storage unit 100, selects a problem solution result regarding the problem number, and returns the selected problem solution result to the system control unit 20. The system control means 20
The selected problem solving result is passed to the reliability evaluation means 130.

The reliability evaluation means 130 determines that a problem solving result including many inference conditions given as a problem for a reason of determination has high reliability, and calculates the reliability based on this.

FIG. 11 is a diagram showing an example of the reliability evaluation result passed from the reliability evaluation means 130 to the system control means 120. The reliability evaluation result is the problem solution result selected from the result storage unit 100 using the problem number as a key (see FIG. 4).
) And reliability. For example, in FIG. 11, the reliability is calculated by the given (number of words) * 10 included in the term of the judgment reason.

The question creating means 170 receives the problem solving failure result (see FIG. 4) from the inference control engine 30,
Information on what conditions are missing for the reason for judgment is obtained, a question is created, and the question is passed to the system control means 20.

FIG. 12 is a diagram showing an example of the contents of a problem solving failure result including a question. The problem solving failure result including the question has a format in which a question is added to the problem solving failure result (see FIG. 4) received from the inference control engine 30. For example, from the problem solution failure result shown in Fig. 4, "What is HW? Please choose from the following. (PC9821A
p, PC9801DX, PC9801FA) ”is generated.

The knowledge replenishment engine 180 is started by the system control means 20 and generates knowledge missing in the knowledge base 60 from the inference history of the inference history storage unit 70 and the problem solving result from the result storage unit 100. The knowledge replenishment engine 180 discovers the type of missing knowledge that caused the inference failure from the inference history related to the problem that failed to be solved and the problem solution result obtained by the other knowledge base device 1 solving the problem, By extracting the missing knowledge from the result storage unit 100, a difference in knowledge is generated and additionally registered in the knowledge base 60.

As an example, if there is an inference history shown in FIG. 13 and there is a problem solution result in the result storage unit 100 shown in FIG. Focusing on “possible” and knowing that the problem solution failed because there is no condition of being impossible. on the other hand,
Since the inference has been successful including the condition of being impossible for the determination reason of FIG. 14, the condition of “impossible” and the solution result of “power source fuse destruction” are added to the knowledge base 60.

The receiving means 140 receives the notification data from the posting / taking-in means 120 of the knowledge base device 1 and notifies the managing means 150 of the received data. Reception means 1
40 receives the investigation result from the management means 150 and notifies the corresponding knowledge base device 1 of the investigation result as notification data. The notification data and the acceptance data are in the format shown in FIG.

The notice content storage section 160 is an area for storing / holding the questions sent from each knowledge base device 1. The problem notified in the form of the notification data in FIG. 8 and the storage time thereof are stored, and at the same time, the name of the knowledge base device that issued the investigation instruction to the problem bulletin board device 3 and the latest reference time are held.

FIG. 15 is a diagram showing an example of the stored contents of the posted content storage section 160. In the bulletin content storage unit 160,
The problem number in the posting content storage unit 160 of each problem, the problem number (local number) in the problem transmission knowledge base device 1, the problem transmission knowledge base device name, the problem registration time, the inference condition, and the problem A problem is stored which includes the name of the inferred execution knowledge base device to be solved and the latest reference time of the problem bulletin board device 3 of each knowledge base device 1.

When the instruction in the reception data is "posting", the managing means 150 adds the problem in the reception data to the posting content storage section 160. When the instruction is “investigation”, the management unit 150 acquires a problem having a reference time later than the reference time of the instruction transmission knowledge base device 1 from the bulletin content storage unit 160, and uses it as notification data as instruction data. 1 to the receiving means 140 to send to In the case of the first “investigation” by the instruction transmission knowledge base device 1, since the reference time is not described, the management unit 150 sets all the problems stored in the bulletin content storage unit 160 as notification data regardless of the time. It requests the receiving means 140 to send it to the instruction transmission knowledge base device 1. The instruction is "Cancel"
In the case of, the management means 150 displays the problem in which the instruction transmission knowledge base device name and the problem number match the posted content storage unit 16
Delete from 0. When the instruction is “execute”, the management means 1
50 adds the instruction transmission knowledge base device name to the inference execution knowledge base device name of the problem whose problem number matches in the bulletin content storage unit 160. If the instruction is “failure”, the management means 1
The reference numeral 50 deletes the instruction transmission knowledge base device name from the inference execution knowledge base device name of the problem having the same problem number in the bulletin content storage unit 160.

The system control means 20 distributes data from each means to appropriate means, and gives an instruction to each means. Further, in response to the instruction from the input / output unit 10, the instruction to each unit is performed. Further, the problem storage unit 40 is checked at regular time intervals, and a problem is re-solved for each problem whose inference state has not been executed. When an inference condition is input from the input / output unit 10, a conversion operation to a problem is performed. The largest of the question numbers of the questions stored in the question storage unit 100 is acquired, and one obtained by adding 1 to the acquired question number is assigned to the question number of the input inference condition, Number. The self-knowledge base device name is entered in the problem origination knowledge base device name, the input inference condition is entered in the inference condition item, and the other items are left blank. This completes the conversion to the problem. The questions input from the instruction / take-in means 120 include, as in the case of the input / output unit 10,
Obtain the problem number and enter the obtained problem number in the item of problem number.

Next, the operation of the collaborative knowledge base system of the present embodiment thus constructed will be described with reference to the flow charts of FIGS. 16 to 29.

(1) First, a case where an inference condition is input as a group of words from the user via the input / output unit 10 will be described (see FIGS. 16, 24 and 25). As an example, it is assumed that “computer, power supply, do not move” is input in the machine name kusui.

When the inference condition is input from the input / output unit 10 (step 1601), the system control means 20 converts the input inference condition into a problem and stores it in the problem storage unit 40 (step 1602). At this time, the inference state in FIG. 5 is assumed to be in execution. Next, the system control means 20
Hands over the problem to the inference control engine 30 and requests a solution (step 1603).

Inference control engine 30 that received the problem
Reads the current activation number 34 of the inference engine 50, compares it with the preset setting value 27 (step 1604), and if the activation number 34 and the setting value 27 are equal (Yes in step 1604), it indicates that. Notify the system control means 20.

The system control means 20 includes a question storage unit 40
The inference state for that problem to unexecuted (step 1
624), and the process ends.

If the number 34 of activations is not equal to the set value 27 (NO in step 1604), the inference control engine 30 increases the number 34 of activations by 1 (step 160).
5) The inference engine 50 is started by passing the inference condition (step 1606). Then, the inference control engine 30
Goes into a standby state.

The activated inference engine 50 executes inference based on the inference conditions (step 1607), and stores the inference history at this time in the inference history storage unit 70 (step 1608). Regardless of success / failure of inference, the inference engine 50 notifies the inference control engine 30 of the inference result (step 1609) and stops.

When the inference control engine 30 receives the inference result (step 1610), it decrements the activation number 34 by 1 (step 1611) and then determines whether or not the resolution status of the inference result is unsolved (step). 1612).
If the solution state of the received inference result is not unresolved (No in step 1612), the inference control engine 30
Add the inference result to the problem and pass it to the system control means 20 as the problem solution result.

Upon receiving the problem solving result (step 1613), the system control means 20 stores the problem solving result in the result storage unit 100 (step 1614), and the inference state of the corresponding problem in the problem storage unit 40 has been solved. (Step 1615). Next, the system control means 20 presents the problem solving result to the user via the input / output unit 10 (step 1616), and ends the processing.

On the other hand, when the solution state of the received inference result is unresolved (Yes in step 1612), the inference control engine 30 passes the question solution failure result obtained by adding the inference result to the question to the question creating means 170.

The question creating means 170 creates a question based on the reason for determining the problem solving failure result (step 1617), and passes the question to the system control means 20.

When receiving the question, the system control means 20 presents the question to the user via the input / output unit 10 (step 1618). When a new inference condition is obtained from the user for the question (Yes in step 1619)
Defines a new problem including the acquired inference condition, and repeats the processing from step 1602.

If a new inference condition cannot be obtained (No in step 1619), the system control means 2
0 requests the bulletin / acquisition unit 120 to post the problem together with the problem (step 1620), sets the inference state of the relevant problem in the problem storage unit 40 to the posting (step 1621), and ends a series of processing. I do.

A bulletin / take-in means 1 receiving a bulletin instruction of a problem
20 creates notification data to the problem bulletin board device 3 based on the received problem (step 1622), and issues a problem posting instruction to the problem bulletin board device 3 via the network (step 1623).

In the problem bulletin board device 3, the knowledge base device 1
Receiving the instruction to post a question from the notice / take-in means 120 via the receiving means 140 (step 15001), the management means 150 stores the question in the notice content storage unit 160 (step 15002).

(2) Next, an operation when a problem solving result is received and a result presentation request is made by a user will be described (see FIGS. 17, 24 and 27).

The coordinating means 110 passes the problem solving result received from the other knowledge base device 1 to the system controlling means 20 (step 1701).

The system control means 20 recognizes which problem is the result of the problem solving from the problem number of the problem solving result (step 1702), and stores the problem solving result in the result storage unit 100 (step 1703). The inference state relating to the problem in the problem storage unit 40 is set to solved (step 1704). Next, the system control means 20
If the number of problem solution results accumulated in the result storage unit 100 for the problem is larger than a preset value (Yes in step 1705), read all problem solution results for the current problem (step 1706),
The reliability evaluation means 13 includes all the read problem solving results.
0 (step 1707). Step 170
If the answer is no in step 5, the system control means 20 shifts the control to step 1717.

Receiving the problem solving result, the reliability evaluation means 130 calculates the reliability (step 1708), and notifies the system control means 20 of the calculation result (step 17).
09).

Upon receiving the calculation result (step 1710), the system control means 20 transmits the calculation result to the input / output unit 10
(Step 1711). Further, the system control means 20 instructs the posting / capturing means 120 to cancel the problem posting on the problem bulletin board device 3 (step 1712).

The notice / take-in means 120 having received the instruction for canceling the notice board creates notification data to the notice board apparatus 3 (step 1713), and instructs the notice board apparatus 3 to cancel the notice notice (step 1714). ).

In the problem bulletin board device 3, the knowledge base device 1
When the instruction for canceling the notice of the problem is received from the notice / take-in means 120 through the accepting means 140 (step 15001),
The management means 150 obtains the instruction transmission knowledge base device name (step 15007), obtains the problem number (step 15008), and deletes the problem whose problem transmission knowledge base device name and problem number match from the bulletin content storage unit 160. Then, a cancellation instruction is sent to the inference execution knowledge base device 1 (step 15010).

On the other hand, the system control means 20 monitors the input of the result presentation request from the user through the input / output unit 10 (steps 1715 and 1716). If the user makes a result presentation request (Yes in step 1717), the problem for which the problem solution result is presented is identified from the problem number entered by the user (step 1718), and step 1
Control is transferred to 706.

(3) Next, a case where a problem is fetched from the bulletin / take-in means 120 will be described (FIGS. 18 and 2).
4 and FIG. 28).

When the problem is received from the posting / capturing means 120 (step 1801), the system control means 20 stores the problem in the problem storage section 40 as the reasoning state is being executed (step 1802). Next, the system control means 20
Hands over the problem to the inference control engine 30 and requests a solution (step 1803).

Inference control engine 30 receiving the problem
Reads the current invocation number 34 of the inference engine 50 and compares it with the preset value 27 (step 1804). If the number of boots 34 is equal to the set value 27 (Yes in step 1804), the fact is notified to the system control means 20.

The system control means 20 includes a problem storage section 40.
The inference state for that problem to unexecuted (step 1
624), and the process ends.

If the number of activations 34 is not equal to the set value 27 (NO in step 1804), the inference control engine 30 determines whether the load on the own knowledge base apparatus 1 is high (step 1805). Own knowledge base device 1
If the load is high (YES in step 1805), the inference control engine 30 notifies the system control means 20 to that effect.
Tell.

The system control means 20 includes a question storage unit 40
The inference state for that problem to unexecuted (step 1
624), and the process ends.

If the load on the self-knowledge base device 1 is not high (NO at step 1805), the inference control engine 3
0 indicates that the system control means 20 executes the problem solution.
(Step 1806).

Upon receiving the problem solving execution notification (step 1807), the system control means 20 instructs the posting / capturing means 120 to notify the problem bulletin board device 3 that the problem solving will be executed (step 1808). .

The notice / take-in means 120 which has received the problem solving execution instruction creates notification data to the problem bulletin board device 3 (step 1809) and notifies the problem bulletin board device 3 of the execution of the problem solving (step 1810). .

In the problem bulletin board device 3, the knowledge base device 1
Is received from the notice / take-in means 120 via the accepting means 140 (step 15001),
The management means 150 acquires the name of the notification transmission knowledge base device (step 15011), acquires the problem number (step 15012), and sets the name of the notification transmission knowledge base device to the name of the inference execution knowledge base device of the problem having the same problem number. It is added (step 15013).

On the other hand, the inference control engine 30 that has completed step 1806 increments the number 34 to be activated by one (step 1811) and passes the inference condition to activate the inference engine 50 (step 1812). Thereafter, the inference control engine 30 enters a standby state.

The started inference engine 50 executes inference based on the inference conditions (step 1813) and stores the inference history in the inference history storage unit 70 (step 181).
4). Inference engine 5 regardless of success or failure of inference
0 notifies the inference control engine 30 of the inference result (step 1815) and stops the processing.

When the inference control engine 30 receives the inference result (step 1816), the inference control engine 30 transfers the problem solution result obtained by adding the inference result to the problem to the system control means 20 regardless of the contents of the solution state of the inference result, and activates it. Number 3
4 is decremented by 1 (step 1817).

When the system control means 20 receives the problem solution result (step 1818), the system control means 20 notifies the problem origination knowledge base device 1 of the problem solution result so that the coordinating means 110 is notified.
(Step 1819).

Upon receiving the problem solution result notification instruction, the coordinating means 110 creates notification data for the problem sending knowledge base device 1 (step 1820) and notifies the problem sending knowledge base device 1 of the problem solution result (step 1820). 1821).

On the other hand, the system control means 20 determines whether or not the problem has been solved based on the solution status of the received problem solving result (step 1822). Only when the solution status is unresolved (NO in step 1822), the notice / take-in means 120 is instructed to notify the problem bulletin board device 3 of the problem solution result (step 1823).

The notice / take-in means 120 receiving the instruction for notifying the problem solving result creates notification data (instruction is “failure”) to the problem bulletin board device 3 (step 1827), and the problem solving device 3 solves the problem. Notify the failure result (Step 1
828).

In the problem bulletin board device 3, the knowledge base device 1
When the problem solving failure result is received from the posting / capturing means 120 via the receiving means 140 (step 15001), the managing means 150 acquires the instruction transmission knowledge base device name (step 15014) and executes the solution. The problem number of the problem bulletin board device 3 is acquired (step 15015).
The instruction transmission knowledge base device name acquired from the inference execution knowledge base device name of the problem whose problem number matches the notice content storage unit 160 is deleted (step 15016).

If the problem is solved (step 1822)
Yes), the system control means 20
The inference state of the corresponding problem is set to solved (step 1
824). If the problem is not solved (step 18
22), the system control means 20
The inference state of the corresponding problem of 0 is unresolved (step 1825). After that, the system control means 20 stores the problem solution result in the result storage unit 100 (step 182).
6), end the process.

(4) Next, a case where a user makes a knowledge supplement request will be described (see FIG. 19).

When a knowledge supplement request is received from the user through the input / output unit 10 (step 1901), the system control means 20 reads the question stored in the question storage unit 40 (step 1902), and stores it in the result storage unit 100. The stored problem solving result is read (step 190)
3). Next, the system control unit 20 sends the read problem solving result to the knowledge replenishment engine 180 to issue a knowledge replenishment instruction (step 1904), and enters a standby state.

Knowledge replenishment engine 1 that has received the knowledge replenishment instruction
The 80 reads the inference history from the inference history storage unit 70 (step 1905), finds the type of missing knowledge using the problem solution result received from the system control means 20 (step 1906), and generates the missing knowledge (step 1906). Step 1907). Next, the knowledge supplement engine 180 registers the generated missing knowledge in the knowledge base 60 (step 190).
8) Further, the extraction means 80 is activated (step 190).
9), end the processing.

The extraction means 80 which has been started is the knowledge base 6
0 is read (step 1910), the contents are summarized (step 1911), and the knowledge base essence 9 is read.
0 is updated (step 1912).

(5) Next, a case where a cancellation instruction is given from the bulletin / take-in means 120 will be described (see FIG. 20).

The system control means 20 receives the cancellation instruction through the posting / capturing means 120 (step 200).
1) The problem stored in the problem storage unit 40 is read (step 2002), and a problem solution stop instruction is issued to the inference control engine 30 (step 2003).

The inference control engine 30, which has received the instruction to cancel the problem solution, specifies the inference engine 50 to be canceled based on the problem number among the problems received from the system control means 20 (step 2004), and specifies the specified inference. Engine 50
Is forcibly terminated (step 2005).

(6) Next, the monitoring of the problem bulletin board device 3 of the system control means 20 will be described (FIG. 21, FIG. 2).
4 and FIG. 26).

The system control means 20 causes the posting / acquisition means 1 after the elapse of a certain set time (step 2101).
20 is instructed to investigate whether or not a new problem is posted on the problem bulletin board device 3 (step 2102).

A notice / incorporation means 120 that has received a survey instruction
Creates notification data to the problem bulletin board device 3 (step 2103) and instructs the problem bulletin board device 3 to investigate a new problem (step 2104). After this, the posting / capturing means 120 enters a standby state.

In the problem bulletin board device 3, the knowledge base device 1
Receiving an instruction to investigate a new problem from the posting / capturing means 120 via the receiving means 140 (step 15001)
The management means 150 acquires the instruction transmission knowledge base device name (step 15003), acquires the reference time (step 15004), acquires a problem having a problem registration time newer than the reference time (step 15005), and acquires the acquired problem. Is sent to the knowledge base device 1 (step 1500).
6). However, if the instruction transmission knowledge base device name does not exist in the bulletin content storage unit 160, all problems are sent to the knowledge base device 1.

The posting / capturing means 120 receives the reply from the problem bulletin board device 3 (step 210).
5) Check whether a new problem exists (step 2106). If there is no new problem (No in Step 2106), the bulletin / capture means 120 ends the processing as it is. If there is a new problem (Yes in step 2106), the posting / capturing means 120 compares the inference condition of the new problem with the contents of the knowledge base essence 90 (step 2107), and the problem may be solved. When the determination is made (step 2108), the control is moved to step 1901 of the system control means 20. If it is determined that the problem cannot be solved (No in step 2106), post /
The capturing means 120 rejects the problem (step 210)
9), end the processing. However, when there are a plurality of new problems, steps 2107 and 2108 are executed for each problem.

(7) Next, the operation of the problem bulletin board device 3 when the user makes a survey request will be described (see FIGS. 22, 24 and 26).

The system control means 20 which has received the investigation request of the problem bulletin board device 3 through the input / output unit 10 displays
An instruction is issued to the fetching means 120 to determine whether a new problem has been posted on the problem bulletin board device 3 (step 220).
1).

The posting / capturing means 120, which has received the instruction to investigate the new problem, creates notification data to the problem bulletin board device 3 (step 2202) and gives an instruction to investigate the new problem to the problem bulletin board device 3 (step 2203). ). After this, post /
The capturing means 120 enters a standby state.

In the problem bulletin board device 3, the knowledge base device 1
Receiving an instruction to investigate a new problem from the posting / capturing means 120 via the receiving means 140 (step 15001)
The management means 150 obtains the instruction transmission knowledge base device name (step 15003), obtains the device reference time (step 15004), obtains a problem having a newer problem registration time than the reference time (step 15005), and obtains it. The problem is sent to the knowledge base device 1 (step 1500)
6).

The posting / capturing means 120 receives the reply from the problem bulletin board device 3 (step 220).
4) Check whether a new problem exists (step 2305). If there is no new problem (No in step 2205), the bulletin / capture unit 120 ends the process. When there is a new problem (Yes in step 2205), the posting / capturing means 120 compares the inference condition of the new problem with the contents of the knowledge base essence 90 (step 2206), and the problem may be solved. When the determination is made (YES in step 2207), the system control means 20
Control is transferred to step 1801. If it is determined that the problem cannot be solved (No in step 2207),
The notice / take-in unit 120 rejects the problem (step 2208) and ends the process. However, if there are a plurality of new problems, step 22
Steps 06 and 2207 are executed.

(8) Next, in order for the system control means 20 to execute the unexecuted problem, the problem storage section 40 is set at regular intervals.
(See FIG. 23).

After the lapse of the set time (step 230
1 is yes), the system control means 20
The problem 0 is read (step 2302), and the problems whose inference state is not executed are listed from the problems (step 2303). Next, the system control unit 20 sets the inference state of the listed problem in the problem storage unit 40 to “in execution” (step 2304), and if the problem is for the own knowledge base device 1 (yes in step 2305), step 160
3; otherwise (NO in step 2305), control is transferred to step 1803.

In the above embodiment, a cooperative knowledge base system used in a series of claim processing operations from failure diagnosis of a device or the like and failure content acceptance to failure diagnosis and troubleshooting has been described as an example. It goes without saying that the invention is not limited to a collaborative knowledge base system for claim processing.

[0114]

As described above, according to the present invention, a problem bulletin board device, which is a common problem posting area, is provided on the network, and the knowledge base device has a problem storage unit, a result storage unit, a posting / capturing means, and a cooperation. By providing the means and the system control means to enable the information exchange between the knowledge base device and other knowledge bases and the information exchange with the problem bulletin board device, the problem not described in the knowledge base is posted on the problem bulletin board device. If you do so, the knowledge base device with a low load solves and directly reports the problem solving result, so the network traffic is lower than issuing a problem solving instruction to all knowledge base devices, and the problem solving job of other knowledge base devices It also has the effect of not hindering.

Also, a problem that is not described in the knowledge base of the own knowledge base device is posted on the problem bulletin board device, the other knowledge base device is instructed to solve the problem, and the result of the solution can be received and presented to the user. There is.

Further, a problem that cannot be solved by another knowledge base device is fetched from the problem bulletin board device, and the solvability is determined based on the solving ability and device load of the own knowledge base device. For example, there is an effect that a problem can be solved and the result of the problem can be sent to the problem transmitting knowledge base device.

Further, by providing the inference history storage unit, the result storage unit, and the knowledge replenishment engine, it is possible to compensate for the lack of knowledge in the knowledge base of the own knowledge base device from the problem solution results from other knowledge base devices. The effect is that the knowledge of the knowledge base is enhanced as it is used.

Further, the inference is not executed in the heavily loaded knowledge base device to prevent an increase in the waiting time for problem solving (inference), while the problem is not transmitted to all knowledge base devices on the network at the same time. There is an effect that it is possible to avoid a temporary increase in the load.

When a plurality of problem solving results for one problem are received, their reliability is evaluated, sorting is performed in descending order of reliability, and the problem solving result can be presented to the user. effective.

Further, failures that are used for a series of claims processing from failure diagnosis of equipment and the like and failure content acceptance to failure diagnosis and failure handling, and that are not described in the knowledge base of the own knowledge base apparatus are problematic. By posting a problem on the bulletin board device, there is an effect that a solution can be requested to another knowledge base device.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a cooperative knowledge base system according to an embodiment of the present invention.

FIG. 2 is a diagram showing a connection example between a knowledge base device and a problem bulletin board device in FIG. 1;

FIG. 3 is a diagram showing a specific example of a problem given to the inference engine in FIG. 1;

4 is a diagram showing a specific example of a problem solving result output from the inference engine in FIG.

FIG. 5 is a diagram showing a specific example of a question held in a question storage unit in FIG. 1;

FIG. 6 is a diagram illustrating a specific example of an inference history stored in an inference history storage unit in FIG. 1;

FIG. 7 is a diagram showing an operation example of an extraction unit in FIG. 1;

FIG. 8 is a diagram showing an example of the content of notification data notified between the knowledge base devices in FIG. 1;

9 is a diagram showing a specific example of a problem given to the posting / capturing means from the system control means in FIG.

FIG. 10 is a diagram showing a specific example of a problem given to the posting / capturing means from the receiving means in FIG.

11 is a diagram showing a specific example of a reliability evaluation result given to the system control unit from the reliability evaluation unit in FIG. 1;

12 is a diagram showing a specific example of a problem solving failure result including a question given to the system control means from the question creating means in FIG.

13 is a diagram showing a specific example of an inference history stored in an inference history storage unit in FIG.

FIG. 14 is a diagram showing a specific example of a problem solution result stored in a result storage unit in FIG.

FIG. 15 is a diagram showing a specific example of a problem held in the bulletin content storage unit in FIG. 1;

16 is a flowchart showing a process when a problem is input from the input / output unit of the system control means in FIG.

FIG. 17 is a flowchart showing processing at the time of receiving a problem solution result from the coordination means of the system control means in FIG. 1;

18 is a flow chart showing a process when a problem is taken in from the notice / take-in means of the system control means in FIG. 1. FIG.

FIG. 19 is a flowchart showing processing when a user of the system control means in FIG. 1 issues a knowledge supplement instruction.

FIG. 20 is a flowchart showing a process of the system control means in FIG. 1 when receiving a cancellation instruction from the posting / capturing means.

FIG. 21 is a flow chart showing the processing when the system control means in FIG. 1 requests the posting / acquisition means to investigate.

FIG. 22 is a flow chart showing a process at the time of a survey request to the posting / capturing means of the system control means in FIG.

FIG. 23 is a flowchart showing a process for executing an unexecuted problem by the system control means in FIG. 1;

FIG. 24 is a flowchart showing a process at the time of receiving an instruction by the management means in FIG. 1;

FIG. 25 is a flowchart showing the processing when the management means in FIG. 1 receives a posting instruction.

FIG. 26 is a flowchart showing a process when the management means in FIG. 1 receives a survey instruction.

FIG. 27 is a flowchart showing processing when the management unit in FIG. 1 receives a cancellation instruction.

FIG. 28 is a flowchart showing a process when the management unit in FIG. 1 receives an execution instruction.

FIG. 29 is a flowchart showing a process when the management unit in FIG. 1 receives a failure instruction.

[Explanation of symbols]

 1 knowledge base device 3 problem bulletin board device 10 input / output unit 20 system control means 27 set value 30 inference control engine 34 number of activations 40 problem storage unit 50 inference engine 60 knowledge base 70 inference history storage unit 80 extraction means 90 knowledge base essence 100 result Storage unit 110 Coordinating means 120 Posting / capturing means 140 Accepting means 150 Management means 160 Posting content storage section 170 Question creating means

Claims (7)

[Claims] 1. A knowledge base that holds knowledge for problem solving, an inference engine that solves a problem using this knowledge base, and a load of an own knowledge base device is detected to determine whether to start the inference engine. In a knowledge base system in which a plurality of knowledge base devices each having an inference control engine and an input / output unit for exchanging information between a user and a knowledge base device are connected via a network, A problem bulletin board device, which is a problem posting area common to all knowledge base devices for allowing other knowledge base devices to solve unsolvable problems, is provided, and in the knowledge base device, a problem storage unit for storing problems,
A result storage unit for storing the result of problem solving, a posting / capturing means for posting a problem on the problem bulletin board device and taking in the problem from the problem bulletin board device, and sending the problem solution result to the problem sending knowledge base device and problem solving A collaborative knowledge base system comprising: a coordinating means for receiving a result from another knowledge base device and a system control means for comprehensively controlling the respective means. 2. The extracting means for extracting a knowledge base essence, which is a summary, from the knowledge base of the self-knowledge base device, and the knowledge base essence when the problem bulletin board device is investigated at regular intervals and a new problem is posted. Using the above, the above-mentioned bulletin board instructing to judge whether the problem can be solved by the self-knowledge base device and try to solve the problem if possible
/ It has a capture means and determines whether a problem that could not be solved by other knowledge-based devices can be solved by the knowledge-base essence and the load on the self-knowledge-based device. The collaborative knowledge base system according to claim 1, which solves a problem of a knowledge base device. 3. An inference history storage unit for storing an inference history of problem solving using the knowledge base device, a problem storage unit for storing a problem solved by another knowledge base device, and a problem solution result. And a knowledge replenishment engine for updating the knowledge base using the problem solving result of the result storing unit, and reflecting the problem solving result of another knowledge base in the knowledge base of the own knowledge base device The collaborative knowledge base system according to claim 1. 4. A reliability evaluation means for calculating a reliability, which is a measure of which of a plurality of problem solution results for the same problem held in the result storage section is the most reliable. The collaborative knowledge-based system of claim 1, comprising. 5. The cooperation according to claim 1, further comprising a question creating unit that receives a problem solution failure result from the inference control engine, obtains information about what is a missing condition for a reason for making, and creates a question. Type knowledge base system. 6. The problem bulletin board device includes a reception means for receiving a problem that cannot be solved by each knowledge base device, a bulletin board content storage section for posting a problem received by this reception means, and a bulletin board content storage section. 2. The collaborative knowledge base system according to claim 1, comprising a management means for managing. 7. The collaborative knowledge base system according to claim 1, which is used for a series of complaint processing operations from failure diagnosis of devices and the like, acceptance of failure details to failure diagnosis, and failure response.