JPH03116330A - Table form knowledge debugging system - Google Patents

Abstract

PURPOSE:To easily verify an inputted knowledge while interlocking with a table form knowledge editor by successively displaying the stage of decision tree search by the input picture of a table form to verify knowledge base. CONSTITUTION:In a knowledge base system 1, inference is performed by a decision tree 12, which is generated by a table form knowledge editor 3 and is stored in a knowledge base 2, to verify the knowledge. In this case, a knowledge expression picture of the same table form as the table form knowledge input picture used by the table form knowledge editor 3 is displayed when a table form knowledge debugger 4 is started. When conditions are gotton one by one, columns whose attributes are designated in accordance with values of conditions are successively emphasized and displayed; and at the time of arrival at a conclusion, the conclusion is emphasized and displayed. Thus, the image of the supplied table is used as it is and the search stage of the decision tree is referred to verify the knowledge supplied in the table form, thereby improving the debugging efficiency.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a tabular knowledge debugging method, and in particular, in a knowledge base system, knowledge is expressed in a tabular format by a tabular knowledge editor and knowledge data is input. The present invention relates to a tabular knowledge debugging method that allows input knowledge data to be easily verified using a tabular knowledge editor that creates knowledge decision trees.

[Conventional technology]

Knowledge base systems have been developed that solve problems on behalf of experts by expressing the knowledge of experts in a predetermined format and storing it in a knowledge base, and making inferences using the stored knowledge. In such a knowledge base system, expert knowledge is input expressed in a predetermined format, edited, and stored. The stored knowledge in a predetermined form is then used to make inferences that lead to conclusions.

The knowledge representation is expressed, for example, as tree-structured data as shown in FIG. Such tree-structured knowledge data is tree-structured data that indicates a path for deriving a conclusion from a plurality of conditions, and is called a decision tree. For example, a decision tree as shown in Figure 6 can be created using a tabular knowledge editor that expresses knowledge in a tabular format and inputs knowledge data.
It is generated from the decision table of the input knowledge.

In the decision tree shown in FIG. 6, 61, 62, 63 indicate condition judgment nodes, and each condition judgment node 61, 62,
The value of the block added below 63 indicates the value of the condition (determination value) for condition determination. The knowledge representation in this decision tree is such that, for example, if the decision value is al for condition A of condition decision node 61, the decision tree is traced downward and then condition B of condition decision node 62 is judged. It shows. Then, the process moves further down, and if necessary, for example, the condition C of the condition determination node 63 is determined. In this way, from the knowledge representation of the decision tree, the decision tree is sequentially traced down and Conclusions 1 to 8 are derived. In this way, a decision tree as shown in FIG. 6 is generated from the input knowledge data, and by following the generated decision tree, inference is executed using the knowledge data.

When different conclusions are drawn by combining the attributes of multiple conditions, a common method is to generate a decision tree as shown in Figure 6 and search this decision tree while obtaining the values of the conditions. .

By the way, in FIG. 6, the judgment values al, a2 . a3 indicates the attribute of condition A, and similarly, the determination values bl, . . . 03, etc. indicate the attribute of condition B, attribute 2, and condition C. In a decision tree search, first, when we obtain the value of condition A, we trace the branches that have attributes that match that value, then we obtain the value of condition B, and trace the branches that have attributes that match that value. . In this way, the decision tree is sequentially traced and a conclusion regarding condition C is reached in the same manner. For example, to reach the conclusion ttI'2, one condition A
.

The attributes of condition B and condition C are al, b2゜c',
1 (hereinafter referred to as (al, b2.cl)) is specified, and the value of the condition that matches this is obtained as a result. In some conclusions, attributes may not be specified for some conditions. If the unspecified attribute is expressed as 11 jj, for example, conclusion 4 has the attribute (B2.--) specified, and condition A is satisfied.
Conclusion 4 can be drawn just by obtaining B2 as the value of .

One method for inputting knowledge data to generate such a decision tree is a tabular knowledge editor that edits and inputs knowledge in a tabular format. In the tabular knowledge editor, conditions and conclusions are placed in the rows and columns of the table, and the attribute of the condition is specified in the column at the intersection of the table to express and input knowledge.

FIG. 7 is a diagram showing an example of a knowledge input screen using the one-table knowledge editor. In the example of the knowledge input screen in Figure 7, Conditions A9, Conditions B2, and Conditions C are arranged in the column direction for Conclusions 1 to 8 in the row direction, and in the column at the intersection of the table, the information for reaching the conclusion is Knowledge is expressed by specifying the attributes of each condition. The knowledge already input on this knowledge input screen corresponds to the decision tree shown in FIG. For example, regarding conclusion 2, the attribute of the condition for reaching that conclusion (a
1. 'b 2', c 1) specifies al in the column corresponding to condition A, and specifies b2. in the column corresponding to condition B2 and condition C. cl is specified and arranged in the vertical direction (column direction). Further, in order to reach conclusion 4, for the attribute of the condition (B2.-,-), B2 is specified in the column corresponding to condition A, and the columns corresponding to condition B2 and condition C are specified as blank. Here, the blank field implicitly specifies that it is not related to the value of the condition, but
For example, this may be specified as "Don't care" or "any" to clearly indicate that it has no relation to the value of the condition. The conclusions may be arranged interchangeably.

[Problem to be solved by the invention]

By the way, as described above, by using the tabular knowledge editor, knowledge can be easily input. However, if there is no debugging method linked to the tabular knowledge editor, there is a problem in that it becomes difficult to verify the input knowledge, and the effect of easy input cannot be fully utilized.

The present invention has been made to solve the above problems.

An object of the present invention is to provide a tabular knowledge debugging method that can easily verify input knowledge in conjunction with a tabular knowledge editor in a knowledge base system.

The above and other objects and novel features of the present invention will become apparent from the description of this specification and the accompanying drawings.

[Means to solve the problem]

In order to achieve the above object, the tabular knowledge debugging method of the present invention uses a tabular input screen in which conditions and conclusions are arranged in rows and columns, respectively, and displays the attributes of the conditions in the columns of the intersections of the table. In a knowledge base system that includes a tabular knowledge editor that generates a decision tree of knowledge data that specifies and leads to a conclusion from the attributes of one condition, a knowledge base that stores the decision tree, and an inference mechanism that makes inferences using the decision tree,
A first process is performed to trace the decision tree of the knowledge base and highlight columns with attributes that match the value of the condition among all columns in the row direction that correspond to one condition. The first process is repeated by sequentially changing the conditions according to key input instructions, and when a conclusion is reached, the second process is performed to highlight the conclusion. It is characterized in that knowledge data is verified by sequentially displaying it on an input screen.

In addition, if a conclusion is reached before all the values of the conditions arranged in the rows or columns of the table are obtained, before the conclusion is highlighted, The present invention is characterized by including a third process of highlighting a column specifying an attribute of an acquisition condition.

[Effect]

According to the aforementioned means, in the knowledge-based system,
Knowledge is verified by making inferences using decision trees created using the tabular knowledge editor and stored in the knowledge base. to display the inference process. That is, a first process is performed to highlight a column in which an attribute that matches the value of the obtained condition is specified among all columns in the row direction corresponding to one condition,
The first process is repeated by changing the conditions one by one in accordance with specific key input instructions. Then, when a conclusion is reached, a second process of highlighting the conclusion is performed.

In this way, the decision tree is followed and columns in the condition column of the knowledge input screen in table format are highlighted one after another.

Knowledge data is verified by sequentially displaying the process of searching a decision tree that leads to a conclusion.

Also, for columns that indicate the value of a condition whose value is not relevant when reaching a conclusion.

Since the conclusion is reached before all the values of the conditions arranged in the rows or columns of the table are obtained, in this case, before highlighting the conclusion, Highlight columns that specify attributes for unobtained conditions.

As a result, when debugging knowledge stored in the knowledge base, a tabular knowledge expression screen similar to the tabular knowledge input screen used in the tabular knowledge editor is displayed, and when a specific key input by the operator is triggered, As the conditions are obtained one by one, the columns with attributes specified according to the value of the conditions are highlighted in sequence, and when a conclusion is reached, the conclusion is highlighted, so knowledge can be expressed in a table format. The search process of the decision tree can be visualized on a tabular screen similar to the input screen where the information was entered. Therefore, the operator can easily verify his or her knowledge on the same screen as the tabular screen on which the input was made, for example.

〔Example〕

Hereinafter, one embodiment of the present invention will be specifically described using the drawings.

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

In FIG. 1, 1 is a knowledge base system, 2 is a knowledge base, 3 is a tabular knowledge editor, 4 is a tabular knowledge debugger, and 5 is an inference mechanism. The tabular knowledge editor 3 receives knowledge input from the expert 6 on a tabular screen, and generates a decision table 13 expressing the knowledge in a tabular format. Decision table 1
3 is a table generated by associating condition items and conclusion items with each row and column of the table, and specifying the condition value of each condition item corresponding to each conclusion item as an element of the table. This allows knowledge to be easily expressed and input by allowing unified handling in a predetermined tabular format. Also 1
The tabular knowledge editor 3 generates a decision tree 12 that derives a conclusion from the condition values of each condition item from the generated decision table 13, and stores the generated decision tree 12 in the knowledge base 2 so that it can be referenced in inference processing. Make it. For example, if user 7 inputs a question to knowledge base system 1,
Based on the decision tree 12 generated as described above, the inference mechanism 5 performs inference processing and derives a conclusion. Further, the tabular knowledge debugger 4 is a process that verifies the knowledge stored in the knowledge base 2, such as the decision tree 12, on a tabular display screen.

When the tabular knowledge debugger 4 is activated, a tabular knowledge expression screen similar to the tabular knowledge input screen used in the tabular knowledge editor 3 is displayed. Then, when the conditions are obtained one by one by a specific key input by an operator such as an expert with knowledge, columns with specified attributes are highlighted in sequence according to the value of the condition. . When a conclusion is reached, the conclusion is highlighted.

In this way, the decision tree search process can be visualized on a tabular screen similar to the input screen on which the knowledge representation is input in tabular form.

Therefore, the operator can easily verify his or her knowledge on the same screen as the tabular screen on which the input was made, for example.

FIG. 2 is a flowchart showing an example of the processing flow of debugging processing by the tabular knowledge debugger. Also,
FIG. 3, FIG. 4, and FIG. 5 are diagrams each showing an example of a display screen of debug processing by the tabular knowledge debugger.

When the tabular knowledge debugger 4 is started in order to verify the knowledge input in tabular form by the tabular knowledge editor 3, the tabular knowledge debugger 4 performs the processing shown in the flowchart of FIG. The screen displays shown in FIG. 4 and FIG. 5 are performed to proceed with the debugging process. That is, first, in step 21, a table format screen is displayed,
The table to be debugged is displayed on the screen, and then, in step 22, the values of the conditions are obtained and the decision tree is traced to search the decision tree. In the next step 23, columns of the table in which attributes matching the obtained values are specified are highlighted. In this example, when the value of one condition is obtained and the decision tree is traversed, the columns of the table whose attributes match the value of the previously obtained condition are highlighted before the next condition is obtained. do. At this time, if no attribute is specified for the condition, that is, the column corresponding to the condition in the table is also highlighted (FIG. 7), where the column is blank. This process is performed in step 24 by highlighting blank columns in the table whose attributes match the obtained values. On such a tabular knowledge screen, the process of highlighting the column of the value of the applicable condition can be done by, for example, retaining the contents of the table being displayed on the screen as a structure and checking all the conclusions. This is achieved by examining the attributes given to the conditions being viewed and highlighting columns on the screen by changing the attributes on the screen to reverse display, etc.

Next, in step 25, it is determined whether or not a conclusion has been reached. If the conclusion has not yet been reached at this point, the process advances to step 26, where the one-screen display is maintained as is.
Determine whether or not the funky key has been pressed. If it is determined that the function key has been pressed, step 22
Returning to step 22, the same process from step 22 is performed for the next condition. In this way, it waits for a specific key input from the operator of the tabular knowledge debugger, and when the key input is made, the value of the condition is obtained for the next condition, and the decision tree is searched. Highlight the columns in the table that have attributes that match the obtained values.

On the other hand, if a conclusion has been reached, the process proceeds to step 27, where columns of all tables included in the conclusion are highlighted. Then, in step 28, the conclusion base is highlighted and the process ends.

By the way, when a conclusion is reached in the search processing, a specific flag is set, and in the conclusion reaching determination processing, the flag is determined to determine whether or not a conclusion has been reached. When a conclusion is reached, the conclusion is highlighted by highlighting the column in which the conclusion base is displayed in the knowledge table expressed in a table format. Furthermore, prior to the process of highlighting a conclusion group, all columns arranged vertically under the conclusion group are highlighted.

According to the above processing flow, the changes in the display screen when the knowledge input to the knowledge base system is verified are explained in the seventh section.
To illustrate the case where a decision tree of input knowledge is verified using a tabular screen as shown in Fig. 3 and Fig. 6, respectively, the display as shown in Fig. 3, Fig. 4, and Fig. 5 is shown. It becomes a screen.

The example of the display screen in FIG. 3 is a display screen in which, for example, immediately after a3 is first obtained as the value of condition A, the value of this condition is determined and the corresponding column of the table is highlighted.

Here, among the columns arranged horizontally on the right side of the condition A column, the column in which the a3 attribute is specified, that is, the conclusion 5 to
The column under Conclusion 8 is highlighted (step 22.
23).

The remaining four columns, that is, the columns below Conclusions 1 to 4, are not highlighted because the condition values are not unspecified blanks (step 24).

In this state, the system waits for the function key to be pressed.

When the operator presses the function key after confirming the screen display shown in FIG. 3, the next condition is obtained and similar processing is performed. In this case, as is clear from the decision tree in Figure 6, the next condition necessary for the search is condition B.
It is. Here, when bl is obtained as the value of condition B, conclusion 5 is immediately drawn. At this time, b of the value of condition B
When l is obtained, among the columns arranged horizontally on the right side of the column for condition B, the columns in which the attribute of bl is specified, that is, the columns under conclusion 1 and conclusion 5, are highlighted, and furthermore, Among the remaining six columns, the blank column, that is, the column below conclusion 4, is also highlighted, resulting in a display screen as shown in FIG.

In this state, the decision tree search has already reached conclusion 5, so next, the column in which the conclusion base of conclusion 5 is specified is highlighted. In this case, by the time conclusion 5 is obtained by searching the decision tree, the blank column corresponding to the value for condition C is also highlighted, although the value for condition C has not been obtained. This is a process executed by highlighting all table columns included in the conclusion (step 27) when the conclusion is reached. In this way, when a conclusion is reached, a display screen as shown in FIG. 5 appears.

As a result, in addition to highlighting all the columns arranged in the vertical direction of conclusion 5, the column of the value of the condition that reaches conclusion 5 and the column of the conclusion base are highlighted.

The present invention has been specifically explained above based on examples, but
It goes without saying that the present invention is not limited to the embodiments described above, and can be modified in various ways without departing from the spirit thereof.

〔Effect of the invention〕

As explained above, according to the tabular knowledge debugging method of the present invention, it is possible to verify knowledge input in tabular format while observing the decision tree search process using the image of the input table as is. This has the effect of greatly improving debugging efficiency.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the main components of a knowledge base system according to an embodiment of the present invention, FIG. 2 is a flowchart showing an example of the processing flow of debugging by a tabular knowledge debugger, and FIG. , FIG. 4, and FIG.
Figure 6 shows an example of a display screen for debugging by a tabular knowledge debugger, Figure 6 shows an example of a decision tree generated from the knowledge representation decision table in Figure 5, and Figure 7 shows a table. It is a figure which shows an example of a knowledge input screen by a formal knowledge editor. In the figure, 1...Knowledge base system, 2...Knowledge base, 3...Tabular knowledge editor, 4...Tabular knowledge debugger, 5...Inference mechanism.

Claims (2)

[Claims] (1) Using an input screen in the form of a table in which conditions and conclusions are arranged in rows and columns, specify the attributes of the conditions in the columns at the intersections of the table, and determine the knowledge data that will lead to the conclusion from the attributes of the conditions. In a knowledge base system that has a tabular knowledge editor that generates a tree, a knowledge base that stores decision trees, and an inference mechanism that makes inferences based on the decision tree, the decision tree of the knowledge base is traced and a row that corresponds to one condition is Out of all the columns in the direction, perform the first process of highlighting the column with the specified attribute that matches the value of the obtained condition, and then change the condition one by one according to the instructions of a specific key input.
The first process is repeatedly performed, and when a conclusion is reached, a second process is performed to highlight the conclusion, and the process of searching the decision tree is sequentially displayed on a tabular input screen,
A tabular knowledge debugging method characterized by verifying knowledge data. (2) If a conclusion is reached before all the values of the conditions arranged in the column direction of the table are obtained, the unobtained conditions included in the column corresponding to the conclusion are highlighted before the conclusion is highlighted. 2. The tabular knowledge debugging method according to claim 1, further comprising a third process of highlighting a column specifying an attribute.