JPH05216674A - Knowledge expression system for explanation with calculator - Google Patents

Abstract

PURPOSE:To perform a detailed explanation on the internal knowledge with a calculator to users. CONSTITUTION:This knowledge expression system is composed of a matrix 100 of the nth level expressing the presence or absence of the relation between plural events, a matrix 200 of the n + 1st level describing the presence or absence of further detailed relation on the two relating events, an explanation 300 related further by a link 2003 on the relation between event items of the n + 1st level and a matrix 400 of more than the n + 2nd level specifying the presence or absence of the relation between further detailed event items which are related by a link 2005. By the matrixes having these plural hierarchies, the knowledge expression existing the hierarchy in the detail is possible, a partial explanation tree is composed of the explanations between events and between event items and the link to the matrix element, and the generation of the detailed explanation with a calculator becomes possible.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a knowledge representation system in a knowledge base system.

[0002]

2. Description of the Related Art Conventionally, in a diagnostic type expert system, a table format (matrix format) is used as a knowledge expression method that is most easily expressed. One example is knowledge representation by S / H matrix. The S / H matrix is a one-layer matrix in which phenomena (Sympton) and causes (Hypothesis) are arranged in rows and columns and the relationship between the two items is represented.

[0003]

However, in the conventional method, when the details of the relation of knowledge have a hierarchical structure,
I couldn't express it well. For example, in the problem of failure diagnosis, the case where the failure part is first roughly specified and then the detailed diagnosis is performed on this part corresponds to this. It is an object of the present invention to represent knowledge having multiple levels of detail etc. in a matrix having a plurality of hierarchies associated with each other, so that a computer can explain a diagnostic problem or an analytical problem to a human. Refine,
It is to provide a knowledge representation method for explanation by a computer for supporting knowledge acquisition from humans.

[0004]

According to the present invention, in a knowledge representation system of a knowledge base system, an n-th level matrix showing the presence / absence of a relation between a plurality of events and the n-th level matrix are used for the i-th row. Between more detailed event items relating to the relationship between the i-th event and the j-th event associated with the matrix element in the i-th row and the j-th column, when the event is marked as being related to the event in the j-th column. Associated with a matrix of the (n + 1) th level indicating the presence or absence of association and a matrix element marked as related between two events and event items,
This is a knowledge representation method characterized in that a partial explanation tree is constructed by having explanatory sentences that explain relationships between events and event items, and that a plurality of matrix hierarchies are provided.

[0005]

The knowledge representation method for explanation by the computer of the present invention will be described with reference to FIG. The n-th level matrix 100 expresses the presence or absence of association between a plurality of events.
The relationship between two events is the event 1001 marked on the line,
It is represented by a circle 1004 in the matrix element at the intersection of the events 1002 marked in the column.

The (n + 1) th level matrix 200 is related by a circle 1004 and a link 1003 marked on the nth level matrix. In this matrix, regarding the two events described as being related in the nth level matrix, the presence or absence of the relationship in more detailed items is described. That is, for the detail item 2001 of the event 1 written in the row and the detail item 2002 of the event 2 written in the column, as in the case of the first level matrix, the circle mark 200 of the matrix element at the intersection is shown.
4 represents the association between the two detail items.

Further, for example, regarding the relationship between the (n + 1) th level event items, the description 300 of the relationship between the event items defined therein is related by the link 2003 from the circle 2004 of the matrix element indicating the relationship between the event items. Described. The description between the event items can be attached to the matrix element at the nth level.

When it is necessary to define the presence or absence of a more detailed relationship between event items at the (n + 1) th level, the relationship is defined by the matrix 400 at the (n + 2) th level or higher. Similar to the matrix of the (n + 1) th level, the matrixes of the (n + 2) th level and above are related by the circles 2004 and the links 2005 provided between the event items.

The matrix having a plurality of hierarchies having the above-described structure enables the expression of knowledge having a hierarchical level of detail, which has been impossible in the past. Further, since the partial explanation tree is configured by the explanation between the events and between the event items and the link to the matrix element, the detailed explanation can be generated by the computer.

[0010]

EXAMPLE FIG. 2 and FIG. 3 show an example of the present invention.
Will be described. FIG. 2 is a diagram showing a part of the hierarchical knowledge for automobile failure diagnosis expressed by using the knowledge expression method of the present invention. The first level matrix 500 and the second level matrix 700 describe in rows and columns the phenomena that occur when an automobile fails and their causes. In the first-level matrix 500, circles 5001, 5002, and 5003 indicate the relationships with the events of “oil depletion”, “engine destruction”, and “cooling water depletion” that cause the phenomenon “overheating”.
Also, the explanations 600, 610, 620 of the relationships between the respective events are related to the circles by the links 5004, 5005, 5006. For example, the link between the event “overheat” and the event “engine destruction” is link 5005.
"Overheat due to engine destruction"
This is explained by (610). The link between the event "Overheat" and the event "Engine destruction" is Link 5
007 connects to a second level matrix 700 showing the relationships between more detailed event items.

The second level matrix 700 describes the presence / absence of a more detailed event necessary for identifying the relationship between the event "overheat" and the event "engine destruction". For example, the circle 7001 of the matrix element indicates that the detailed event “radical engine brake use” is considered as the cause of the detailed event “suddenly metallic noise from the engine”. Further, the triangle mark 7002 of the matrix element indicates that the degree of association between the events is lower than that indicated by the circle mark. The symbol written on the matrix element of the present invention is 1 or 0.
It is obvious that not only the digital amount such as the above, but also the analog amount represented by the triangle mark 7002 is possible. Further, as in the case of the first-level matrix, circles 7001 and 700 of matrix elements indicating the relationships between event items are shown.
3 and triangle 7002 and links 7004, 7005
7006 binds the related descriptions 800, 810, 820.

The knowledge of the vehicle failure diagnosis described above is
It is used in the process of the flowchart shown in FIG. P
At 1, the matrix level is initialized to 1. At P2, the input of the phenomenon of automobile failure is accepted from the user. P3, P
In step 4, the n-th level matrix matching the input phenomenon is searched, and the processing from the next P6 is performed each time a matching item is searched. Further, in P5, when the search for all the nth level matrices is completed, all the processing steps are completed. For the matching item, the partial explanation of the item is stored (P6). Matching item is the nth
If it does not have a +1 level matrix, the description is generated using the partial description stored in P8. When the matched item has a matrix of the (n + 1) th level, the level is incremented by 1 at P9 and the processing from P2 is repeated.

In the example of FIG. 2, the event "overheat" is first input by the user at P2. In the process of P3 to P8, the explanation about the two matching items that do not have the second level matrix is output. For the items having the second level matrix, the partial explanation “overheat due to engine destruction” is stored, and the second level matrix is searched. At this time, an input request for another phenomenon is issued to the user at P2. If the phenomenon input here is, for example, "there was a metallic noise from the engine before,"
Through the process described above for the first level matrix, a partial explanation that "metal noise is generated from the engine due to engine deterioration" can be obtained. Partial explanation and first obtained here
By using an appropriate explanation generation rule in P8 based on the partial explanation obtained from the level matrix, it is possible to obtain an explanation such as, for example, "engine aging caused engine destruction, resulting in overheating". ..

By the above, it is possible to obtain a computer explanation of the result of the diagnosis performed by the computer.

[0015]

As described above, according to the present invention, it is possible to give a detailed explanation to the user about the knowledge inside the computer. Furthermore, it is also possible to support knowledge acquisition from humans to the system in the expert system. Also, since it is a simple knowledge representation method, explanations can be generated at high speed.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing a knowledge representation method for explanation by a computer of the present invention.

FIG. 2 is a diagram showing an embodiment of the knowledge representation system of FIG.

FIG. 3 is a flowchart of a process using the embodiment of FIG.

[Explanation of symbols]

100 nth level matrix 200 n + 1th level matrix 300 Partial explanation of relationships between events and event items 400 n + 2 level matrix 500 First level matrix 600, 610, 620 for automobile fault diagnosis Inter-level event related description 700 Second level matrix for automobile fault diagnosis 800, 810, 820 Second level event Inter-item related description 1001 Event described in row of nth level matrix 1002 nth level Events marked in columns of matrix 1003 Links linking relationships between events and n + 1 level matrix 1004 Circles indicating relationships between events 2001 Event items marked in row of n + 1 level matrix 2002 Level n + 1 level Event items marked in the matrix columns 200 3 Link that links the relationship between event items and their partial description 2004 Circle mark that shows the relationship between event items 2005 Link that links the relationship between event items and the matrix of n + 2 level or higher 5001, 5002, 5003 Between the events of the first level A circle 5004, 5005, 5006 A link that links the relationship between events and their partial explanations 5007 A link that links the relationship between events and a second-level matrix 7001, 7003 Circles that show the relationship between event items 7002 Events Triangle mark showing relation between items 7004, 7005, 7006 Event Link linking relation between item and its partial explanation

Claims (1)

[Claims] 1. A knowledge representation method for a knowledge-based system, wherein an nth level matrix showing the presence or absence of a relationship between a plurality of events, and an event in the ith row and the jth column in the nth level matrix (N + 1) th level, which indicates whether there is a more detailed relationship between event items related to the relationship between the i-th event and the j-th event, which is associated with the matrix element at the i-th row and the j-th column, when it is marked as related to And a descriptive sentence that explains the relationship between the events and the event items, which is associated with the matrix element that is marked as related between the two events and the event items, and forms a partial explanation tree by having A knowledge representation method characterized by having multiple matrix layers.