JPH0228726A - Arrangement/integration system for frame type knowledge expression - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Summary] Regarding the organization and integration method for organizing and integrating frame-type knowledge expressions, duplicate descriptions of frame-type knowledge expressions are extracted and compiled as a superordinate frame, and the duplicated knowledge is organized and integrated. The aim is to reduce the burden on the developer by having the computer perform the process.The frame-type knowledge representation is a system that expresses knowledge by having multiple frames composed of slots and slot values in a hierarchical relationship, and A similarity calculation unit that calculates similarity in association with slots and slots 1 to values, or the matching number of slots, and a common attribute when the similarity calculated by this similarity calculation unit exceeds a predetermined threshold. The present invention includes a grouping processing section that creates a higher-level frame having a common attribute and deletes the common attribute from the original frame, and the grouping processing section organizes and integrates the common attributes.

[Industrial application field]

The present invention relates to an organization/integration method for organizing/integrating frame-type knowledge representations.

As knowledge representation in expert systems, there is a method that uses a tree-like data structure based on frames. When the amount of knowledge handled using this frame becomes enormous, the amount of data to be described will vary significantly depending on how the knowledge is organized, which will also affect its maintainability. Therefore, it is necessary to reduce the amount of data by organizing and integrating it as much as possible. It is hoped that

[Problems to be solved by conventional technology and inventions] Frame-based knowledge representation, which describes knowledge about a certain event on a computer, treats knowledge as a collection of various pieces of information and divides the subdivided information into items called slots. The slot attributes are further structured as slot values. For example, knowledge about trucks is expressed as shown in Figure 4 (a), where trunk is the frame name, year of production, capacity, etc. are slots,
Furthermore, in 1988, 3 people, etc. are slot values. In addition, as shown in Figure 4 (b), we create a hierarchical relationship between multiple frames, systematize knowledge, and allow lower-level concepts to inherit and use higher-level concepts, so that knowledge can be expressed without waste. I have to. For example, a car is a superordinate concept of a bus, a trunk, and a private car, and these subordinate concepts inherit the superordinate concept.

However, the frame-type knowledge representation that is not well organized is
It contains a lot of duplicate information, and as the amount of knowledge to be handled increases, the amount of data described by frames becomes enormous. For this reason, in the past, these overlapping contents have been sorted out by individual knowledge engineers based on their experience, which poses the problem of not being able to do it satisfactorily and creating a heavy burden.

The present invention extracts duplicate descriptions of frame-type knowledge expressions and compiles them into upper-level frames, and organizes and organizes duplicate knowledge.
The purpose is to reduce the burden on developers by having computers perform the integration.

[Means to solve problems]

FIG. 1 shows a basic configuration diagram of the present invention.

In FIG. 1, a frame-type knowledge representation 1 describes knowledge about frames having slots and slot values, with a hierarchical relationship between them.

The similarity calculation unit 2 calculates the similarity in association with the slots and slot values of a plurality of frames, or the number of matching slots.

When the calculated similarity exceeds a predetermined threshold, the grouping processing unit 3 creates a higher-rank frame having a common attribute and deletes the common attribute from the original frame.

[Effect]

In the present invention, as shown in FIG. 1, a similarity calculation unit 2 searches a frame-based knowledge representation 1 to find out slots and slot values, or the number of frames in which the slots match, and calculates the similarity from this number. When the calculated similarity exceeds a predetermined threshold, the grouping processing unit 3 creates the common attribute as a higher-rank frame and deletes the common attribute from the original frame.

Therefore, it is possible to extract common attributes (slots and slot values that a plurality of frames have in common, or slots) from the frame-type knowledge representation 1 and organize and integrate them into a higher-level frame. This can reduce the burden on knowledge engineers.

〔Example〕

Next, the configuration and operation of one embodiment of the present invention will be explained in detail using FIGS. 1 and 3 in accordance with the order shown in the flowchart of FIG. 2.

In FIG. 2, ■ performs 5ORTING (sorting). This is for efficiently processing the frames constituting the frame-type knowledge representation 1 in FIG. 1 by arranging them in ascending order of the number of slots.

② Select a frame. This means that the items are selected in order from the first item sorted by ■.

(2) calculates the degree of similarity. This means that when slots with the same name and the same slot value exist in a frame, the similarity is set to, for example, 0.9/number of slots for each slot, or when only slots with the same name exist, the similarity is set to, for example, 0.9/number of slots for each slot. This means that the total score is calculated by changing the weighting based on the number of 17 slots. For example, frame F001 in frame-type knowledge representation 1 in Figure 1,
When only slots S1 and S3 exist in FOO2, F2O3, 0.1/3 for each thrower I.
becomes.

(2) determines similarity>threshold. This means that it is determined whether the scores (similarities) sequentially calculated in (■) exceed a predetermined threshold. If YES, perform ■. In the case of No, perform the grouping process. This means that a new upper frame is created that has the common attributes (slots, or slots and slot values) of the frames whose similarity was calculated in step ①, and the common attributes are deleted from the original frame. There is.

Specifically, for lower frames F001 to FOO5 of frame type knowledge representation 1 in Figure 1, common attributes are extracted and new upper frames MOOL and M2O3 are created, and the common attributes are extracted as shown in Figure 3. Delete the attribute from the original frame Fool or F2O3. At this time, when frames that can be grouped are found, they are sequentially grouped into higher-level frames and common attributes are deleted from the original frame, so there is no need to have multiple higher-level conceptual frames.

(2) It is determined whether or not all slots have been completed.

If YES, perform ■. If NO, repeat the steps below.

(2) determines whether all frames have been completed.

If YES, the process ends. If NO, repeat the steps below.

As described above, for multiple frames constituting frame-based knowledge representation 1, by creating a new upper frame with a common attribute and deleting the common attribute from the original frame, the amount of data can be reduced by combining duplicate information. It becomes possible to reduce the

FIG. 3 shows an example of grouping. This means that the lower frames FO01 to F2 of frame-type knowledge representation 1 in Figure 1
Duplicate information from O3 to upper frame MOOL, M2O3
The duplicate information is deleted from the original frames F001 to FOO5.

This can be done by a computer, reducing the developer's burden.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram showing the principle configuration of the present invention, FIG. 2 is a flowchart explaining the operation of the present invention, FIG. 3 is an example of grouping, and FIG. 4 is an explanatory diagram of the prior art. In the figure, 1 is a frame-type knowledge representation, 2 is a similarity calculation unit, and 3
represents a grouping processing section.

Claims (1)

[Claims] In an organization/integration method that organizes/integrates frame-based knowledge representations, a frame-based knowledge representation (one ), a similarity calculation unit (2) that calculates similarity in association with the slots and slot values of multiple frames, or the matching number of slots; a grouping processing unit (3) that creates a higher-rank frame having a common attribute and deletes the common attribute from the original frame when the degree exceeds a predetermined threshold; A frame-type knowledge expression organization/integration method characterized by being configured to organize/integrate.