JPS6383837A - Predicate control system - Google Patents

Abstract

PURPOSE:To contrive to improve the predicate processing performance by con trolling intensively the predicate expression knowledges related to the same predicate in a single control unit. CONSTITUTION:The knowledges expressed in predicates are stored in a control unit 14 of an object frame, etc., that is set in response to the predicate equal to that shown in a name expressed in a predicate. Then an object frame control table 13 decides for control a specific predicate to which the control unit set for each predicate corresponds. Therefore the knowledges expressed in plural predicates and having the same predicate are controlled intensively in a single control unit. Thus it is possible to narrow a predicate range where the unification is carried out and therefore to improve the predicate processing performance. Furthermore it is possible to refer to the fragmentary knowledge expressed in a predicate in the form of a group of predicates.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a management technique for predicate notation in knowledge expression expressed using predicate notation. The present invention relates to a management method that aggregates and manages predicate notations related to each predicate expressed by a notation.

[Conventional technology]

Conventionally, when predicate notation is used to express knowledge in an artificial intelligence system, regardless of the format of the file contents, such as text format or program format, they are physically managed within the knowledge storage file. However, the management format on the knowledge storage file does not take into account whether the predicate names in the predicate expressions have the same name or not, and the expression content is managed as written according to the defined order. It was becoming like this. This point will be explained below based on the drawings.

FIG. 7 shows a general format example of predicate representation of target knowledge managed by the predicate management method. General format example 8 of predicate representation consists of a predicate name indicating a predicate and a plurality of arguments ( 1st argument, 2nd argument, . . .)) For example, if the knowledge that ``Abe is good at skiing'' is expressed as a predicate, it becomes ``I am good at (Abe is good at skiing)''.

To explain the conventional predicate management method using specific predicate notation based on the general format example above, for example, "I'm good at it (Tabe, skiing)", "It's my hobby (Tabe, reading)" , "My hobby (Jibe Travel)", "My father (Abe Part)", "My father (Abe Tsujibe)", "I'm good at (Aabe Cooking)", "I'm good at... When knowledge with a predicate notation such as ``(next part: swimming)'' is defined (registered) in that order, conventionally, the management form of each knowledge in the knowledge storage file is shown in Figure 8, which is a simulated diagram. As such, each piece of knowledge is stored in, for example, consecutive areas of the knowledge storage file 9 in the defined order. Therefore, even if the predicate expressions have the same predicate name (such as "I'm good at it"), they are stored scattered in partial parts of the knowledge storage file 9.

[Problem that the invention seeks to solve]

In the conventional predicate management method described above, as shown in Figure 8, multiple predicate expressions are managed in the order of definition, so when unification is executed, even if the predicate is to be processed by a known predicate expression, the As shown in Figure 8, it is necessary to unify the knowledge based on the managed predicate notation for all predicate notation descriptions in order from the beginning. Therefore, when dealing with a huge amount of descriptions such as knowledge expressions, the performance of predicate processing is greatly affected.
The disadvantage is that performance deterioration is inevitable.

The present invention solves these conventional drawbacks, and its purpose is to improve the performance of predicate processing by narrowing down the range of predicates that require unification in predicate processing for knowledge expressed in predicate notation. It's about improving.

[Means for solving problems]

In order to achieve the above object, the present invention provides a predicate management method in which knowledge expressed in predicate expression is stored and managed in a knowledge storage means such as a knowledge storage file. The data is stored in a management unit such as an object frame, which is provided corresponding to the same predicate as the predicate, and the management unit, such as an object frame management table, stores what kind of predicate the management unit provided for each predicate corresponds to. It is configured to be managed using a table.

[Effect]

Knowledge based on predicate notation to be registered in the knowledge storage means is registered in the management unit corresponding to the same predicate as the predicate expressed by the predicate name it has, so knowledge of multiple predicate notations having the same predicate are collectively managed in the same management unit.

〔Example〕

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention, showing an embodiment in which a frame representation is used as a management unit in a knowledge storage file that stores a plurality of predicate representation knowledge storage units, and P is a predicate management unit. , 5 is a knowledge input support means, 6 is an input/output device such as a console having a keyboard or CRT, and 7 is a knowledge storage file.

Referring to the figure, the predicate management unit P of the present embodiment includes a predicate notation knowledge predicate identification unit 1 to which knowledge based on predicate notation input from the input/output device 6 via the knowledge input support unit 5 is provided, and a predicate notation knowledge predicate identification unit 1. It is composed of knowledge registration means 2, predicate-by-predicate notation knowledge management means 3, and predicate notation knowledge predicate management means 4. The knowledge storage file 7 also includes a plurality of object frames 14. ~147 and each object frame 14. ~14. An object frame management table 13 is stored that manages which predicates the object frame corresponds to.

FIG. 2 shows an object frame management table 13° and an object frame 14. ~14. ．． It shows an example of the content at a certain point in time and the relationship between the two. The object frame management table I3 has a plurality of slots 13. ~13. 1, and each slot stores a predicate name and a pointer value to an object frame. Also, the object frame 14. ~147 is also a plurality of slov) 14+t~14.
,, 14□, ~14z-,,1431~143-,
. . , and in each slot, in the example shown in the figure, the object name (fat part, ski, etc.) of the first argument and the second argument is stored.

In the initial state where no predicate notation knowledge is stored in the knowledge storage file 7, no predicate name, pointer value, or target name is stored in the object frame management table 13 and the object frames 141 to 14a.

FIG. 3 is a flowchart for explaining the operation of FIG.
1 to S5 indicate each step. Hereinafter, the operation of this embodiment will be explained with reference to each figure.

In FIG. 1, when the operator displays the knowledge of the predicate notation he wants to register on the CRT display by operating the keyboard of the input/output device 6, and then issues a registration instruction, the knowledge of the predicate notation is transferred to the knowledge input support. captured by means 5;
The predicate notation knowledge is passed to the predicate identification means 1.

When the predicate notation knowledge predicate identification means 1 receives knowledge of a predicate notation, as shown in step S1 in FIG. Identify predicates. Now, suppose that each predicate expression in the predicate expression group 10 in FIG. Therefore, we identify “I am good at” as a predicate. Then, the predicate notation knowledge predicate identification means 1 determines whether the identified predicate has already been recognized, as shown in step S2 in FIG.
The determination is made based on whether "I am good at" is registered as a predicate name in any slot of the object frame management table 13 of the knowledge storage file 7. As mentioned above, since no predicate name is registered in the object frame management table 13 in the initial state,
The determination in step S2 is No, and the predicate notation knowledge predicate identification means 1 sends the input knowledge of predicate notation to the predicate notation knowledge management means 3 for each predicate.

When the predicate-by-predicate notation knowledge management means 3 receives the above-mentioned knowledge of predicate notation, as shown in step S3 in FIG. 3, it assigns a different identifier to each predicate to the predicate name in the predicate notation, and One object frame that has a pointer value ((001) in the example in Figure 2) corresponding to this assigned identifier and has "I am good at" as a predicate name (
In FIG. 2, an object frame 141) is generated, and then the predicate notation knowledge predicate management means 4 is activated.

When the predicate notation knowledge predicate management means 4 is activated, as shown in step S4 in FIG. , registers "I am good at" as the predicate name, registers the value (001) corresponding to the identifier issued to object frame 141 as its pointer value, and inserts slot 13+ and object frame 14. Link with.

Completion of the above processing by the predicate notation knowledge predicate management means 4 is as follows:
This is notified to the predicate notation knowledge management means 3 for each predicate, and upon receiving this notification, the knowledge management means 3 for predicate notation for each predicate registers the knowledge of the predicate notation to be registered (in this example, "I am good at it (Tabe Ski)". The information is passed to the predicate notation knowledge registration means 2 to request registration.

When the predicate notation knowledge registration means 2 receives this request, it checks the predicate name of each slot in the object frame management table 13 in the knowledge storage file 7, as shown in step S5 in FIG. Slot 1 where is registered
3. is found, the empty slot (second
In the example shown in the figure, by registering "thick part" in the first argument field of slot 14□) and "skiing" in the second argument field, the "skiing" input from the input/output device 6 is inputted from the input/output device 6. (
This means that the knowledge of the predicate notation "Tabeski)" has been registered in the knowledge storage file 7.

According to FIG. 4, the knowledge of the predicate notation to be input next is "It's a hobby (Tabe Reading)", and since the predicate name "It's a hobby" is not in the object frame management table 13, Through the same process as described above, a new object frame 142 with (002) as the pointer value and "hobby" as the predicate name is generated, and is stored in slot 13□ of the object frame management table 13 with "hobby" as the predicate name. " is in the pointer column (002
) is registered to link the slot 13□ and the object frame 148, and the object names "fat section" and "reading" are registered in the slot 14t1 of the object frame 14□.

Next, when the knowledge of the predicate notation "My hobby (next trip)" is input from the input/output device 6, since the predicate name "My hobby" is already registered in the object frame management table 13, Predicate notation knowledge predicate identification means 1
Then, it is determined that the predicate name has been recognized (S2>), and the knowledge of the input predicate notation is passed to the predicate notation knowledge registration means 2. When the predicate notation knowledge registration means 2 receives this, it checks each slot in the object frame management table 13 of the knowledge storage file 7, and selects slots 13 .
According to the pointer value (002) of object frame 14. empty slot (slot 14 in the case of Figure 2)
゜), register the target names "Next time" and "Travel".

From now on, as shown in Figure 4, "I am the father (part of the fat part)"
, “I am a father (Abe next time)”, “I am good at it (Aabe next time)”
By sequentially inputting the knowledge of the predicate notation "cooking)" and "I'm good at (next swimming)" from the input/output device 6,
As shown in FIG. 2, they are stored in object frames having the same predicate name. The frame 11 in FIG. 4 is a simulation of the management form of knowledge stored in the knowledge storage file 7 when each predicate notation knowledge in the predicate notation group 10 in the same figure is registered in order from the highest level. It is expressed,
As illustrated by the broken line in the figure, knowledge of predicate expressions having the same predicate is managed in the same management unit.

FIG. 5 is a block diagram showing an example of a predicate processing device using the predicate management method of the above embodiment, in which a processing device 51,
An input/output device 52 connected to the processing device 51 via an ink face 54 and an interface 5 connected to the processing device 51
and an external storage device 53 connected via 5. The processing device 51 includes an input means 510 corresponding to the knowledge input support means 5 in FIG. The external storage device 53 contains an object frame management table 5 similar to that shown in FIG.
30 and a plurality of object frames 531 . ~531
A knowledge storage file 532 having ``n'' is stored. In this knowledge storage file 532, necessary predicate notation knowledge is stored in advance by the predicate management means 511 through the process described above.

The predicate processing means 513 receives information from the input/output device 52 from the ink face 54 . The processing target input via the input/output means 512 is processed according to the knowledge in the knowledge storage file 532 in the external storage device 53, and the processing results are sent to the input/output means 512.
It is output to the input/output device 52 via the ink face 54. The predicate processing means 513 executes unification in its processing process, but in the present invention, predicate notation knowledge related to the same predicate is stored collectively in the same management unit, so for example, as shown in FIG. Processing efficiency can be improved by unification processing. That is, when knowledge of the predicate notation to be processed (for example, a predicate notation in which the first argument is a variable) is input from the input/output device 52 and received by the predicate processing means 513, first, the sixth
As shown in the figure, the predicate expressed by the predicate name to be processed is identified (310), and then the knowledge storage file 532
Obtain the pointer of the object frame that stores the knowledge related to the predicate from the object frame management table 530 in the object frame to specify the target object frame (5ll), and within the specified object frame,
For example, a slot having a value matching the argument to be processed is searched for (S12).

〔Effect of the invention〕

As explained above, the present invention focuses on predicates in knowledge expressions expressed by predicate names in predicate notation, and manages predicate notation knowledge related to the same predicate in the same management unit. It has the following effects.

(1) The performance of predicate processing can be improved by narrowing down the range of predicates for which unification is performed in predicate processing for knowledge expressed in predicate notation.

(2) Fragmentary knowledge based on predicate notation can be referenced as a predicate-related group.

(3) Maintenance work for knowledge regarding the same predicate becomes easier and work efficiency improves.

(4) Access efficiency to knowledge regarding the same predicate is improved.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, and FIG. 2 shows an object frame management table 13 and an object frame 14. -1411, FIG. 3 is a flowchart for explaining the operation of FIG. 1, FIG. 4 is a flowchart for explaining the operation of FIG. 1, FIG. 5 is a block diagram of an application example of the present invention, and FIG. FIG. 5 is a flowchart for explaining the operation, FIG. 7 is a diagram showing an example of a general format of predicate expression, and FIG. 8 is an explanatory diagram of a conventional predicate management system. In the figure, P... predicate management section, 1... predicate notation knowledge predicate identification means, 2... predicate notation knowledge registration means, 3.
... Predicate notation knowledge management means for each predicate, 4... Predicate notation knowledge predicate management means, 5... Knowledge input support means, 6...
Input/output device, 7...Knowledge storage file, 13...Object frame management table, 141-14R...
・Object frame.

Claims (1)

[Claims] In a predicate management method in which knowledge expressed in predicate expression is stored and managed in a knowledge storage means, knowledge expressed in predicate expression is stored in a predicate expression corresponding to the same predicate as the predicate name expressed by the predicate name in the predicate expression. 1. A predicate management method, characterized in that the predicates are stored in a management unit provided for each predicate, and a management table manages what kind of predicate the management unit provided for each predicate corresponds to.