JP2000155682A - Analogizing device, its method and computer readable recording medium recording analogical program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an analogizing device, etc., which has high accuracy, utilizes the knowledge of a wide world that can not intentionally be grasped by a human and also rationally fetches such knowledge. SOLUTION: A rule group managing part 1 manages a rule group in a problem area being an analogical object. A fact group managing part 2 manages a known fact group in the problem area. An analogically possible body group constructing part 3 refers to the parts 1 and 2 and an analogical result group managing part 6 and constructs an analogically possible body group such as an analogy expression. An analogically possible body group managing part 4 manages the analogically possible body group constructed by the part 3. An analogizing part 5 performs analogical processing based on the analogically possible body group managed by the part 4. The analogical result group managing part 6 manages an analogical result group obtained through the analogical processing by the part 5. The part 3 successively improves an analogically possible body group while reflecting the evaluation results of the analogical result group managed by the part 6 in the process of repeatedly performing analogical processing by the part 5.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to analogy processing using a computer system, and more particularly to analogy processing based on a group of rules and facts in a problem area to be analogized and a group of analogy results obtained by analogy processing. TECHNICAL FIELD The present invention relates to an analogy device that forms an analogous body group used in a computer, a method thereof, and a computer-readable recording medium that stores an analogy program.

[0002]

2. Description of the Related Art Many analogy processes using a computer system have been proposed so far.
Japanese Patent Publication No. 251467 discloses an analogy process using a data mining technique. Here, the data mining system described in Japanese Patent Application Laid-Open No. 9-251467 has a rule base for managing rules assumed in advance, a receipt database for managing receipt data as purchase data, a rule base and a receipt database. And data mining means for generating rules used in analogy processing based on the rules and receipt data managed by. In this data mining system, first, analogy targets (products and their attributes, etc.) are extracted from rules managed in a rule base, and a multidimensional database is created using a receipt database based on the extracted contents. Generate. Then, a new rule candidate is extracted by the data mining means using the multidimensional database generated in this way, and an important rule is selected from the extracted rule candidate to be used in analogy processing. Rules that are being generated.

[0003]

However, in the data mining system described in the above-mentioned Japanese Patent Application Laid-Open No. 9-251467, the application is limited to the extraction of simultaneous purchase rules using basket analysis. Is limited to rules that say, "If product A is sold, product B is also sold." Therefore, it is not possible to perform highly-accurate analogy processing such as "which product can be sold when and how many." There is.

[0004] The rules used in the analogy process include data of the product itself stored in the product database,
Since it is generated based on the sales record of the product stored in the receipt database, even if it is generally known that the weather greatly affects the sales of the product, it seems that such a weather was involved It is not possible to derive a rule that reflects a very large world, and there is a problem that the world behind the rule is very narrow. Here, the motivation for performing analogy processing using a computer system other than human beings originally has the expectation of discovering hidden rules that humans cannot consciously grasp. Simply performing the analogy processing based on the rule that does not meet the original request.

[0005] The simplest method for solving such a problem is to widen the types of facts used in analogy processing. However, it is unrealistic to create a database of all event elements that are regarded as facts. It is a target.

[0006] The present invention has been made in view of the above points, and utilizes knowledge of a wide world that is highly accurate and cannot be consciously perceived by humans. It is an object of the present invention to provide a computer-readable recording medium that records an analogy apparatus, a method thereof, and an analogy program that can be captured.

[0007]

A first feature of the present invention is as follows.
A rule group management unit that manages a rule group in the problem area to be analogized, a fact group management unit that manages a known fact group in the problem area, a rule group managed by the rule group management unit, On the basis of the fact group managed by the fact group management unit, an analogous body group constituting unit constituting the analogous body group used for the analogy processing, and an analogous body constituted by the analogous body group configuring unit An analogy-possible body group management unit that manages a group; an analogy unit that performs analogy processing based on the analogy-possible body group managed by the analogous body group management unit; and an analogy obtained by the analogy processing by the analogy unit. An analogy result group management unit that manages a result group, wherein the analogizable body group constituent unit reflects the evaluation result of the analogous result group managed by the analogous result group management unit and reflects the analogous body group. Form, the analogy can body group construction section,
An analogization apparatus characterized by sequentially improving the analogizable group in the course of repeatedly performing analogy processing by the analogizable body group management unit, the analogization unit, and the analogy group result group management unit.

In the first feature of the present invention described above, it is preferable that the analogous body group forming unit configures the analogous body group in a non-deterministic improvement process. Also,
It is preferable that the analogous body group constituting unit appropriately selects and refers to a part of the rule group managed by the rule group managing unit in the process of improving the analogous body group. further,
It is preferable that the analogous body group constituting unit appropriately selects and refers to a part of the fact group managed by the fact group management unit in the process of improving the analogous body group.

In the above-mentioned first feature of the present invention, the analogizable body group forming unit includes a rule group managed by the rule group managing unit and a fact group managed by the fact group managing unit. A variable group extraction unit that extracts a variable group in the problem area based on a group, an analogy item group management unit that manages analogy item groups to be analogized in the problem area, and a variable group extraction unit that is managed by the analogy item group management unit. The analogy formula group used to configure the analogy formula group used for analogy processing based on the analogy item group as the analogous body group, and the analogy result group managed by the analogy result group manager are managed by the fact group. An evaluation unit that evaluates based on the fact group managed by the unit, and an evaluation result group management unit that manages an evaluation result group evaluated by the evaluation unit. Department More analogy treated with repeated process, and it is preferable to successively improve the analogy expression group while reflecting the evaluation result group managed by the evaluation result group management unit.

Further, in the first aspect of the present invention described above, the fact group management unit includes external data access means for accessing an external data source and adding a fact group included in the external data source. Is preferred. Further, it is preferable that the external data access means learns an external data source to be accessed based on the usage status of the added fact group.

A second feature of the present invention is that a rule group management unit that manages a rule group in a problem area to be analogized,
A fact group management unit that manages a known group of facts in the problem area, a rule group managed by the rule group management unit, and an analogy process based on the fact group managed by the fact group management unit. An analogy part to perform,
An analogy apparatus, wherein the fact group management unit includes an external data access unit that accesses an external data source and adds a fact group included in the external data source.

[0012] In the second aspect of the present invention, it is preferable that the external data access means learns an external data source to be accessed based on the use status of the added fact group.

A third feature of the present invention is that a step of managing a rule group in a problem area to be analogized, a step of managing a known fact group in the problem area, a managed rule group and a fact group Configuring an analogous body group used in the analogy processing based on, and managing the configured analogous body group,
A step of performing an analogy process based on the managed analogable body group, and a step of managing the analogy result group obtained by the analogy process, wherein the step of configuring the analogous body group includes performing the analogy process. An analogization method characterized by sequentially improving the analogizable body group while reflecting the evaluation result of the managed analogy result group in the process of repeating.

A fourth feature of the present invention is a procedure for managing a rule group in a problem area to be analogized, a procedure for managing a known fact group in the problem area, a managed rule group and a fact group. A procedure for configuring a group of analogyable bodies used for analogy processing based on the above, a procedure for managing the configured group of analogyable bodies, a procedure for performing analogy processing based on the managed group of analogyable bodies, And a step of managing the analogy result group obtained by the processing.The procedure of configuring the analogous body group reflects the evaluation result of the managed analogy result group in the process of repeatedly performing the analogy processing. A computer-readable recording medium storing an analogy program, wherein the analogous group is sequentially improved.

According to the first, third and fourth features of the present invention, the analogy result group is sequentially improved while the analogy result group is evaluated based on the fact group. A possible body group can be obtained, and therefore, analogy processing can be performed with high accuracy in various problem areas.

Further, according to the first, third and fourth features of the present invention, by making the process of improving the analogous body group non-deterministic, the subtle behavior in the process of improving the analogous body group can be reduced. Can be modified so that it is possible to reflect knowledge that is more complicated than humans can grasp in analogical expressions, and therefore, it is possible to use knowledge of the wide world that humans cannot consciously grasp. Analogy can be performed with higher accuracy.

Further, according to the first, third and fourth features of the present invention, the external data access means for accessing the external data source by the fact group management unit and adding the fact group included in the external data source is provided. By having
Analogy processing can be performed with higher accuracy by reasonably taking in knowledge of the wide world.

On the other hand, according to the second feature of the present invention, the fact group management unit has the external data access means for accessing the external data source and adding the fact group included in the external data source, so that a wide world Can be reasonably taken in to perform analogy processing with higher accuracy.

[0019]

Embodiments of the present invention will be described below with reference to the drawings. 1 and 2 are views showing an embodiment of an analogization device according to the present invention. here,
FIG. 1 is a block diagram showing the overall configuration of the analogization device, and FIG. 2 is a block diagram showing details of the analogization device shown in FIG.

First, the overall configuration of the analogization device will be described with reference to FIG. As shown in FIG. 1, the analogization apparatus includes a rule group management unit 1 that manages a rule group in a problem area to be analogized, a fact group management unit 2 that manages a known fact group in the problem area, and a rule group management. An analogous body group forming unit that forms an analogous body group such as an analogous expression used in analogy processing based on the rule group managed by the unit 1 and the fact group managed by the fact group managing unit 2. 3, an analogizable body group management unit 4 that manages the analogizable body group configured by the analogizable body group forming unit 3, and an analogy based on the analogizable body group managed by the analogizable body group management unit 4. An analogy result group management unit 6 that manages analogy result groups obtained by analogy processing performed by the analogy unit 5; Evaluation results of managed analogy result groups And the analogy-possible body group is formed by repeating the analogy process by the analogy-possible body group forming unit 3, the analogy-possible body group management unit 4, the analogization unit 5, and the analogy result group management unit 6. Are sequentially improved.

The rule group management unit 1 manages one or more rule groups considered to be satisfied in the problem area to be analogized in a storage device such as a hard disk. The rules here include "the product A sells when the temperature rises" and "the product B stops selling when the word" yen depreciation "increases on a certain homepage." Note that the contents of the rule need only be assumed to be satisfied, and need not necessarily be true. In addition, the expression of the rule is, in addition to the plain text as described above, a mathematical expression,
Any expression such as an expression suitable for symbol processing may be used. Further, the method of storing the rules may be any method using a database or the like, other than the method using a simple file. It should be noted that the rule group management unit 1 manages many rule groups in advance, and selects a part of the rule group managed by the rule group management unit 1 as needed in the process of improving the analogous expression group described later. It is good to be referred to.

The fact group management unit 2 manages one or more fact groups, which are known data in a problem area to be analogized, in a storage device such as a hard disk. The fact referred to here is “E1 (time transition of sales of the product A): (t, f) = ｛(t0, f0), (t1, f1), (t2, f2),.
In this case, the direction t of the fact E1 is “time”, and the content f at the time t of the fact E1 is “sales of the product A”. The expression of the fact may be any expression such as an expression using a table, in addition to the enumerated expression described above. Also, the method of storing facts may be any method using a database or the like, other than the method using a simple file. The fact group management unit 2 manages a large number of fact groups in advance, and in the process of improving the analogous expression group described later,
A part of the fact group managed by the fact group management unit 2 may be appropriately selected and referred to.

The analogizable body group forming unit 3 is configured to be able to perform analogy based on the rule group managed by the rule group managing unit 1 and the fact group managed by the fact group managing unit 2. It constitutes a body group and will be described later in detail.

The analogizable body group management section 4 manages the analogizable body group constituted by the analogizable body group forming section 3 in a storage device such as a hard disk. The method of storing the analogyable body may be any method using a database or the like, other than the method using a simple file.

The analogization unit 5 performs analogy processing based on the analogous group managed by the analogiable group management unit 4. Specifically, the content of calculation such as an analogous expression is adjusted to the expression. And interpret and execute arithmetic processing according to the expression.

The analogy result group management unit 6 manages one or more analogy result groups estimated by analogy processing by the analogy unit 5 in a storage device such as a hard disk. The analogy result referred to here is “E1 (time transition of sales of the product A): (t, e) = ｛(t0, e0), (t1, e1), (t2, e2),..., (Tm, e
m)｝ ”. In this case, the analogy result E1
Is the time t, and the content e of the analogy result E1 at the time t is "sales of the product A". If the calculation contents of the analogy formulas and the like that are analogizable bodies are definitive, the same analogy result is always obtained when the same analogy condition is given to the same analogy formula. In addition, the expression of the analogy result may be any expression such as an expression using a table, in addition to the enumerated expression as described above. Further, the method of storing the analogy result may be any method using a database or the like, other than the method using a simple file.

Next, referring to FIG. 2, the details of the above-described analogizable body group forming unit 3 will be described. As shown in FIG.
The analogizable body group forming unit 3 extracts a variable group in a problem area based on the rule group managed by the rule group managing unit 1 and the fact group managed by the fact group managing unit 2. Unit 31, an analogy item group management unit 32 for managing analogy item groups based on the variable groups extracted by the variable group extraction unit 31, and an analogy item group based on the analogy item groups managed by the analogy item group management unit 32. Analogy expression group forming unit 33 that configures analogy expression groups used for processing as analogizable body groups
And an evaluation unit 34 that evaluates the analogy result group managed by the analogy result group management unit 6 based on the fact group managed by the fact group management unit 2, and an evaluation result group evaluated by the evaluation unit 34. An evaluation result group management unit 35 that manages the evaluation result group. The analogy expression group configuration unit 33 reflects the evaluation result group managed by the evaluation result group management unit 35 in the process of repeatedly performing the analogy processing by the analogization unit 5. While forming an analogy formula group, the analogy formula group forming unit 33, the analogizable body group management unit 4, the analogy unit 5,
The analogy result group management unit 6, the evaluation unit 34, and the evaluation result group management unit 35 sequentially improve the analogy expression group.

The variable group extraction unit 31, based on the rule group managed by the rule group management unit 1 and the fact group managed by the fact group management unit 2, sets the background of the rule group and the fact group. This is to extract one or more variable groups considered to exist in the problem area. The variable referred to here is the above-mentioned “When the temperature rises, the product A
For example, the rule that "sells"
And "sales of product A". In this example, a very simple extraction is performed. Alternatively, for example, “the existence of a rule” may be extracted as a variable. In this case, if one rule exists, one variable is extracted correspondingly. Note that the method of extracting variables is determined by the extraction strategy knowledge held in advance inside the variable group extraction unit 31. In addition to simply extracting variables, additional information can be given at the same time as extracting variables. As an example, the presence / absence of a fact corresponding to the extracted variable is checked with reference to the fact group management unit 2, and the result is given as additional information. In addition, additional information such as “extracted from the same rule” can be given to a plurality of variables extracted from one rule. Furthermore, if a significant dependency between variables can be recognized, the dependency is expressed in a form, for example, "It is the increasing tendency of the variable Z that controls the increasing tendency of itself (which is a variable)." It can also be given as additional information.

The analogy item group management unit 32 manages one or more analogy item groups serving as containers for a direct solution and a secondary solution in a problem area to be analogized in a storage device such as a hard disk. It is. The analogy item here is “sales of the product A (in the future)”. Note that the direction of analogy may be not only in the future but also in the past, may be not only time but also space, and may be determined depending on the product series or the like, or may be determined depending on the combination of these. The method of inputting the analogy item may be given based on the variable group extracted by the variable group extracting unit 31 in addition to the external direct instruction. When the variable is assigned based on the variable group extracted by the variable group extracting unit 31, the variable is managed after being replaced with a concept of analogy item. The method of storing analogy items may be any method using a database or the like, other than the method using a simple file.

The analogy formula group forming unit 33 is used for analogy processing based on the analogy item group managed by the analogy item group management unit 32 and the evaluation result group managed by the evaluation result group management unit 35. This constitutes one or more analogous expression groups. The analogy formula here is

(Equation 1) And so on. Note that the calculation content of the analogy formula must be deterministic, that is, the same analogy result can be obtained when the same analogy condition is given to the same analogy formula. However, the calculation content of the analogy equation must be deterministic, but the process of constructing the analogy equation may be non-deterministic. That is, even when the same analogy item group and the same evaluation result group are given, it is not always necessary to configure the same analogous equation group. The expression of the analogy expression may be any expression such as a procedural expression in addition to the mathematical expression described above.

The evaluation unit 34 evaluates the analogy result group managed by the analogy result group management unit 6 on the basis of the fact group managed by the fact group management unit 2. Specifically, for example, The sum of the absolute value of the difference between the analogy result corresponding to the analogy item and the fact corresponding to the analogy item is obtained, and the degree of coincidence between the analogy result and the fact is evaluated based on the result. In addition, the evaluation method is not limited to the above-described method, and may be any method that is significant.

The evaluation result group management unit 35 manages the evaluation result group evaluated by the evaluation unit 34 in a storage device such as a hard disk. The evaluation result here is, in the above-described evaluation method, “the degree of coincidence between the analogy result E1 obtained by the analogy equation E1 and the fact E1 is 0.7”.
And so on. Such an evaluation result is associated with the analogy expression used in the analogy processing, and the analogy expression group forming unit 3
In the process of repeating the analogy process by the analogy formula group constituent group 33, the analogizable body group management unit 4, the analogization unit 5, the analogy result group management unit 6, the evaluation unit 34, and the evaluation result group management unit 35, It is used to construct an analogous expression with a higher degree of coincidence. The expression of the evaluation result may be any expression such as an expression using a table, in addition to the plain text as described above. Also, the storage method of the evaluation result may be any method using a database or the like, other than the method using a simple file.

Next, the operation of the present embodiment having such a configuration will be described by taking a specific problem area (analysis of sales at a convenience store) as an example.

First, an outline of the convenience store business will be described. Convenience stores are usually affiliated with a particular convenience store chain. The minimum unit constituting such a convenience store chain is a “store”, and a “store code” for identifying each store is assigned to each store.

Various merchandise / services are sold at each store, and a “merchandise code” is issued for each type of merchandise. Also, services are considered in the same category as products, and product codes are issued. Note that the product code here is the same for all the same products A, and is not a serial number for identifying a product individual. Also, the product code is assumed to be common throughout the convenience store chain.

Here, all sales of merchandise are made by POS (Point
It is performed under the management of the (Of Sales) system, and when a product is sold, the product code, the number, and the sales time of the sold product are recorded. At this time, sales to the same customer are managed as one collective sale. That is, a “sales number” is issued for each sale to a customer, and the product code / quantity / sales time is managed under this sales number. This sales number becomes a unique number throughout the convenience store chain by adding the store code to itself.

Each store orders merchandise to the convenience store chain headquarters (managed as a type of store and has a store code) according to the respective circumstances.
With a single order, multiple types of products can be ordered. The content of the order is determined by the store owner of each store. An “order number” is issued and managed for each order. This order number can be obtained by adding a store code to itself.
It is a unique number throughout the convenience store chain.

It is assumed that delivery for an order is performed, for example, three times a day every eight hours. The delivery time is determined for each store, and delivery for a certain order is made at the next delivery time after the closest delivery time from the time of ordering. For example, if a store whose delivery time is determined at 2: 00/10/18 is ordered at 1:00, the delivery for the order is performed at 10:00. In this case, it is assumed that the ordered product is always delivered, and no shortage occurs.

FIG. 3 is an ER diagram showing an outline of the above convenience store business as an ER (Entity Relation) model. Here, the ER diagram shown in FIG.
It is described according to the notation EF1X (I
Regarding DEF1X, reference 1 (Thomas A Bruce: “Des
igning Quality Databases with IDEF1X InfomationMod
els, "DOREST HOUSE PUBLISHING, 1992). Since the actual code system used in such a convenience store business is not an essential problem, it will not be described further here.

In such a convenience store business, it is very important to prevent the occurrence of out-of-stock items while minimizing store inventory. Therefore, "what / how / when, Is the most important issue? In order to solve this problem, it is necessary to analogize the accurate sales.Therefore, the problem area to be analogized for such a convenience store business is, "Which products are available at each convenience store? / How many / when to sell. "

The operation of the analog inference apparatus shown in FIGS. 1 and 2 will be described below, taking the problem area set as described above as an example. Here, a flowchart for explaining the operation of the analogization device shown in FIGS. 1 and 2 is shown in FIG.

First, the rule group management unit 1 manages sale of goods in a convenience store and one or more rule groups that are generally considered to be established (step A1). Note that the rules described here need not always be true. Here, for example, it is assumed that three rules as shown in FIG. 5 are managed. These rule groups can be managed using, for example, a relational database. FIG. 6 is an ER diagram showing a schema for managing these rule groups. FIG. 7 is a diagram showing an image in which these rule groups are stored in a table of a relational database. As shown in FIGS. 6 and 7, the rule group management unit 1 can manage the rule group by a general technique that is not special, so that the rule group management unit 1 can be easily mounted. There are various rules input methods,
As an example, FIG. 8 shows a method of inputting a rule from a window displayed on a screen. In addition to the above, various input methods such as input from a file can be considered. However, since the input method of the rule is not an essential problem, no further description will be given here. Whichever input method is adopted, the input function can be realized by a general technique that is not special. Further, such an input function can be implemented as a part of the function of the rule group management unit 1.

Next, the fact group management section 2 manages a fact group existing in the problem area to be analogized (step A2). Here, for example, it is assumed that a fact group as shown in FIGS. 9 to 12 is managed. These fact groups can be managed using, for example, a relational database. FIGS. 13A and 13B
(C) and (d) are ER diagrams showing schemas for managing these fact groups. FIGS. 14 to 17 show images in which these fact groups are stored in a table of a relational database. As shown in FIGS. 13 to 17, in the fact group management unit 2,
Since the fact group can be managed by a general technique that is not special, the fact group management unit 2 can be easily mounted. There are various methods for inputting facts. As an example, an input method from a file is shown in FIGS. In addition to the above, various input methods such as input from a window displayed on the screen are conceivable. However, the fact input method itself is not an essential problem, and therefore will not be described further here. Whichever input method is adopted, the input function can be realized by a general technique that is not special. Further, such an input function can be implemented as a part of the function of the fact group management unit 2.

Next, the variable group extracting unit 31 of the analogizable body group forming unit 3 uses the rule group managed by the rule group managing unit 1 and the fact group managed by the fact group managing unit 2. Then, a variable group in the problem area is extracted (step A3). In the example shown so far, “game”, “game magazine”, “potato chips”, and “circle” are extracted from the rule group, and “game”, “game magazine”, “potato chips”, and “circle” are extracted from the fact group. Is done. Various methods for extracting the variables can be considered according to the method of expressing the rule group and the fact group and the content of the specific rule and the fact. In the examples shown so far, since the rule group is expressed as plain text, it is possible to use a method of extracting nouns from the plain text by using a generally used phrase analysis method. Here, whether or not a noun is determined can be determined by collating with a dictionary. Note that the dictionary may be held inside the variable group extraction unit 31, or may be an external dictionary such as a so-called electronic dictionary referred to via an interface or the like. On the other hand, in the examples described so far, the fact group is represented using a table, and therefore, for example, a method of extracting a variable name from a table name can be used. Thus, the variable group extraction unit 3
In 1, the variable group can be extracted based on the rule group and the fact group by a general technique that is not special, so that the variable group extracting unit 31 can be easily mounted.

Various methods for extracting variables can be considered in addition to a method for extracting nouns from plain text using a lexical analysis technique. The extraction strategy knowledge on the variable extraction method is held by a program code or the like mounted inside the variable group extraction unit 31. In addition to extracting variables, additional information can be given at the same time as extracting variables. Here, in order to obtain additional information, the fact group management unit 2 is referred to again. The reason why the fact group management unit 2 is referred to is to check whether or not a fact corresponding to the extracted variable exists. In addition, a method of interpreting the context of the rule using a generally used syntax analysis method can be used for extracting additional information other than the fact information. As described above, the additional information can be extracted by a general technique that is not special, so that the variable group extraction unit 31 can be easily mounted. In the example shown so far, the following is obtained as additional information.・ The variable “game” has “fact” ・ The variable “game magazine” has “fact” ・ The variable “game magazine” depends on “variable“ game ”” ・ The variable “potato chips” has “fact”・ The variable “Potato Chips” depends on the variable “Game”. ・ The variable “Potato Chips” depends on “Friday 19:00”. ・ The variable “Yen” has “Fact”. The variable "yen" is "increased"

Next, the analogy item group management unit 32 of the analogous body group configuration unit 3 manages, as an analogy item group, one or more commodities that need analogy for ordering a convenience store (step A4). . In the examples described so far, “game”, “game magazine”, and “potato chips” are used as analogy items, and sales trends of these items are analogized.
Note that, here, these analogy items are containers for direct solutions in the problem area of analogy of sales transition, that is, analogy items directly related to accurate ordering. To be managed. Further, the variable group extracted by the variable group extraction unit 31 is managed as an analogy item group (step A5). In the example shown so far, the variable "yen" may be added as an analogy item. Since the analogy item “yen” is a container for secondary solutions in the problem area of analogy of sales transition, that is, an analogy item that may be related to accurate ordering but is not explicit, It is managed as additional information together with the distinction as described above. In the examples shown so far, the following items are obtained as analogy items. -Analogy item "game" ... direct analogy item-Analogical item "game magazine" ... direct analogy item-Analogical item "potato chips" ... direct analogy item-Analogical item "yen" ... indirect analogy item

The analogy item group and the variable group described above can be managed using, for example, a relational database. FIG. 22 is an ER diagram showing a schema for managing these analogy item groups and variable groups. FIGS. 23A, 23B, and 23C are diagrams showing images in which these analogy item groups and variable groups are stored in a table of a relational database. As shown in FIGS. 22 and 23, the analogy item group management unit 32 can manage analogy item groups by a general technique that is not special. Therefore, it is easy to implement the analogy item group management unit 32. Can be. Note that there are various methods for inputting analogy items that are containers for direct solutions. FIG. 24 shows a method of inputting analogy items from a window displayed on the screen as an example. In addition, various input methods such as input from a file can be considered. However, since the input method of analogy items is not an essential problem, no further description will be given here. Whichever input method is adopted, the input function can be realized by a general technique that is not special. Such an input function can be implemented as a part of the function of the analogy item group management unit 32. In addition to the above, it is assumed that the analogy item group management unit 32 manages an analogy item “t” representing time as a special analogy item. The analogy item "t" is
In the analogy formula group forming unit 33, it is used in the form of “t”.

Next, the analogy group managed by analogy group management unit 4 and analogy item group management unit 32 manage analogy group managed by analogous group management unit 4 by analogy formula group configuration unit 33 of analogous body group configuration unit 3. One or more analogy expression groups are configured based on the analogy item group and the evaluation result group managed by the evaluation result group management unit 35 (step A6). In the example shown so far, for example, FIG.
A group of analogy formulas is constructed by the method shown in FIG.

FIG. 25 is a flowchart for explaining the details of step A6 shown in FIG. Since the analogy process has not yet been performed at the initial time, the analogizable body group management unit 4 and the evaluation result group management unit 35 are empty. In FIG. 25, first, analogy items are retrieved from the analogy item group management unit 32, and corresponding empty analogy expressions are generated (step B1). In the example shown so far, the following analogous expression group is generated.・ Analog expression “Game-1: φ” ・ Analog expression “Game magazine-1: φ” ・ Analog expression “Game magazine-2: φ” ・ Analog expression “Potato chips-1: φ” ・ Analog expression “Potato chips- 2: φ ”・ Analog expression“ Yen-1: φ ”

Here, “φ” indicates that the content of the analogy expression is empty. "Game-1""GameMagazine-1"
“Game magazine-2”, “Potato chips-1”, “Potato chips-2”, and “Yen-1” are identifiers of analogical expressions. The information after the hyphen attached to the end of the identifier of each analogy expression indicates that, for example, “game-1” is the first analogy expression for the analogy item “game”.
The reason why a plurality of analogy expression groups are made to correspond to one analogy item is that the present embodiment is based on a process of improving the analogy expression group. As one of the processes for improving the analogy equation group, for example, a method of preparing a plurality of analogy equation groups for one analogy item and constructing an analogous equation group having excellent analogy precision by a so-called genetic algorithm approach. (For a genetic algorithm, see Reference 2 (David E. Goldin
g: “Genetic Algorithms in Search, Optimization and
Machine Learning, “ADDISION WESLEY, 1989)). It is not essential how many analogy formulas should be prepared for one analogy item.
No further mention is made here.

Next, the knowledge given as a rule, particularly the relation between variables, is reflected in the analogical expression (step B2).
FIG. 26 is a flowchart for explaining the details of step B2 shown in FIG. In FIG. 26, first,
One of the analogy items managed by the analogy item group management unit 32 is specified (Step C1), and a variable serving as a basis of the analogy item specified in this way is specified (Step C).
2). Next, the additional information of the variable specified in this way is checked, and a variable group related to this variable is specified (step C3). After that, the analogy item group corresponding to the variable group specified in this way is specified (step C4), and any one of the analogy expression groups corresponding to the analogy item group specified in this way is determined. All analogy formula groups corresponding to the analogy item specified in step C1 are made to correspond (step C1).
5). Then, the above-described steps C1 to C5 are repeated for all the analogy items (step C6). In the example shown so far, the variables "game magazine" and "potato chips" depend on the variable "game", so that the following analogical expression group is formed, for example. In the following analogical expression, “int” represents a function for converting to an integer.・ Analog expression “Game-1: φ” ・ Analog expression “Game magazine-1: int (0.9 ד Game-1 ”)
(t-30 * 60))]-Analogical formula "game magazine-2: int (0.8 x" game-1 ")
(t-20 * 60))] ・ Analog expression "Potato chips-1: int (" Game-1 "(t
-30 * 60))] ・ Analog expression "Potato chips-2: int (" Game-1 "(t
-40 * 60))] ・ Analog expression "Yen-1: φ"

In these analogy expressions, the analogy item “game”
Since there is only an analogy expression corresponding to "game-1", the analogy expressions "game magazine-1", "game magazine-2", "potato chips-1", and "potato chips-2" respectively have analog expressions "game-1". Is reflected. This step B2
In, if the additional information has more usable information, further information can be given to the content of the analogy expression. For example, if the additional information “increase in proportion to the decrease rate of the game” is given to the variable “game magazine”, then the analog expression “game magazine-1” is given by “−d
(“Game-1”) / dt ”. Here, dx / dt is obtained by replacing the variable “x” with the variable “t”.
Means differentiating with Thus, it is considered that the analogy formula group corresponding to the analogy item is formed based on the variable group extracted from the rule group and the fact group and the additional information, so that the analogy formula group is established in the problem area to be analogized. Knowledge can be effectively used, and therefore, a group of analogical expressions having excellent analogical precision can be obtained. Also, since the analogy group is improved starting from the rule group given in advance, it is possible to obtain a group of analogy groups that is faster and has higher analogy accuracy than the case of improving a group of analogy groups that are completely random. Can be.

Next, the remaining knowledge given as the rule is reflected in the analogical expression (step B3). FIG. 27 shows FIG.
12 is a flowchart for explaining details of step B3 shown in FIG. In FIG. 27, first, one of the analogy formulas is specified (step D1), and then the analogy item group management unit 32
An arbitrary analogy item group is selected from (step D2). Next, a variable group as a basis of the analogy item group selected in this way is specified (step D3), and additional information of the variable group specified in this way is obtained (step D4). Then, after specifying the additional information other than the relationship between the variables (step D5), the content of the additional information specified in this way is arithmetically processed and reflected on the analogy expression specified in step D1 ( Step D6). Then, the above-described steps D1 to D6 are repeated for all analogical expressions (step D7). In the example shown so far, for example, the following analogous expression group is configured. In the following analogical expression, “if C then R1 else R
"2" represents a function that returns R1 when the condition C is satisfied, and returns R2 when the condition C is not satisfied.・ Analog expression “Game-1: φ” ・ Analog expression “Game magazine-1: int (0.9 ד Game-1 ”)
(t-30 * 60))]-Analogical formula "game magazine-2: int (0.8 x" game-1 ")
(t-20 * 60))] ・ Analog expression "Potato chips-1: int (" Game-1 "(t
-30 * 60) × (if cos (2π (t / 7/24/60 / 60- (24 * 6 + 19) / 24/7))
> = 0.5 then 90 (cos (2π (t / 7/24/60 / 60- (24 * 5 + 19) / 24 /
7)) + 1) else 1) "・ Analog expression" Potato chips-2: int ("Game-1" (t
-40 * 60) × (if cos (2π (t / 7/24/60 / 60- (24 * 6 + 19) / 24/7))
> = 0.5 then 80 (cos (2π (t / 7/24/60 / 60- (24 * 5 + 19) / 24 /
7)) + 1) else 1) "・ Analog expression" Yen-1: 120 + t / 24/60/60 "

In these analogical expressions, "depends on Friday 19:00" which is additional information for the variable "potato chips" which has not been used so far, and "increases" which is additional information for the variable "yen" Are applied in a processed form. "Friday 19:00", which is additional information for the variable "Potato chips", is "t / 7/24/60 / 60-
"(24 * 5 + 19) / 24/7" and "increase" which is additional information for the variable "yen" is expressed as "+ t / 24/60/60". In this way, by giving rules other than the relationship between variables, it is possible to effectively utilize knowledge that is considered to be established in the problem area to be analogized, and therefore, a group of analogical equations having excellent analogical precision. Can be obtained.
Further, similar to the processing in step B2 described above, it is possible to obtain a group of analogy equations with higher speed and higher analogous precision.

Next, knowledge other than that given as a rule is reflected in the analogical expression (step B4). FIG. 28 shows FIG.
13 is a flowchart for explaining details of step B4 shown in FIG. In FIG. 28, first, one of the analogy formulas is specified (step E1), and then the analogy item group management unit 32
An arbitrary analogy item group is selected from (step E2). Next, after an arbitrary analogy expression is selected from the analogy expression group corresponding to the analogy item group selected in this way (step E3), the analogy expression thus selected is arithmetically processed. Then, it is reflected on the analogy expression specified in step E1 (step E4). Then, the above steps E1 to E4 are repeated for all the analogy expressions (step E5). In the example shown so far, for example, the following analogous expression group is configured.・ Analog expression “Game-1: int (if cos (2πt / 365/24/60/6)
0)> = 0 and sin (2πt / 365/24/60/60)> = 0 then 100co
s (2πt / 365/24/60/60) else 0) "・ Analog expression" game magazine-1: int (0.9 × "game-1")
(t-30 * 60) -d ("yen-1") / dt) "・ Analog expression" game magazine-2: int (O.8 x "game-1")
(t-20 * 60) -d ("yen-1") / dt) "· Analogy formula" Potato chips-1: int ("game-1" (t
-30 * 60) × (if cos (2π (t / 7/24/60 / 60- (24 * 6 + 19) / 24/7))
> = 0.5 then 90 (cos (2π (t / 7/24/60 / 60- (24 * 5 + 19) / 24 /
7)) + 1) else 1) "・ Analog expression" Potato chips-2: int ("Game-1" (t
-40 * 60) × (if cos (2π (t / 7/24/60 / 60- (24 * 6 + 19) / 24/7))
> = 0.5 then 80 (cos (2π (t / 7/24/60 / 60- (24 * 5 + 19) / 24 /
7)) + 1) else 1) "・ Analog expression" Yen-1: 120 + t / 24/60/60 "

In the above analogy equation, the analogy equation "game-
"1" is composed of the analogy item "t" which is always managed by the analogy item group management unit 32. In addition, the analogy expression "yen-1" corresponding to the analogy item "yen" which has not been used so far.
Is applied in a form processed as “d (“ yen−1 ”) / dt”. In this way, even if not explicitly given as a rule, by relating analogical expressions to each other or analogical items, knowledge that is not recognized by humans can be reflected in the analogical expression. It is possible to obtain a group of analogy expressions having higher analogy accuracy. In other words, analogy processing can be performed using knowledge of a wide world that cannot be consciously grasped by humans. In step B4 described above, there is no guarantee that a group of analogy equations having higher analogy precision will be obtained in a monotonically increasing manner, and the effect may be short-term. However, it is expected that by forming the analogy equation group in the improvement process, an analogous equation with more excellent vertebral accuracy can be obtained in the long term. Note that the process of improving the analogy formula group may be non-deterministic as described above. Processing strategy knowledge such as what kind of analogy item is selected and what kind of processing is performed is held by a program code or the like mounted inside the analogy formula group forming unit 33.

Here, returning to the flowchart shown in FIG. 4, the analogizable body group management section 4 manages the analogous equation group formed by the analogous equation group forming section 33 of the analogizable body group forming section 3 ( Step A7). Here, the analogy expression group is pe
rl script, and is managed in a file format.
rry Wall and Randal L. Schwartz: “Programming Per
1, "O'REILLY and ASSOCIATES Inc., 1996). Fig. 29 is a diagram showing an image in which the analogy expression" potato chips-1 "is stored in a file. As described above, since the analogy group can be managed by a general technique that is not special, the analogous body group management unit 4 can be easily mounted.

Next, the analogization unit 5 performs analogy processing based on the analogy expression group managed by the analogizable body group management unit 4 (step A8). In the example shown so far, the analogy inference unit 5
Can be configured from a perl script engine. The analogization unit 5 configured in this way reads a perl script managed by the analogizable body group management unit 4 and executes it on the perl script engine to perform analogy processing. Do.

Next, the analogy result group management section 6 manages analogy result groups obtained by analogy processing by the analogy section 5 (step A9). These analogy result groups are, for example,
It can be managed using a relational database. FIGS. 30A and 30B are ER diagrams showing a schema for managing these analogous inference result groups. FIG. 31 (a)
(B) is a diagram showing an image in which analogy results of analogy items “potato chips” and “circle” are stored in a table of a relational database. As shown in FIGS. 30 and 31, the analogy result group can be managed by a general technique that is not special, so that the analogous result group management unit 6 can be easily mounted.

Next, the evaluation section 34 of the analogizable body group forming section 3
Thus, the analogy formula result group managed by the analogy formula result group management unit 6 is evaluated based on the fact group managed by the fact group management unit 2 (step A10). In the example shown so far, the evaluation as shown in FIG. 32 is performed. That is, the sum of the absolute value of the difference between the analogy result and the fact at each time is obtained, and a value obtained by raising the value to a power of 1/2 is defined as the evaluation result. In this case, the closer the evaluation result is to 1, the more accurate the analogy is.
If the evaluation result of the analogy expression group corresponding to the analogy item group as a container for all direct solutions is sufficiently close to 1, it is considered that the analogy processing has been completed, and the processing of this embodiment is performed. finish. In the case of the evaluation result as shown in FIG. 32, it cannot be said that it is sufficiently close to 1, so it is determined that analogy processing needs to be continued.

Here, the evaluation was performed by the above-described method, but any method may be employed as long as it is significant.

Next, the evaluation result group management unit 35 of the analogizable body group forming unit 3 manages the evaluation result group evaluated by the evaluation unit 34 (step A11). These evaluation result groups can be managed using, for example, a relational database. FIG. 33 is an ER diagram showing a schema for managing these evaluation result groups. FIG. 34 is a diagram showing an image in which these evaluation result groups are stored in a table of a relational database. FIG.
As shown in FIG. 3 and FIG. 34, the evaluation result group management unit 35 can manage the evaluation result group by a general technique that is not special, so that the evaluation result group management unit 35 can be easily mounted. it can.

In the example shown so far, the accuracy is not sufficient, and it is necessary to repeat the analogy process (step A1).
2) Again, the analogy expression group forming unit 33 improves the analogy expression group by a method as shown in FIG. 35 (step A13).

FIG. 35 is a flowchart for explaining the details of step A13 shown in FIG. In FIG. 35, first, for the analogy formula group managed by the analogous body group management unit 4, the evaluation result corresponding to each analogous formula is given by referring to the evaluation result group management unit 35 (step F1). . In the examples described so far, the analogy expression and the evaluation result are associated with each other as follows. -Analogical formula "Game-1" → Evaluation result "7.29E-64"-Analogical formula "Game magazine-1" → Evaluation result "7.29E-13"-Analogical formula "Game magazine-2" → Evaluation result " 6.35E-10 "-Analogy expression" Potato chips-1 "→ Evaluation result" 1.29E-26 "-Analogical expression" Potato chips-2 "→ Evaluation result" 3.51E-52 "-Analogical expression" Yen- 1 "→ Evaluation result" 4.24E-14 "

Here, regarding the analogy item "game magazine", the analogy formula "game magazine-1" has a better evaluation result than the analogy formula "game magazine-2", so the analogy formula "game magazine-1". And weakens the analogy expression "game magazine-2" (step F2). Here, the weakening means that the existence of the inferior analogy group is relatively weakened in order to leave the dominant analogous group. Here, if there is no other analogy group that depends on the inferential analogy, the analogy is discarded, and if there is another analogy group that depends on the analogous inferiority, dominant Try to increase the number of analogical expressions. In the example shown so far, there is no analogy equation group that depends on the analogy equation "game magazine-2", so the analogy equation "game magazine-2" is discarded. However, if this is left as it is, the group of analogy formula groups will shrink,
A new analogy formula "game magazine-3" is constructed based on the analogy formula "game magazine-1" (step F3).・ Analog expression "game magazine-3: int ((0.85 x" game-
1 "(t-30 * 60) + 2 * d (" yen-1 ") / dt) × (1 + 0.01 × sin2
πt)) ''

In this case, a new analogy equation is obtained by simply discarding the analogous equation group with a bad evaluation result and correcting the analogous equation group with a good evaluation result. May be obtained. For example,
In the example shown so far, a method of calculating the arithmetic mean of the analogy formulas "game magazine-1" and "game magazine-2" and using the arithmetic average as a new analogy formula "game magazine-3" may be adopted. good. In this way, by making corrections that allow for subtle behavior in the process of improving the analogy formula group, more complex knowledge than humans can grasp is reflected, and as a result, better analogy accuracy Can be obtained by analogy. That is, analogy can be performed with higher accuracy by using knowledge of a wide world that cannot be consciously grasped by humans.

The analogy group improved in this way is again managed by the analogizable body group management unit 4 (step A7), and then analogization processing is performed again by the analogy unit 5 based on the analogous group. (Step A8). Then, the analogy result group management unit 6 manages the analogy result again (step A9). Similarly, the evaluation unit 34 evaluates the analogy result (step A10).
In step 5, the evaluation result is managed (step A11).
At this time, if the evaluation result is not good (step A1)
2) Returning to the analogy expression group forming unit 33 again, the analogous expression group is improved (step A13). Then, when the evaluation result is sufficiently improved, the processing of the present embodiment ends.

Incidentally, the analogy expression obtained as the best obtained in this way is used as follows. That is, the analogy device of the present embodiment is incorporated into the ordering system of each store, and the analogy device performs analogy processing for analogy items that are containers for direct solutions for a necessary period to obtain future forecast data. After that, a list of commodities required for the next order and the number of commodities are arranged by the ordering system of each store based on the forecast data, and the order is placed to the convenience store chain headquarters.

As described above, according to the present embodiment, the analogy result group is sequentially improved while the analogy result group is evaluated based on the fact group, so that it is very excellent for inferring accurate sales of goods / services. The analogy formula of the analogy precision can be obtained. For this reason, the next order content can be inferred with high accuracy in a convenience store business or the like, and the occurrence of shortage can be prevented while minimizing store inventory.

Further, according to the present embodiment, by making the process of improving the analogy formula group non-deterministic by using a genetic algorithm or the like, a subtle behavior can be added in the process of improving the analogy formula group. So you can make corrections,
More complicated knowledge than a human can grasp can be reflected in the analogy expression, and therefore, analogy processing can be performed with higher accuracy using knowledge of a wide world that cannot be consciously grasped by humans.

In the above-described embodiment, the fact group management unit 2 may access an external data source such as a homepage on the Internet and add a fact group included in the external data source. In other words, in order to perform analogy processing with high accuracy, it is necessary to perform analogy processing using more facts, but it is impractical to create a database of all event elements that are regarded as facts in advance. It is. Therefore, in the process of improving the analogous inference formula group as described above, the number of fact groups managed by the fact group management unit 2 is dynamically increased. Specifically, as shown in FIG. 36, a home page on the Internet (a group of WWW servers) is periodically analyzed by the external data access unit 21 based on the URL list 21a. Then, as shown in FIG. 37, the fact group managed by the fact group management unit 2 is enhanced by regarding the appearance frequency and the like of the words included in the homepage 10 as facts. FIG. 38 is a diagram showing an example of a home page analyzed by the fact group management unit 2. From the home page 10 shown in FIG. 38, for example,
Words such as "administrative reform" and "global warming" are extracted, and the appearance frequency of these words becomes a fact, and is finally used as an analogy item. As a result, the knowledge of the wider world can be reflected in the analogy expression, and therefore, the knowledge of the wider world that cannot be consciously grasped by humans can be rationally captured and the analogy processing can be performed with higher accuracy. . Further, since the number of analogy items is dynamically increased, the current trend is appropriately reflected, and the arithmetic processing is performed without difficulty. Therefore, very realistic analogization processing can be performed.

FIG. 39 is a block diagram showing details of the fact group management unit 2 including the external data access means 21, and FIG.
5 is a flowchart for explaining a procedure for increasing a fact group managed by the fact group management unit 2. First,
As a preparation stage, a U serving as an address for specifying a resource on the Internet is provided inside the fact group management unit 2.
A URL list 21a listing RLs (Uniform Resource Locators) is stored. Here, the URL list 21
a is, for example, as shown in FIG. Also, U
The method of storing the RL list 21a may be any method using a database or the like, other than the method using a simple file. Next, the fact group management unit 2 uses the URL list selection unit 22 to select one U from the URL list 21a.
The RL 29 is selected (step G1, FIG. 42), and the contents of the corresponding homepage 10 are acquired by the page acquisition unit 23 (step G2, FIG. 43). Here, the page acquisition unit 23 uses an HTT such as a general WWW browser performs.
By performing access according to P (Hyper Text Transport Protocol), the contents of the home page 10 are acquired. Then, lexical analysis is performed on the obtained contents of the homepage 10 by the page lexical analysis unit 24, and a list of words (word list 25) included in the homepage 10 is created (step G3, FIG. 44). Then, it is determined by the variableizing unit 26 whether or not each word has been obtained so far (step G4, FIG. 4).
5) If it is a newly obtained word, it is regarded as a new variable and managed as a fact in the fact storage unit 27 (step G5, FIG. 45). On the other hand, if the word has already been obtained, the appearance frequency is updated and managed as a fact in the fact storage unit 27 (step G6, FIG. 45). Here, it is assumed that a word list obtained up to the time of the operation exists inside the variable conversion unit 26. Perform the above operations
This is performed for a sufficient number of URLs 29 included in the RL list 21a (step G7).

It is preferable that the contents of the home page 10 analyzed in this way be updated frequently in order to grasp some tendency.

FIG. 46 is a diagram showing the fact group management unit 2 including the update degree evaluation unit 28 for evaluating the update degree of the homepage. The fact group management unit 2 shown in FIG. 46 has basically the same configuration and operation as the fact group management unit 2 shown in FIG. Here, the same parts as those of the fact group management unit 2 shown in FIG. 39 are denoted by the same reference numerals, and detailed description is omitted. In FIG. 46, the update degree evaluation unit 28
Evaluates the update degree of the content of the homepage 10 acquired by the same method as the fact group management unit 2 shown in FIG. Here, the following two methods can be adopted as a method of evaluating the degree of update. The method of evaluating the update degree is not limited to these two methods, but may be any method that is significant.

First, the first method is to evaluate the degree of homepage update using the degree of volatility (volatility) of the checksum. In this method, the update degree evaluation unit 28 holds a checksum list 28a therein (see FIG. 47). As shown in FIG. 47, the checksum list 28a has a table structure in which a combination of the URL and the evaluation date and time of the update degree is used as a key, and the checksum is a value. Here, various methods can be adopted as a method of calculating the checksum. Here, as one example, the HTML document constituting the home page is regarded as a byte string, and the remainder obtained by dividing the total by 65536 is the checksum. Shall be adopted. The update degree evaluation unit 28 checks the checksum of each URL in a time series and calculates the volatility. Various methods can be used to calculate the volatility of the checksum. Here, a method of obtaining the sum of the absolute values of the deviation from the average value is adopted (see FIG. 48). In this method, it can be considered that the greater the volatility of the checksum of the home page indicated by each URL, the more the content changes.

On the other hand, the second method is to evaluate the degree of update of the homepage using a specific keyword and information on the surroundings. As shown in FIG.
The last update date 10a is often embedded in a predetermined location (for example, at the end). In this method, on such a homepage 10, "last updated on" is regarded as a keyword, and the date expression "February 13,
By recognizing "1998", the update date of the homepage 10 is grasped. For this reason, the update degree evaluation unit 28 holds therein a keyword list 28b and a last update date list 28c (see FIG. 50). in this way,
The shorter the update interval of the home page indicated by each URL is, the more the content can be regarded as having a change. In this method, a change in the contents of the home page without a change in the checksum, that is, a “steady change” in which the contents are not changed superficially despite being updated frequently (for example, There is an advantage that a variety of fact groups (for example, a fact expressed as a stationary function) can be extracted without fail.

As described above, since the homepage whose contents are frequently updated can be specified by the update degree evaluation unit 28, the URL included in the URL list 21a can be specified.
Can be brushed up, and the fact group managed by the fact group management unit 2 can be more effectively enhanced.

As described above, a homepage suitable for analysis is preferably one whose contents are frequently updated, but in addition, it is necessary that words useful for analogy processing appear frequently. It is. Therefore, the external data access unit 21 included in the fact group management unit 2
Based on the usage status of the added fact group, it is preferable to appropriately learn a homepage where words effective in analogy processing frequently appear as access targets.

Here, the step of learning a home page where words effective in analogy processing frequently appear is shown in FIGS.
This is performed in the URL selection unit 22 of the fact group management unit shown in FIG. 6 and can be regarded as included in step G1 in FIG.

FIG. 51 shows the URL of the fact group management unit 2.
FIG. 52 is a block diagram showing details of the selection unit 22, and FIG.
It is a flowchart for demonstrating the learning procedure of an L list. First, as a preparatory stage, as additional information given to a variable serving as a basis of an analogy item, a corresponding U is used so that it can be understood from which home page the variable was extracted.
RL is given. Next, the variable reference unit 41 specifies a variable serving as a basis of an analogy item included in the analogous expression group managed by the analogical expression group forming unit 33 (step H).
1). Here, a variable corresponding to the analogy item is set, and variable additional information is set to the variable (see FIG. 22). Therefore, the URL acquisition unit 42 refers to the additional information of the variable specified in this manner,
The URL of the home page as the extraction source can be obtained (step H2). Next, the reference counter increment unit 43 increments the reference counter 29a of the URL obtained in step H2 in the above-mentioned URL list 21a (step H3, FIG. 53). That is, as shown in FIG.
9a represents information indicating how many variables each URL 29 is extracted from. After repeating this operation for all variables (step H4), the URL
The deletion unit 44 leaves only the URL 29 with the larger value of the reference counter 29a from among the URLs 29 managed in the above-described URL list 21a, and discards the other URLs 29 (step H5). For example, the URL deletion unit 44
Operates to discard the lower 20% of the URL list 21a. Further, the URL supplementing unit 45 newly creates a URL 29 corresponding to the discarded URL 29 (step H6). In addition, as a method of newly creating the URL 29, in addition to a method of fusing the remaining URL groups, extension, deletion, or replacement of a part of the remaining URL, or a link in the homepage indicated by the existing URL is used. A method of tracing is conceivable.

A specific example of a new URL generation method will be described below. First, a method of expanding a part of the remaining URL is shown in FIG. In a large Web site, multiple Web sites
Due to the fact that the server b has been started, there is a case where management is performed such that a serial number is included in the URL. FIG.
The method shown in is for such a case.
In this example, “www” in the URL is extended to “www2”. Next, a method of deleting a part of the remaining URL is shown in FIG. The method shown in FIG. 55 is the reverse of the method shown in FIG. 54. In this example, “sub1” in the URL is deleted. Next, a method of replacing a part of the remaining URL is shown in FIG. When a large company or organization has launched a website in multiple countries, the first half of the URL is often the same except for the last part that indicates the type of country or organization in the URL. Is a valid URL by replacing the last part that represents the type of country or organization.
Can be newly generated. The method shown in FIG.
This corresponds to such a case, and in this example, UR
"Com" at the end of L is replaced with "co.jp". Finally, FIG. 57 shows a method of following a link in a home page indicated by an existing URL. Usually, a homepage (HTML document) specified by a URL has a plurality of links for accessing other homepages 12a, 12b, and 12c as shown in FIG. 57 because of its nature as a hypertext. Therefore, a new URL can be generated by extracting information of these links (described as a URL in the HTML document) from the homepage 10. In this example, "htt
p: //www.foo.com/ ”to“ http://www.foo.com/theme1
/ ”,“ Http://www.foo.com/theme2/ ”, and“ http:
//www.foo.com/theme3/ ”are newly generated.

As described above, by referring to the homepage reflecting the usage status of the variable serving as the basis of the analogy item used in the analogy formula, the external database to be accessed can be learned. Become like As a result, a home page linked to the remaining tendency of the analogy expression itself remains, and a home page in which words effective in analogy processing frequently appear is referred to.

The method of enhancing the fact group managed by the fact group management unit 2 is not limited to the above-described method using the homepage on the Internet, but may be, for example, a method using a database of another department outside or inside the company. Etc. can be adopted. That is, any method can be adopted as long as it takes in information existing outside the range of a system prepared in advance as an analogy device. In the present embodiment, a homepage provided via the Internet is merely exemplified as a representative of content that is not placed under the conscious control of a person who uses the analogization device.

In the above embodiment, the rule group management unit 1, the fact group management unit 2, the analogizable body group forming unit 3, the analogizable body group management unit 4, the analogization unit 5, and the analogization result group management unit 6 Can be realized by dedicated hardware, or can be realized as a program module operating on the computer system 50 as shown in FIG. Here, the computer system 50 includes a computer main body 51 and a CRT (Cath
ode Ray Tube), a printer 53, an input device 54 such as a keyboard and a mouse, and an FD (Floppy Dis
k) Drive 56, CD-ROM (Compact Disk-Read)
Only Memory) Various peripheral devices such as a drive 57 and a hard disk device 58 are provided. The program including the program module as described above is stored in the memory 57 or the hard disk 5 on the computer main body 51.
8 and a flexible disk 59
And stored in various computer-readable recording media such as an external storage device such as a CD-ROM 60 or the like, and sequentially read and executed from a CPU (Central Processing Unit) of the computer main body 51 as described above. Realizing various functions.

[0085]

As described above, according to the present invention, it is possible to use knowledge of a wide world that is highly accurate and cannot be consciously grasped by humans, and to rationally incorporate such knowledge. Analogical processing can be realized.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall configuration of an embodiment of an analogization apparatus according to the present invention.

FIG. 2 is a block diagram showing details of the analogization device shown in FIG. 1;

FIG. 3 is an ER diagram showing an example of a business (convenience store business) to which the analogization device shown in FIGS. 1 and 2 is applied;

FIG. 4 is a flowchart for explaining the operation of the analogization device shown in FIGS. 1 and 2;

FIG. 5 is a diagram showing an example of a rule group managed by a rule group management unit.

FIG. 6 is an ER diagram showing a schema for managing a group of rules in a database.

FIG. 7 is a diagram showing a management image of a rule group by a database.

FIG. 8 is a diagram showing an example of an interface used to input a rule group.

FIG. 9 is a diagram showing an example of a fact group (sales of a game).

FIG. 10 is a diagram showing an example of a fact group (sales of a game magazine).

FIG. 11 is a diagram showing an example of a fact group (sales of potato chips).

FIG. 12 is a diagram showing an example of a fact group (exchange rate of yen).

FIG. 13 is an ER diagram showing a schema for managing a group of facts in a database.

FIG. 14 is a view showing a management image of a group of facts (sales of a game) by a database.

FIG. 15 is a diagram showing a management image of a group of facts (sales of game magazines) by a database.

FIG. 16 is a diagram showing a management image of a fact group (sales of potato chips) by a database.

FIG. 17 is a view showing an image of managing a group of facts (exchange rate of yen) by a database.

FIG. 18 is a view showing an image of inputting a group of facts (sales of a game) from a file.

FIG. 19 is a view showing an image of inputting a group of facts (sales of a game magazine) from a file.

FIG. 20 is a view showing an image of inputting a group of facts (sales of potato chips) from a file.

FIG. 21 is a view showing an image of inputting a group of facts (exchange rate of yen) from a file.

FIG. 22 is an ER diagram showing a schema for managing an analogy item group and a variable group in a database.

FIG. 23 is a diagram showing a management image of analogy item groups and variable groups by a database.

FIG. 24 is a diagram showing an example of an interface used to input a group of analogy items.

FIG. 25 is a flowchart for explaining the details of step A6 (steps forming an analogous expression group) shown in FIG. 4;

FIG. 26 is a flowchart for explaining details of step B2 (step of reflecting a part of a rule in an analogy expression) shown in FIG. 25;

FIG. 27 is a flowchart for explaining details of step B3 (step of reflecting the remaining rules in the analogy expression) shown in FIG. 25;

FIG. 28 is a flowchart for explaining details of step B4 (step of reflecting knowledge other than rules in an analogy expression) shown in FIG. 25;

FIG. 29: Analogy expression in a file (Potato chips-
The figure which shows the storage image of 1).

FIG. 30 is an ER diagram showing a schema for managing a group of analogy results in a database.

FIG. 31 is a diagram showing a management image of a group of analogy results by a database.

FIG. 32 is a diagram for explaining an evaluation method in the evaluation unit of the analogizable body group forming unit.

FIG. 33 is an ER diagram showing a schema for managing an evaluation result group in a database.

FIG. 34 is a view showing a management image of an evaluation result group by a database.

FIG. 35 is a flowchart for explaining the details of step A13 (step for improving the analogous expression group) shown in FIG. 4;

FIG. 36 is a diagram showing external data access means incorporated in the fact group management unit.

FIG. 37 is a conceptual diagram for explaining a method of enhancing a fact group by analyzing a homepage.

FIG. 38 is a diagram showing an example of a home page analyzed by the fact group management unit.

FIG. 39 is a block diagram showing details of a fact group management unit including external data access means;

FIG. 40 is a flowchart for describing a procedure for increasing a fact group managed by a fact group management unit;

FIG. 41 is a diagram showing an example of a URL list managed by a fact group management unit.

FIG. 42 is a diagram showing an example of selecting an arbitrary URL from a URL list.

FIG. 43 is a view showing an example of acquiring contents of a homepage based on a URL.

FIG. 44 is a diagram showing an example of obtaining a word list by lexical analysis of contents of a homepage.

FIG. 45 is a view showing an example of obtaining a fact based on the contents of a homepage and a word list.

FIG. 46 is a block diagram showing details of a modification of the fact group management unit including external data access means.

FIG. 47 is a view showing an example of a checksum list held by an update degree evaluation unit shown in FIG. 46;

FIG. 48 is a conceptual diagram for explaining a method of calculating the degree of change (volatility) of the checksum.

FIG. 49 is a view showing an example of a home page in which the last update date is embedded.

FIG. 50 is a view showing an example of a keyword list and a last update date list held by the update degree evaluation unit shown in FIG. 46;

FIG. 51 is a block diagram showing details of a URL selection unit of the fact group management unit;

FIG. 52 is a flowchart illustrating a learning procedure of a URL list related to a homepage analyzed by a fact group management unit.

FIG. 53 is a view showing an example of a URL list to which a reference counter for grasping a URL reference status is added;

FIG. 54 is a view showing an example of a new URL generation method (expansion of a part of the URL).

FIG. 55 is a view showing another example (a part of a URL is deleted) of a new URL generation method.

FIG. 56 is a view showing another example (replacement of a part of a URL) of a new URL generation method.

FIG. 57 shows still another example of a new URL generation method (UR
FIG. 9 is a diagram showing an example of an L document that uses link information included in an HTML document to which L is specified.

FIG. 58 is a block diagram showing an example of a computer system on which an analogy program for realizing the analogy apparatus and the method thereof according to the present invention is implemented.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Rule group management part 2 Fact group management part 3 Analogical body group composition part 4 Analogical body group management part 5 Analogical part 6 Analogical result group managing part 21 External data access means 31 Variable group extracting part 32 Analogical item group managing part 33 Analogous Expression Group Configuration Unit 34 Evaluation Unit 35 Evaluation Result Group Management Unit 59 Flexible Disk (Recording Medium) 60 CD-ROM (Recording Medium)

Claims (11)

[Claims] 1. A rule group management unit for managing a rule group in a problem area to be inferred, a fact group management unit for managing a known fact group in the problem area, and managed by the rule group management unit. Based on a rule group and a fact group managed by the fact group management unit, an analogous body group configuring unit that configures an analogous body group used in analogy processing, and configured by the analogous body group configuring unit An analogy-possible body group management unit that manages the analogy-possible body group, an analogy unit that performs analogy processing based on the analogy-possible body group managed by the analogous body group management unit, and an analogy process by the analogy unit. An analogy result group management unit that manages the analogy result group obtained by the above, wherein the analogizable body group constituting unit is configured to reflect the evaluation result of the analogy result group managed by the analogy result group management unit, The analogizable body group is configured, and the analogizable body group is sequentially improved in the process of repeatedly performing the analogy processing by the analogizable body group constituent unit, the analogizable body group management unit, the analogization unit, and the analogy result group management unit. An analogy device characterized by: 2. The analogy inference apparatus according to claim 1, wherein the analogizable body group forming unit configures the analogous body group in a non-deterministic improvement process. 3. The method according to claim 1, wherein the analogizable body group forming unit appropriately selects and refers to a part of the rule group managed by the rule group managing unit in the process of improving the analogizable body group. The analogization device according to claim 1. 4. The analogizable body group constituting unit is configured to appropriately select and refer to a part of a fact group managed by the fact group management unit in the process of improving the analogous body group. The analogization device according to claim 1. 5. A variable group in the problem area based on a rule group managed by the rule group management unit and a fact group managed by the fact group management unit. A variable group extracting unit for extracting the analogy item group to be analogized in the problem area; and an analogy item group management unit for managing analogy item groups to be analogized in the problem area. An analogy formula group that constitutes the analogous formula group obtained as the analogizable body group, and an analogy result group managed by the analogy result group management unit based on a fact group managed by the fact group management unit. An evaluation unit to be evaluated, and an evaluation result group management unit that manages an evaluation result group evaluated by the evaluation unit. Analogy apparatus according to claim 1, wherein the sequentially improving the analogy expression group while reflecting the evaluation result group managed by the evaluation group managing unit. 6. An apparatus according to claim 1, wherein said fact group management unit has external data access means for accessing an external data source and adding a fact group included in said external data source. Analogue device as described. 7. The external data access means learns an external data source to be accessed based on the use status of the added fact group.
Analogue device as described. 8. A rule group management unit for managing a rule group in a problem area to be analogized, a fact group management unit for managing a known fact group in the problem area, and managed by the rule group management unit. A rule group and an analogy unit that performs analogy processing based on the fact group managed by the fact group management unit, wherein the fact group management unit accesses an external data source and includes the external data source in the external data source. An analogy inference apparatus comprising external data access means for adding a group of facts to be added. 9. The external data access means learns an external data source to be accessed based on the use status of the added fact group.
Analogue device as described. 10. A step of managing a group of rules in a problem area to be inferred, a step of managing a group of known facts in the problem area, and a step of managing analogies based on the managed group of rules and facts. Constructing an analogous body group that is to be obtained, managing the configured analogable body group, performing analogy processing based on the managed analogable body group, and analogy results obtained by the analogy processing Managing the group, the step of configuring the analogizable body group is a step of repeatedly performing the analogy processing, sequentially reflecting the evaluation result of the managed analogous result group while sequentially reflecting the analogizable body group. An analogy method characterized by improvement. 11. A procedure for managing a group of rules in a problem area to be inferred, a procedure for managing a group of known facts in the problem area, and a procedure for analogy processing based on the managed rules and group of facts. The procedure for constructing a set of analogable bodies that can be used, the procedure for managing the configured analogable bodies, the procedure for performing analogy processing based on the managed analogable bodies, and the analogy result obtained by the analogy processing And a step of managing the group, the step of configuring the analogous body group is a step of repeatedly performing the analogy processing, and reflecting the evaluation result of the managed analogous result group while reflecting the analogous result group. A computer-readable recording medium on which an analogous inference program is sequentially improved.