JPS60254326A - Expressing method of knowledge - Google Patents

Abstract

PURPOSE:To access attributes included in plural objects by linking objects having the same viewpoint as that of the objects having said attributes and referring to the attributes from the linked respective objects. CONSTITUTION:When a pair S1 of a class and an attribute is applied, a model selecting part 10 retrieves a corresponding table 12 for attributes and models to find out a model. The obtained model S2 is sent to a data discrimination part 14 together with the pair S1 of the class and the attribute and the data discrimination part 14 retrieves a data base 20 on the basis of the S1 and S2 to retrieve the required attribute. When the required attribute is obtained, it is outputted, but if there is no required attribute, the information S1 and S2 are sent to an object determining part 16 and its high-order class is outputted in the S2 route. The data discrimination part 14 accesses the data base 20 on the basis of the outputted class. Thereafter, similar processing is repeated to retrieve the data base 20 and search the required attribute.

Description

[Detailed Description of the Invention] [Industrial Field of Application] The present invention relates to a method of representing knowledge in an information system that utilizes knowledge, and in particular to a system that simultaneously allows users to view a plurality of different views of the target world in one system. Concerning knowledge representation methods that can be incorporated into

[Conventional technology]

In an information system that uses knowledge, such as a database system, how to represent knowledge is an important issue because it determines how to access the system and how to use it. There are many different ways of looking at knowledge. For example, if we consider "human being" as an example of knowledge, this human being can be understood as a living thing, an R source (human resource), or as a man/woman, adult/child, parent/child, and there are various ways of looking at it. .

Conventional knowledge representation methods include the frame-type method, which can naturally describe the relationships between objects in the target world, but it can only show one view of the target world, and can express the object from multiple perspectives. Even if I wanted to describe the world, this was difficult.

[Problem that the invention seeks to solve]

However, the present invention aims to provide a knowledge representation method that allows multiple views of the target world to be described in an easy-to-understand manner, thereby allowing access to the attributes of multiple target objects.

[Means for solving problems]

The present invention provides an object in which, in a knowledge representation system for an object world consisting of a large number of objects, at least some of which have attributes, each object having the attributes can be viewed in the same way as the object. This method is characterized in that objects are linked and the attributes can be referenced from each linked object.

The elements that constitute knowledge of the target world are objects. When an object with an attribute is linked to an object that can be viewed in the same way as the object, for example, in a superior-subordinate relationship, the link provides the inheritance path of the attribute.

The model created in this way gives only one perspective. Therefore, in the present invention, various views are linked, so that multiple views of the target world can be expressed simultaneously. Next, the configuration and operation of the present invention will be explained in detail with reference to the drawings.

〔Example〕

Figure 1 is a diagram that organizes knowledge from multiple perspectives. What is surrounded by a solid circle is an object that is a unit of knowledge, and t
(uman is the aforementioned human being, Man/Father is a man or father (both a man and a father), Woman/Mot
"her" indicates a woman or a mother (both a woman and a mother), and the same applies hereinafter. Arrows connecting objects link each object from a certain perspective, and arrows with symbols such as age indicate attributes or access paths of the objects. ^ge surrounded by a dotted circle indicates age, which is an attribute of Human, and a database containing the age of each individual may be assumed. Similarly, 5upport (to support a child) enclosed in a dotted circle is an attribute of the parent, and it may be assumed that it is a database containing the names of dependents.
(Respect) is a child's attribute, and a database containing people that Child respects may be assumed. Of course, these databases may be independent or each field of one database. An unsigned arrow or link is called a super link, and the arrow indicates the superior/subordinate relationship of each object. For example, man (man) is below "Human", and conversely, "human" is above men and women. Although not shown, the superordinate concept of humans is animals and living things.

If links are established in a superior/subordinate relationship, the lower-level object can refer to the attributes of the higher-level object by following the link. For example, if the higher-level object ``animal'' has attribute information that ``has a head,'' then the lower-level object ``human'' has attribute information that ``has a head,'' and the attribute is inherited.

In FIG. 1, links are set in a superior-subordinate relationship with a superior object having an attribute as the vertex. i.e. boy or son B
oy/Sob is subordinate to a man or father, a girl or daughter (iirl/[loughte
r is a subordinate of Woman/Mother,
may also be used as Father or Mo.
"ther" is subordinate to Parent, and "Boy" or "Girl" is subordinate to Child. For each object, the one indicated by the arrow is higher, and the opposite is lower.

If we view these objects in a superior-subordinate relationship to living things (called a biological view B-View), we can
There is a series of n women, boys, girls, and they are viewed from a parent-child relationship (parent-child relationship perspective P Vie
), Parent, Child, and father F
Mother, Son, Daughter
I can do it in the er series. In Figure 1, the target world is B - V
iew, P-View (7) It is described from two perspectives. In conventional methods, the world of objects can only be viewed from one perspective, but with the present invention, the world can be viewed from multiple perspectives.

As described above, in the present invention, a plurality of series are set in the target world corresponding to a plurality of views of the target object. The series of upper-lower relationships is expressed by 5upper, and the subscript i=1゜2,...
・is used to distinguish them from each other), and the object is 0ij (i
is the number of the series and j is the number within the series), the above-mentioned plural series (superlinks) are expressed as follows.

5uperl -t・o 1r, OI 2', ・・
""5upper2-= 021, 022, '・
...As shown in Figure 1, 5upper 1 is, for example, H
umar++ Man+Boy series, 5uper 2
is the series of Human+Woman and Girl, and these are the series of 13-View. In contrast, ParenL, Father, Son, and P.
arent.

There are also the Mother and Daughter series, which are P-View series.

In Figure 1, the object Humar++ Parent, C
hild attribute agel 5upport, resp
ect and Human -(age)-AgeP
arent" (support) -"Child
Child-(respect)-”Paren
Connect with t. Other objects are connected in two directions by super links, and are ultimately connected to objects with the above attributes.

Man/Father-(B-5upper)
→Human-(P-5upper)-” Par
entWoman / Mother-(B-5up
er) MI(uw+an(P-5upper)
-”ParentBoy/Son -(13-5
upper) -”Man/Father-(P
-5uper) -+ChildGirl/Dau
ghter-(B-5upper)-”Woman/
Mother-(P-5upper)-Chill
In the object world, there is a series or way of looking at attributes.
If you use the correspondence table (View) (this is called the first method), you can perform a desired search using only the object you want to search and the attribute name. For example, when calculating "boy's age," one knows from the correspondence table that this view is B-View, and specifies B 75upper for Boy 7'Son. In this way, you can get Man/Father from Boy/Son's B-3uper, get HuIIlan from that B-5uper, and get that ^ge.
You can get the answer from. Similarly, ``the person my daughter respects'' is a P-View, so Girl/D
Specify PView for older, and this will cause Ch
Obtain ild and then reach Parent. Previously 5u
There is only one type of per link and only one definition is possible, but the present invention (provides multiple 5uper links),
It can be defined by mixing multiple viewpoints, and is superior in terms of efficiency.

FIG. 2 shows a database system according to the first method, in which 10 is a model (series or Vies) selection section, 12
14 is a data discrimination section, 16 is an object determination section, and 20 is a database. When this system is given a bear S1 of class (boy in the above example) and attribute (same age), the model selection unit 10 searches the attribute-model correspondence table 12 to find a model. In this example, this is B-5upper. This obtained model B-3upper (referred to as S2) is sent to the data judgment section 14 together with the class and attribute bear 3.1, and the data judgment section 31. In S2, the database 20 is searched for attributes (age) that can be claimed. If you can get the attribute you want (Y), output it, but if not (N) information 3
1. S2 is sent to the object determination unit 16, and S2, that is, 13-5
The upper class is output using the upper route, the data discrimination unit accesses the database using it, and the same process is repeated. In this example, the class is Man, )luma
Change it to n and search for -20 in the data for each class to find the attribute that can be requested. In this example, this is a class of Human, and its data (age) is output.

Lower object Boy/Son + Girl/Da
``ughter'' etc. also have attributes and are usually stored in a database. However, if all the data for each object were recorded in the database for each object, the database would become enormous. Therefore, if there is a separate database containing information on age, hobbies, dependents, respected persons, etc., the capacity can be reduced by referring to that database and omitting the relevant object from the database.

Regarding FIG. 2 above, if the data discriminator 14 is given the ability to judge whether the attribute in question is in the class or not, the database will be accessed every time a new class is output from the object decider. There is no need to do so; it is possible to check the class output by the object determining section to see if it has the desired attribute, find the next class if it does not exist, and access the database using the class that is found to have the desired attribute. This point also applies to FIG. 3, which will be described later.

Furthermore, "class" and "object 0bjectJ" are the same thing, but in detail, "class" refers to something abstract such as Human, and object refers to an individual concrete thing such as a personal name.

Figure 1 (DHuman+ Parent+ Child
Man/F
If you have multiple 5uper links extending from there, such as ather, you may do the following.

Man / Father Super (FIt+man age) (Paren
t support) Woman / Mother Super (Human age) (Paren
t support) Boy /5on Super (Man / Father age)
(Child respect) Girl/Daug
hter Super (Woman/Mother age)
) (Child resper, t ) i.e. 5uper
There is only one link, but for example the object Man/
For Pa Lher, age is Human ag.
e If %5upport, Parent 5upport
I instruct them to fetch t. By doing so, the model selection section and the correspondence table between attributes and models become unnecessary, and FIG. 2 becomes as shown in FIG. 3. This method is called the second method. When determining the above-mentioned "boy's age" using this second method, the object (boy) and the attribute (age) S3 are input to the data determining section 14, and in response to this, the data determining section 1
4 searches the database 20 for the object Boy/Son and finds the attribute age. In this example, Boy/Sob
There is no age data (N), so database 20
The data read from is sent to the object determining section. When the object determining unit 16 checks the data, it determines whether Man/Fath
The data judgment section 14 is informed that the message "er" should be looked at. The data determination unit 14 determines the target object Man/F.
When the database 20 is accessed with ather and the read data is checked, there is no data "age" there, so when it is sent to the object determination section, if age is "Human", the data determination section 14 accesses "Human" this time. The age of the data boy is obtained (Y), and the data determination section outputs this.

There may be cases where the desired data is not in the database, in which case the output will be that there is no corresponding data.

Figures 2 and 3 are constructed from a computer system, and to explain the search operation in this case, in this case, the attribute of the object, for example the age of Human, is the information on how to retrieve the data from the database. Assume that an access pass has been specified, and that the question is "Please tell me the age of Fat."

In the method shown in Figure 2, this question is written as follows.

? (Boy/Son age) s, t, (B
oy/5on-bold part) When this is input to the model selection unit 10, B-5upper is selected from age and a question sentence in the following format is sent to the data determination unit 14.

? (Boy/Son age B-5upper)s
, t, (Boy/Son = thick)Boy/So
Since n does not have an access path to aBe, the object determining unit outputs the next object Man/Father, and since there is no access path to aBe, the next object Human is output.

)1uman has an access path, so what about the data discrimination unit 14? (Human age B-5upper)s, t,
The database 20 is accessed with (Bo-y/Son = fat part), the answer "10 years old" is obtained, and this is output.

The case of FIG. 3 is similar to y' and is described as follows.

'>, (Boy/Son age) s, t,
(Boy/Son = thick part) This is the data judgment section 14
If you enter , it will be checked whether the Boy/Son has an access path for age.

Since there is no such case, the process is moved to the target object determining section 16.

Here, we find Man/Father among the 5uppers of Boy/Son as having age as an attribute.

? (ffan/Father age)s,
t, (Boy/5on-thick part) This is sent to the data determination section 14 again. In this case, Man/Father also has a
Since there is no access path for ge, it is sent to the object determination section. Here, Human is obtained from Man/Father as having age as an attribute.

? (Human age) s, t, (Boy/
Human
Since there is an Aksho pass for age, execute this and get "10 years old" as the answer.

〔Effect of the invention〕

As explained above, according to the present invention, multiple views on knowledge of the target world can be described simultaneously, and when applied to a database system, multiple databases can be used, and storage capacity can be reduced by avoiding duplication. Benefits can be obtained.

[Brief explanation of drawings]

Figure 1 is an explanatory diagram of the target world of the present invention, Figures 2 and 3 are
The figure is a block diagram and does not show the configuration of the system. In the drawing, 10 is a model selection section, 12 is an attribute-model correspondence table, 14 is a data discrimination section, 1G is an object determination section, 20
is a database. Figure 1 Figure 2 Figure 3 Procedural amendment (voluntary) 1. Display of the case Patent Application No. 111790 of 1982 2. Name of the invention Knowledge expression method 3. Person making the amendment Relationship with the case Patent applicant residence Address: 1015 Kamiodanaka, Nakahara-ku, Yozaki City, Kanagawa Prefecture Name (522) Fujitsu Limited Representative: Takashi Yamamoto 4, Agent: 101-5, Date of amendment order: None 6, Number of inventions increased due to amendment None 8. Contents of amendment (i) "Database" on page 9, line 8 of the specification shall be amended to "database". (2) In the 7th line from the bottom of page 9, "to record" is corrected to "to record".

Claims (1)

[Claims] In a knowledge representation system for an object world consisting of a large number of objects, at least some of which have attributes, for each object that has the attributes, an object that can be viewed in the same way as the object is linked. , a knowledge representation method characterized in that the attributes can be referenced from each linked object.