JPH02133823A - Relative item retrieving system for knowledge arrangement editor - Google Patents

Abstract

PURPOSE:To easily verify the matchability of inter-item relation in the arranging process of knowledge by defining hierarchically the relation between items in multiple names, and retrieving the relative items based on the higher rank concepts and different concepts of the defined relation names. CONSTITUTION:A retrieving means 5 consists of a relation defining/changing means 7 which defines the relations among items via a graphic editor (capable of editing rectangles and segments) and a text editor and an object retrieving means 6 which retrieves actually the item objects of a graphical net by reference to the defined relations. Thus it is possible to retrieve a specific item that links a certain item via a specific relation. This relation can be hierarchically defined in multiple names, and the questions are possible with the higher rank concepts and in different viewpoints. As a result, the matchability of relations can be verified graphically and easily among items in the arranging process of knowledges.

Description

[Detailed Description of the Invention] <Industrial Application Field> The present invention relates to an information organization support function, and more specifically, to the acquisition of knowledge from experts, which is said to be a bottleneck in constructing an expert system. Relationships in the knowledge organization editor that supports the work associated with general problem solving and organization represented by the KJ method, etc., and allows even people who are not familiar with computers to easily verify the knowledge given to the computer. This relates to an item search method.

Fields of use include construction of expert systems related to LR (artificial intelligence), database search systems, qualitative simulations of social phenomena, information management and idea support for general individuals (any subject matter is acceptable), etc.

<Conventional technology> When structuring and organizing unorganized knowledge and managing it, an information organization support function that can be easily expressed and manipulated while appealing visually is required. In response to such needs, software for managing information on computers using diagrams such as cards, arrow diagrams, and spreadsheets has recently been increasing.

However, when it comes to managing a certain amount of knowledge while maintaining consistency, or when using and searching knowledge after it has been organized, there are currently few methods that fully meet the needs of users. In particular, it is rare to find a search that takes into consideration the hierarchy of target items and inter-item relationships.

Therefore, the purpose of this invention is to be able to easily verify the consistency of item quantity relationships in the process of organizing knowledge, and to search for related items taking into account the hierarchy of target items and item quantity relationships. The purpose of this invention is to provide a related item search method in a knowledge organization editor.

Means for Solving the Problem> In order to achieve the above object, the related item search method in the knowledge organization editor of the present invention graphically represents the relationship between items on a display using cards, arrows, etc. Create a graphical net, search for related items from the items in the graphical net using the relationship between items as a key, and if the search item has relationships with the same name that are connected to other items in a chain, Search for related items by tracing the relationship up to an arbitrary step, define the relationship between items multiple times hierarchically using several names, and search for related items based on the superordinate concept or different concept of the defined relationship name. The feature is that it allows you to search for.

<Function> ■It functions so that it is possible to search what kind of items are linked to a certain item through what kind of relationship.

■Also, for relationships with chain attributes, the relationship chain can be traced to any depth to search for related items. Even if the item itself has a hierarchy, it functions so that chained attributes between hierarchies can be defined, and it functions so that searches can be performed across upper and lower concepts.

■It works so that relationships can be defined multiple times in a hierarchical manner using several names, and it works so that questions can be asked from a superordinate concept or from a different perspective of the defined relationship name.

<Embodiment> An embodiment of the related item search method of the present invention will be described below with reference to the drawings.

Figure 1 shows an example of the tools needed to implement this method.

The input means 1 is a computer peripheral device for inputting characters and figures, such as a keyboard and a mouse.

The display means 2 refers to a display device of a computer that can display graphics using a bitmap method, and an O device that allows multiple screens to be displayed on the same screen, such as a window system.
9, and includes screen switching means and graphic display means.

The search means 5 includes a relationship definition/change means that defines and changes relationships between items using a graphic editor (one that can edit rectangles and line segments) and a text editor, and a relationship definition/change means that refers to the defined relationships and searches the graphical network This is computer software that consists of an object search means that actually searches for item objects.

The storage means 8 includes an object storage means 9 that stores in on-memory a graphical net converted into an internal structure form, and a relation definition that stores on-memory the defined relationship converted into an internal structure form. The storage means 10 corresponds to the storage device of the computer.

The control means 11 is a computer software environment for managing and controlling these means.

Figure 2 shows an example of the screen specifications of this method. In this example,
A list of relationships that item rXJ has is called up on the right side of the screen. Select "Cause" in the relationship list and have the graphical net trace the "Cause" relationship from rXJ to the item rAJrBJrcJ that is the cause.

The illustration is shown in reverse video.

In order to implement this method, a graphic editor that can edit a graphical net and a list display/search request menu screen that allows searching for items while viewing the graphic editor are required, for example.

The objects that are edited in the graphic editor include cards, nesting, and the arrows (
I will explain by taking up the arc) and the text. The relationship list in FIG. 2 shows that the names (relationship names) of arrows going in and out of card rXJ are "result,""relationship," and "cause."

Generally, items are linked through a specific relationship, so it is possible to search for related items from a specific item using the specific relationship as a key.

This search step is not limited to one step, but also needs to be compatible with a multi-step search (chain search) when a search item has a relationship with the same name that is linked to another item in a chain.

FIG. 3(a) shows an example of the definition of whether or not a certain relationship can be searched in a chain, and FIG. 3(b) shows a specific example of a chain/non-chain relationship. The examples here are not absolute and can be defined by the knowledge organizer depending on the situation.

Furthermore, when considering the path of the chain, relationships need to be classified into at least the following three types. In other words, there are three types: ・Bidirectional search is possible and the same name can be searched from both directions ・Bidirectional search is possible but the relationship name is different on one side and the other ・Three types that can only be searched in one direction need to be classified into one type. An example of this relationship type definition is shown in FIG.

Furthermore, the nested and nested 1,000 yen items are also special "relationships" (parent-child relationships), and it is necessary to define chain attributes and relationship types for these relationships as well. This relationship type is a "bidirectional synonym" in the words of Figure 4, but defining chain attributes requires some steps. In other words, as shown in Figure 5, when a new nest is created, after defining whether to allow chaining to its own parent nest2) whether to allow chaining to its own children, On the other hand, 3) Define whether or not to allow chaining to llll, and 4) Redefine whether or not to allow chaining from the parent.

・Means for realizing function ■ In order to realize this function, create a data structure in which a bidirectional reference pointer is attached to each object, ``card,''``nested,'' and ``arc,'' which is the object of editing in the graphical net. It is necessary. An example of this data structure is shown in FIG. With such a data structure, for example, the parent of a specific nest can be found using the link "5uper", and the child can be found using the link "°children".

To find descendants and ancestors, you can follow the links many times.

Similarly, use the “arcs” link to find the arrows (arcs) that go in and out of itself, and
The item to which the arrow (arc) is directed can be found using the "e" or "target" link.

Once you have the data structure as described above, you can create a program that follows the links of the necessary data. An example of the basic procedure of the program is shown in FIG.

・Means for achieving effect (■) In order to enable chained searches, it is necessary to define chained attributes and relationship types as shown in Figure 3 or 4, and then incorporate them into the basic data structure ( Figure 6 chai
n-from”, “chain-to”,
"chain-attrib", "type-der"
), check whether the search is possible or not while referring to these data at the time of search. If the search is possible, an l-step search is performed using the basic search procedure shown in FIG. You can perform a chain search by repeating this procedure until you run out of destinations.

・Means for realizing effect (■) In order to realize effect (■), it is necessary to define a hierarchical relationship in the relationship itself. Here, we will present two types of relationship hierarchy definition methods that are considered useful from the viewpoint of item retrieval.

One of them is multiple relationship hierarchy definition, and an example of how to define it is shown in FIG. As shown in this figure, an arc defined by the name indicated by the root can be searched by using the name indicated by any leaf above it (leaves 0 to 12). Furthermore, as shown on the concrete side of this figure, the multiple relationship hierarchy can be used, for example, when one concept is viewed from multiple different perspectives. In this case, it is possible to search for ``the subversion of a is c, the opposite concept of Jra is C'', ``the unified concept of a is c', the opposite concept of Jra is b''.

The other one is a single relationship hierarchy definition, and an example of how to define it is shown in FIG. As shown in this figure, an arc defined by the name indicated by a leaf can also be searched using the name indicated by the root above it. Also, if you use a single relationship hierarchy as shown in the example in this figure, for example, when you ask, "What is the purpose of a?", you can say, "A's purpose 1 is Jra's purpose 2."
In a sense, the efficiency of questioning is better than asking a different question twice. In other words, a single relational hierarchy is characterized in that a search request can be abstracted using a more general concept.

The work procedure for defining a relationship hierarchy is generally provided by a graphic editor, and the data structure (6th chapter) is created by interpreting the graphical net of the defined relationship hierarchy.
"5upper" and 5upper-type of the arc in the figure
All you have to do is create a program that writes the value to ”).

Figure 10 shows the names of cards, nests, and arcs, appropriate modifiers, final forms of fixed question sentences, and case particles that are selected from the presented set by menu selection, and a fixed question sentence is created by combining them. This is an illustrative example showing the possibility of devising a way to make inquiries as if they were made using natural language. As can be seen from this example, if you prepare the starting point of the combination sentence, the method for specifying the selected relationship, and many search methods, you can search for effects ■～■ using a limited number of types of sentences. Can be done.

<Effect of the invention> By implementing this method, the following effects can be obtained.

I) Verification of the consistency of item quantity relationships in the process of organizing knowledge can be easily performed graphically.

2) Flexible relationship definitions can be made when organizing knowledge. It also allows flexible questions to be asked when searching for knowledge.

[Brief explanation of the drawing]

Fig. 1 is a diagram showing an example of the overall configuration of a system for realizing an embodiment of the present invention, and Fig. 2 is an example of specifications of a computer display screen necessary for this embodiment and relation name calling from a specific item. FIG. 3 is a diagram showing an example of chain relationship definition in this embodiment. FIG. 4 is a diagram showing an example of relationship type definition in this embodiment. FIG. 6 is a diagram showing an example of a circular chain definition, FIG. 6 is a diagram showing an example of the basic data structure necessary to implement this embodiment, and FIG. 7 is an example of the basic procedure of a search operation in this embodiment. FIGS. 8 and 9 are diagrams each showing an example of relational hierarchy definition in this embodiment, and FIG. Figure 1 Patent Applicant Sharp Corporation Agent
1 patent attorney Qishaku 葆 Haka Fig. 7 Fig. 8 - I column of the definition of Sui Circulation of the current heavy related matter Fig. 9 Fig. 10

Claims (1)

[Claims] (1) Create a graphical net that graphically expresses the relationship between items on a display using cards, arrows, etc., search for related items from the items in the graphical net using the relationship between items as a key, If a search item has a chain of relationships with the same name that connects to other items, the relationship can be traced to any step to search for related items, and the relationships between items can be hierarchically determined. A related item search method in a knowledge organization editor, characterized in that related items are defined multiple times by the name of the relationship name, and related items are searched based on a superordinate concept or a different concept of the defined relationship name.