JPH0464130A - Knowledge base searching system - Google Patents

Abstract

PURPOSE:To operate a high-speed searching process by searching the object being an object to be searched in the connection direction of a genealogical tree according to an index from a high-order class object searched by a width priority searching means. CONSTITUTION:When a class object 3(P) is searched, a designation P of the class object 3 to be searched and a designation C of the high-order class object 3 are set in an index setting means. Then, the width priority searching means width-priority searches each class object 3 from a root object 2 based on the index C, so that a class object 3(C) can be searched. Then, a depth priority searching means depth-priority searches each object 3 and 4 from the searched class object 3(C) according to the index P, so that an object 3(P) being the object to be searched can be searched. Therefore, a useless search for the objects 3 and 4 lower than the high-oder class object 3 can not be operated. Thus, a range to be searched is reduced so that the searching process can be operated at the high-speed.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a knowledge base search system in which knowledge base objects are connected in a phylogenetic tree.

In conventional technical knowledge base search systems, when performing inference or knowledge editing, the search means searches for one object from a large number of objects in the knowledge base based on a predetermined index. Here, the knowledge base has objects connected in a phylogenetic tree, and in this case, the objects are a root object located at the tip, a class object located in the middle, and an instance located at the end. Consists of objects.

In the conventional method of searching for a single object from a knowledge base in which objects are connected in a phylogenetic tree as described above, the name of the object to be searched is usually set as the index of the search method, and the phylogenetic tree is searched from the root object. Objects are searched by performing breadth-first search, which searches sequentially in the branching direction, and depth-first search, which searches sequentially in the direction of connection of the phylogenetic tree.

Problems to be Solved by the Invention By doing the above, it is possible to search for a single object from a knowledge base in which objects are connected in a phylogenetic tree, but this method does not allow for high-speed search processing because the search range is large. It is difficult to

In order to solve this problem, it has been proposed to manage object names and pointers to entities in a table such as a binary tree or to perform hashing.

However, these methods require a mechanism to manage object information separately from the knowledge base structure, resulting in a complex system structure.Furthermore, the above-mentioned management table cannot be used when the knowledge base structure changes. The system is difficult to maintain because it needs to be updated.

Furthermore, when actually searching for objects using a knowledge base search system, it is inevitable that related objects are searched sequentially. In other words, since it is possible to speed up the search process by preferentially searching for objects related to the object searched in the previous search process, there is a need for a system that easily realizes this.

Means for Solving the Problems The invention according to claim 1 is a knowledge base in which a root object located at the tip, a class object located at the middle, and an instance object located at the end are connected in a phylogenetic tree. In a knowledge base search system provided with a search means for searching for one object, an index setting means is provided for setting the name of the object to be searched and the name of a class object located above the object as an index of the search means, and A breadth-first search means is provided to search for upper class objects from the root object in the branching direction of the phylogenetic tree according to the index set in the index setting means, and from the upper class objects searched by this breadth-first search means in the concatenation direction of the phylogenetic tree according to the index. A depth-first search method was installed to search for the object to be searched.

The invention according to claim 2 provides a search for searching for one object from a knowledge base in which a root object located at the tip, a class object located at the middle, and an instance object located at the end are connected in a phylogenetic tree. In the knowledge base search system provided with means, a search range defining means is provided in which a plurality of object search ranges are set in advance according to predetermined conditions for the previously searched object, and a last object storage means is provided for temporarily storing a pointer of a searched object. and an in-range search means for reading the search range from the search range defining means to search for the next object based on the object of the previous search selected from the pointer temporarily stored by the last object storage means; An effective search range detection means is provided for accumulating the number of searches for objects in each search range, and the priority order of the search range output from the search range definition means to the within-range search means is determined according to the integrated value of the effective search range detection means. We have provided means to improve search efficiency.

The invention according to claim 1 searches for one object from a knowledge base in which a root object located at the tip, a class object located at the middle, and an instance object located at the end are connected in a phylogenetic tree. In a knowledge base search system provided with a search means, an index setting means is provided for setting the name of an object to be searched and the name of a class object located above the object as an index of the search means, and A breadth-first search means is provided to search for higher class objects from the root object in the branching direction of the phylogenetic tree according to the index, and the object to be searched is searched from the upper class object searched by this breadth-first search means in the concatenation direction of the phylogenetic tree according to the index. By providing the depth-prioritized search means, unnecessary searches for objects lower than the upper class object of the object to be searched are not performed, so the search range is reduced and the search process is speeded up.

The invention according to claim 2 provides a search range defining means in which a plurality of object search ranges are set in advance according to predetermined conditions for the previously searched object, and a last object storage means for temporarily storing a pointer of the searched object; The last object storage means reads the search range from the search range defining means based on the object of the previous search selected from the temporarily stored pointer, and searches for the next object. Objects related to objects are searched preferentially, so it is possible to speed up the search process when sequentially searching for related objects.Furthermore, the number of times the within-range search method searches for objects can be adjusted for each search range. By providing an effective search range detecting means that performs integration, and providing a search efficiency improving means that changes the priority order of the search range output from the search range defining means to the within-range searching means according to the integrated value of the effective search range detecting means. By repeating the search process, the priority order of the search range is changed according to the worker's work style.

Embodiment An embodiment of the invention set forth in claim 1 will be described based on FIG. First, in the knowledge base search system (not shown) of this embodiment, as shown in the figure, the knowledge base 1 has a phylogenetic tree structure, and the objects A to Y are the root objects 2 (A ), class objects 3 (B-E, H to L, O, P) located in the middle, and instance objects 4 (F, G, M
,N.

Q-Y).

This knowledge base search system is provided with an index setting means for setting the name of the object 3.4 to be searched and the name of the class object 3 located above this as an index of the search means. A breadth-first search means searches for a higher class object 3 from the root object 2 in the branching direction of the phylogenetic tree according to an index set in the means, and a concatenation direction of the phylogenetic tree from the upper class object 3 searched by this breadth-first search means according to the index. A depth-priority search means for searching objects 3 and 4 to be searched is provided.

Therefore, the operation when searching for class object 3 (P) using this knowledge base search system will be described in detail based on the drawings. First, for example, the name P of the class object 3 to be searched and the name C of the upper class object 3 are set in the index setting means. Then, Figure 1 (a
), the breadth-first search means performs a breadth-first search from the root object 2 to each glass object 3 from the root object 2 to (A→to→B-C) based on the set index C, thereby finding the class object 3 ( C) is retrieved, and then, as illustrated in FIG. P
), the object 3(P) to be searched is retrieved by depth-first searching.

As mentioned above, in this knowledge base search system, the search for objects 3 and 4 is performed by performing a breadth-first search on the class object 3 located above them, and then performing a depth-first search. Since unnecessary searches for objects 3 and 4 lower than the upper class object 3 are not performed, the search range becomes smaller and the search processing is faster. Moreover, this knowledge base search system does not use a table such as a binary tree to manage or hash object names and pointers to entities, and does not require a mechanism to manage object information separately from the knowledge base structure. , the system structure is simple and the system is easy to maintain.

Next, an embodiment of the invention set forth in claim 2 will be described based on FIG. 2. This knowledge base search system (not shown)
Objects 2 to 4 searched in the previous search process are used as the last objects in the next search process, and the search range for the next objects 2 to 4 relative to the last objects 2 to 4 is determined by the search range specifying means. There are multiple settings in advance. A last object storage means for temporarily storing the pointers of the searched objects 2 to 4 is provided, and the last object storage means selects the last objects 2 to 4 from the temporarily stored pointers and selects the last objects from the search range defining means. By reading , the within-range search means searches for the next object.

Now, an example of the search range set in this search range defining means will be described in detail below based on the drawings.

First, the last object is class object 3 (f
), this last object 3(f) becomes the first search range 5, as illustrated in FIG. 2(a).

Then, as illustrated in FIGS. 2(b) to (h), the last object 3 (f) is used as the second search range 58.
) and the set of class objects 3 at the same level (H, Z
), and the last object 3 (
f) a set of lower class objects 3 (○, G),
Last object 3 (f) as the fourth search range 54
Upper class object 3 (B), fifth search range 5. Instance object 4 (N) below class object 3 (H, Z) at the same level as last object 3 (f), and class object 3 below last object 3 (f) as sixth search range 5 □
Set of instance objects 4 below (0, G) (S, U, W), seventh search range 5. As class object 3(B) above last object 3(f)
Upper root object 2 (A), search range to 5. The root object 2 (which is the upper class object 3 (B) which is the upper class object 3 (B) of the last object 3 (f))
A set of objects 3 and 4 below A) (K, E...
Y, X), etc. are set. Note that the search range 5 as described above is similarly set for each of the objects 2 to 4.

In other words, in this knowledge base search system, for example, if the last object searched in the previous search process is class object 3 (■), the search range is 5. ~
5. The next objects 2 to 4 are searched for in sequence.

In actual search work, this is related objects 2 to 4.
This is in response to the fact that searches are often performed sequentially.
Search processing is sped up by preferentially searching for objects related to the last object.

Furthermore, in this knowledge base search system, the number of searches for objects 2 to 4 in the search process as described above is accumulated for each search range by the effective search range detection means, and the search efficiency improvement means defines the search range according to this integrated value. The priority order of the search range output from the means to the within-range search means is changed.

By doing this, this knowledge base search system changes the priority of the search range according to the worker's work style when the search process is repeated, so that the search process is not repeated and unnecessary search processes are executed. The number of searches is reduced, and the search process becomes faster.

Note that this knowledge base search system also does not require a mechanism for managing object information apart from the structure of the knowledge base, so the system structure is simple and the system is easy to maintain.

Effects of the Invention The invention as claimed in claim 1 is based on a knowledge base in which a root object located at the tip, a class object located at the middle, and an instance object located at the end are connected in a phylogenetic tree. In a knowledge base search system provided with a search means for searching for one object, an index setting means is provided for setting the name of the object to be searched and the name of a class object located above the object as an index of the search means, and A breadth-first search means is provided to search for upper class objects from the root object in the branching direction of the phylogenetic tree according to the index set in the index setting means, and from the upper class objects searched by this breadth-first search means in the concatenation direction of the phylogenetic tree according to the index. By providing a depth-first search method to search for the object to be searched, unnecessary searches are not performed for objects lower than class objects higher than the object to be searched, so the search range becomes smaller and the search process becomes faster. In addition, the system does not require management or hashing of object names and pointers to entities using a table such as a binary tree, and does not require a mechanism to manage object information separately from the knowledge base structure. The structure is simple and the system is easy to maintain.

The invention according to claim 2 provides a search range defining means in which a plurality of object search ranges are set in advance according to predetermined conditions for the previously searched object, and a last object storage means for temporarily storing a pointer of the searched object; The last object storage means reads the search range from the search range defining means based on the object of the previous search selected from the temporarily stored pointer, and searches for the next object. Objects related to objects are searched preferentially, so it is possible to speed up the search process when sequentially searching for related objects.Furthermore, the number of times the within-range search method searches for objects can be adjusted for each search range. By providing an effective search range detecting means that performs integration, and providing a search efficiency improving means that changes the priority order of the search range output from the search range defining means to the within-range searching means according to the integrated value of the effective search range detecting means. When the search process is repeated, the priority order of the search range is changed according to the worker's work style, so the number of times the work is repeated and unnecessary search processes are reduced, speeding up the search process. , there is no need to manage or hash object names and pointers to entities using a table such as a binary tree, and there is no need for a mechanism to manage object information apart from the knowledge base structure, so the system structure is simple and the system This has the advantage of being easy to maintain.

[Brief explanation of drawings]

The drawings show an embodiment of the present invention, and FIG. 1 is an operation explanatory diagram showing the embodiment of the invention as claimed in claim 1, and FIG. 2 is an illustration of the search range showing the embodiment of the invention as claimed in claim 2. It is a structural explanatory diagram. 1...Knowledge base, 2...Root object, 3
...Class object, 4...Instance object, 5...Search scope Applicant Ricoh Co., Ltd.

Claims (1)

[Claims] 1. Search for one object from a knowledge base in which a root object located at the tip, a class object located at the middle, and an instance object located at the end are connected in a phylogenetic tree. In a knowledge base search system provided with a means, an index setting means is provided for setting the name of an object to be searched and the name of a class object positioned above the object as an index of the search means, and the index setting means is provided with the name of a class object located above the object. Breadth-first search means is provided to search for higher class objects from the root object in the branching direction of the phylogenetic tree according to the index, and from the upper class object searched by the breadth-first search means, the objects to be searched are searched in the concatenation direction of the phylogenetic tree according to the index. A knowledge base search system characterized by providing a depth-first search means for searching. 2. A knowledge base equipped with a search means for searching for a single object from a knowledge base in which a root object located at the tip, a class object located at the middle, and an instance object located at the end are connected in a phylogenetic tree. In the search system, there is provided a search range defining means in which a plurality of object search ranges are set in advance according to predetermined conditions for the previously searched object, and a last object storage means is provided for temporarily storing a pointer of a searched object. An in-range search means is provided for reading a search range from the search range defining means to search for the next object based on the object of the previous search selected from a pointer temporarily stored by the means, and the number of times the within-range search means searches for the object is provided. An effective search range detection means is provided for integrating the search range for each search range, and the priority order of the search range output from the search range definition means to the within-range search means is changed according to the integrated value of the effective search range detection means. A knowledge base search system characterized by providing efficiency improvement means.