JPS6249437A - Inference result displaying system - Google Patents

Abstract

PURPOSE:To grasp easily a whole inference process as the relation of an object, and to explain it visually by storing knowledge as an object, deducing the knowledge by an object unit, executing the inference, storing the inference process as a relation of the object, and displaying the whole process of the inference, and the result by an object unit. CONSTITUTION:For instance, it is supposed that a user pictures an image of a cheetah in his mind, and desires to infer its name by a system. First of all, since it is clear that its object is an animal, an object name of 'animal' is inputted as an argument. An inference mechanism 2 searches an object of 'animal' from an object memory device 1, and checks whether a value of its actual example (young=an object of a subordinate concept) is stored or not. In this case, two objects of 'Mammalia' and 'carnivorous animal' are stored, therefore, first of all, 'Mammalia' of the first order is fetched. An affinitive attribute of 'Mammalia' is 'animal' only, therefore, an operation for following up a parent is not executed.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field of the Invention The present invention relates to an inference result display method for displaying an inference process and an inference result when inference is performed using knowledge stored in an electronic computer.

[Technical background of the invention and its problems]

In recent years, research has been actively conducted on expert systems that have the ability to store specialized knowledge in a computer, make inferences using this knowledge, and substitute or assist the role of an expert. This expertise is usually expressed using facts of the form ``one is true'' and rules of the form ``if A then B.''

A representative method of reasoning is the backward reasoning method.
This method starts from a hypothesis that should lead to a conclusion, sequentially applies rules backwards, and attempts to prove whether or not a known fact is reached.

For example, as a well-known example, the problem of finding a corresponding animal from among the hierarchical relationships of animals shown in FIG. 2 will be explained. In the conventional method, knowledge is expressed as a rule in the format shown in FIG. 7, and this is sequentially applied to perform C inference. In the backward reasoning method, two lists ((animal is cheetah) and (animal is)) are given as hypotheses, and in order to prove these, they are checked against the then part of the rules shown in Figure 7, and the rules that are matched are The if part of is used as the next hypothesis and C inference is proceeded. For example, I will now explain the reasoning behind proving that (animal is a cheetah). First, a match is made with the then part of Rule 4, and the inference is made using the if part of Rule 4 as a hypothesis. In other words, we are trying to prove that (animal is mammals). Next, rule 1 th
It is matched with the en part, and attempts to prove the if part (animal hos hair). Since there is no rule that has this in the then part, the user is asked "Do animals have hair?" and if the answer is true (animal is a mammal), it is proven.

In the C1 conventional method of displaying the inference results proven in this way together with the inference process, as shown in Fig.
For ``why (animal is a cheetah),'' Rule 4 applied to the proof is displayed, and for the if part of Rule 4, ``The animal is a mammal,'' ``Why is the animal a mammal?''"If there is," that is, "rhy" (animal is a mammal).
Show Rule 1 applied to the proof of. Therefore, the inference results are displayed as they are together with the rules used in the inference process.

In other words, although rules are used as display units, these rules merely indicate the process of inference.

For example, for "rwhy (animal is cheetah)", although inference is made using rules 1 and 4, in the conventional display method, rule 4 used in the previous stage of inference is written. Just do it.

The method of displaying the inference process on a rule-by-rule basis as in the conventional example described above is for debunking the rules, and has the disadvantage that it is difficult to follow the entire process of inference, making it difficult for the user to understand.

[Purpose of the invention]

An object of the present invention is to provide a method for visually explaining the entire inference process so that the user can easily understand it as a relationship between objects.

[Summary of the invention]

The present invention stores knowledge centering around objects (and their corresponding attributes and values), extracts knowledge and makes inferences for each object, and stores the inference process as a relationship between the objects. It displays the entire inference process and inference results for each object.

〔Effect of the invention〕

According to the present invention, even if the rules become complicated or the inference process becomes complicated, the entire relationship between the objects is displayed using the object as a unit, so compared to the method of displaying the rules,
This has the effect of making it easier for the user to understand the reasoning process.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to the drawings.

(Hereinafter, a target object will be referred to as an object.) FIG. 1 is a schematic configuration diagram of an embodiment apparatus. The object memory device 1 is a storage device that stores information regarding objects. The inference mechanism 2 is a part that performs inference using objects, and is a storage device that stores intermediate results of inference in a short-term memory device 3. The inference result display device 4 is a device that displays intermediate results of inference stored in a short-term memory device.

The control unit 5 is composed of the above-mentioned parts, and includes a human output unit 6
is used for user input, displaying inference results, etc.

As shown in FIG. 3, in each object memory device 1, a plurality of attributes and corresponding values are stored for each object name. For example, let's say a user envisions a teak (a yellow, spotted, carnivorous animal) and wants the system to infer its name. At first, it is clear that the object in C news is an animal, so it is called "animal".
Enter the object name as an argument. The inference mechanism 2 searches the object memory device 1 for an object called "animal" and checks whether the value of its example (child = object of a subordinate concept) is stored.

In this case, two objects, "mammalian" and "carnivorous", are stored, so first, the first "mammalian" is taken out. Since "animal" is the only parent attribute for "mammals," no action is taken to trace the parent.

Next, a question is generated based on the two necessary condition attributes C of the object "mammalians", and if the answer is YES, the inference is correct from "animals" to "mammalians", and 8TM31. Stored in units of objects (animals are mammals). Similarly, this time, using the object "mammalian" as an argument, the example attribute value of "mammalian" is checked, and the parent attribute is checked for the first value. in this case"
``teak'', so we ask the parents ``mammals'' and ``carnivores'' of cheeks whether they satisfy the required condition attributes. (However, it is STM3 that it is a “mammalian” (:
By referring to the stored statement (animals are mammals), we assume that it is correct, so we do not ask any questions. ) Since the animal is certainly a carnivore, the answer to the question "Is the animal a carnivore?" is YES, and the parent attribute "carnivore" of "cheek" is satisfied. In STM3 C, this part of the inference process is stored in units of objects (animal carnivores).

From the above, both of the requirements for "cheek" are satisfied, and the answers to the questions regarding the required attributes of "cheek", "What color is the animal?" and "Does the animal have spots?" are "yellow". If ``YES'', the required attribute is satisfied and it is determined that the item is ``cheek''.

In STM3, this inference process of ``cheek'' between ``mammalia'' and ``carnivores'' is carried out in units of objects ((
is stored as (mammal, carnivore) teak), and the inference that ``that animal is teak'' is completed.

At this time, the inference result list stored in STM3 is the list (animals, mammals) (animals, mammals) stored at each stage of inference.
carnivores) ((mammalians, carnivores) teak) by (
This is a list that expresses the relationship between objects: (animals, mammals) (animals, carnivores) ((mammals, carnivores) teak).

When the inference is completed, the inference result display device 4 refers to the STM 3 and assigns coordinates to each object in the inference result list. The problem is solved by using the example case in which it is inferred that the animal is a teak. The flowchart shown in FIG. 4 shows the specific operation of the coordinate assignment. First, set the initial value (1°1) to the first argument "animal" (Fig. 5, 11), and extract the first list (animals, mammals) from the inference result list (Fig. 5). 12). This is not an empty list, so we will calculate the coordinates. Since the coordinates (1, 1) of "animals" are on the left side, l = 2 + r = 1, and "mammals" has (2°1). (Fig. 5, 13.15) In the second list (Animals, Carnivores), "Animals" is (1,1), and "Carnivores" already has (2,1) Since it is used in "class", (Fig. 5 14) r = 2, and (2, 2)
is set. (Fig. 5 16) In the third list ((mammals, carnivores) teak), on the left side "mammals" (=(
2, 1) is set for “carnivore”, so t”3 * r=l becomes “cheek” (=
is set to (3, 1). In the manner described above, the object with coordinates as shown in FIG. 5 is displayed at the designated position. Since the left side and right side of each element in the inference result list represent the inference between the left side and the right side in the inference process,
By connecting the objects on the left side and the right side with a line, it becomes a display method that makes the relationship between each object clear as shown in FIG. 6.

It should be noted that the present invention is not limited to the above embodiments, and can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of the drawing]

FIG. 1 is a schematic configuration diagram of a device according to an embodiment of the present invention, FIG. 2 is a diagram showing the hierarchical relationship of objects, FIG. 3 is a diagram showing an example of storing knowledge in an object memory device, and FIG. 4 is an inference result display. Figure 5 is a diagram showing the correspondence between the inference result list and each element of the object (: attached coordinates).
The figure shows objects arranged based on coordinates and their relationships. Figure 7 shows an example of the description of inference rules in the conventional method. Figure 8 shows the display method of the inference process in the conventional method. It is. 1... Object memory device 2... Reasoning mechanism 3... Short-term memory device 4...
・Inference result display device 5... Control unit 6... Input/output section agent Patent attorney Nori Chika Ken Yudo Kikuo Takehana 8th diagram Figure 4

Claims (2)

[Claims] (1) An object storage means in which the object name to be inferred is stored in advance together with the corresponding attributes and values, and inference regarding the object name stored in this object storage means is performed based on the attributes and values. Inference means, short-term memory means for sequentially storing the inference process regarding the object name performed by the inference means, and inference result display means for displaying the inference process regarding the object name memorized by the short-term memory means. An inference result display method characterized by the following. (2) The inference result display means displays the inference process in relation to each other by assigning coordinates using object names as units and connecting them with line segments. Display method.