JPH0279123A - Structure expression inferring system - Google Patents

Abstract

PURPOSE:To prevent a decline in the inferring ability of the title system and, at the same time, to eliminate the necessity of a large storing capacity and an increase in calculating quantity by excluding loops chained to rules used for inference without limiting the rules, and performing inference by a depth- preferential retrieval strategy. CONSTITUTION:Oriented flags expressing structure descriptions are held at a structure description holding section 101 as initial data and plural rules expressed in sets of three each are stored in an inference rule holding section 102. While inference controlling means 103 manages sequences of paired elements and natures of the oriented flags, an empty sequence appears in the initial state and whether or not an oriented flag is contained in the sequence managed by a given goal can be discriminated. Control of inference by means of the inference control means 103 is added to the basic inference by a backward depth-preferential retrieval strategy having a back track mechanism of a backward inference means 104. Therefore, necessity of a large storing capacity and an increase in calculating quantity can be eliminated without lowering the inferring capacity.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a structural representation inference method, particularly to a system that structurally represents the input/output functional specifications of a program and performs verification by inferring the properties of the representation. It concerns the structural representation inference method used.

[Conventional technology]

When performing inference using rules, a problem arises in the inference method when the chain of rules forms a loop. Conventionally, a method is known in which (1) inference is performed using only rules that do not create loops, and inference is performed using a 0 breadth-first search strategy.

[Problem to be solved by the invention]

However, the conventional method ■ limits the rules used for inference, which reduces inference ability, and the conventional method ■ requires a large amount of storage capacity due to breadth-first search, and it requires a large amount of calculation because it allows a chain of rules that form a loop. This brings about the inconvenience of increasing numbers of people.

An object of the present invention is to provide a structure representation inference method that solves this problem.

[Means to solve the problem]

According to the present invention, a structural description holding unit holds a structural description, and an inference rule holding unit holds a plurality of rules that associate inferable properties of description elements in the structural description with the structural description. an inference control means for managing an inference process of the nature of the description element of the structural description held in the structural description holding unit, and a description in the structural description for the structural description held in the structural description holding unit; A structural representation inference method characterized by comprising a backward inference means for backward inference while controlling the inference process by the inference control means using the rules held in the inference rule holding unit to determine the properties of the elements. can get.

[Effect]

A structure description in the present invention is expressed as a directed graph in which structures are notes and relationships are links.

FIG. 1 is a block diagram showing the principle of the structural representation inference method of the present invention. In the figure, 101 is a structure description holding unit;
02 is an inference rule holding unit, 103 is an inference control means, 10
4 is a backward inference means.

As initial data, the structural description holding unit 101 holds a directed graph G representing a structural description. The inference rule holding unit 102 has a plurality of rules represented by triplet<
G+, T'+, C+ > (1=1,...
, N) are stored. G1 is a digraph representing contics to which the rule can be applied, and T is the element of the digraph G+ that can be inferred by the rule and its property bare (<g
+, p+ >, g+G+), C,+ is T1
In order to infer the condition, the bare arrangement of the elements of the directed graph G and their properties ((<g+', p+'>),
g+'EG+.

j=1. ..., Ml). The inference control means 103 manages a bare series of elements of the digraph G and their properties, but in the initial state, the series is empty.

The inference control means 103 infers a certain goal <gC, pC> at a time when it attempts to infer a certain goal <gC, pC>.

In the AND-OR proof book with p〉 as the vertex, the goal -
It always maintains and manages all the goal series existing on the bus connecting <g, p> and goal <g', po>, and a given goal is included in the managed goal series. It is a means of determining whether or not.

The backward inference means 104 uses the programming language PROLO.
This system is based on inference based on a backward depth-first search strategy with a backtrack mechanism similar to that of the G processing system, and inference control by an inference control means 103 is added.

That is, inference is started with the first goal <g, p>, but when inferring a certain goal <g', p'>, the inference control means 103 is used to set the goal <g'.

It is determined whether or not p 〉 is included in the series of goals maintained and managed by the inference control means 103, and if it is included, it is determined that a loop due to a chain of rules has occurred, and the goal g d.

The loop is eliminated by assuming that the inference of p > has failed.

■The goal < g d , p d > is the inference control means 1
03, the applicable rules <G K, T K,
Regarding the rule that T matches < g d , p d > as Cx> and GK becomes a partially directed graph of G,
The sequence of goals called the condition part CK is inferred. Each time the rule is backtracked, the next candidate is selected.

repeat.

〔Example〕

FIG. 2 is an example of representation of nodes and links of a directed graph in the structure description holding unit 101. A directed graph is held in the structure description holding unit 101, and in this embodiment, it is expressed as a set of nodes and links of the directed graph.

Figure 2 (a) is a frame representation of a node, consisting of an identifier slot and a structure type slot indicating the type of structure. Figure 2 (b) is a frame representation of a link,
It consists of an identifier slot, a relationship type slot that indicates the type of relationship, a relationship source note slot that holds the identifier of the source note of the relationship, and a relationship destination node slot that holds the identifier of the destination node of the relationship.

FIG. 3 is a representation of the rules stored in the inference rule holding unit 102 used in this embodiment. A directed graph, which is a consequent for rule application, is stored in a concise slot as a set of nodes and links, and a rule for deriving a conclusion from conditions is stored in a rule slot.

Inference control means 103 and backward inference means 1.0 in this embodiment
4 uses the backtrack mechanism supported by the logic system of the programming language PROT-OG.

The flow chart shown below assumes the existence of a backtrack mechanism.

FIG. 4 shows inference control means 1 and 03 used in the embodiment of the present invention.
This is a flowchart. The input is the goal A and the goal sequence B to be inferred, and the output is the updated goal sequence C.

Step 401 is a starting step, step 40
Proceed to step 2. In step 402, goal A is goal series B.
If it is included, go to step 403; if not, go to step 4.
Proceed to 04. Step 403 fails. Inevitably, this inference control means itself also fails.

In step 404, goal A is added to goal series B to create a new goal series C, and the process proceeds to step 405. Step 405 is a step that ends with success.

FIG. 5 is a flowchart of the backward inference means 104 used in the embodiment of the present invention. The input is a goal or a sequence of goals X and a series of goals to be inferred.

Step 501 is a starting step, step 50
Proceed to step 2. In step 502, if the input X is a single goal, the process proceeds to step 506, if the input X is a line of goals, the process proceeds to step 503, and if the human power X is an empty line, the process proceeds to step 510.

In step 503, the backward inference means 1 inputs the first goal and the series of goals of the input goal sequence X.
This step is to recursively call 04. Step 5
04 is a step in which the backward inference means 104 is recursively called by inputting the remaining goal sequence and the goal series excluding the first goal of the input goal sequence X.

Step 505 is a step that ends with success.

Step 506 is a step of calling the inference control means 103 with the goal X and the goal series as input, and the output goal series is set as E. Step 507 is a step of obtaining applicable rules for goal X. This step sequentially provides applicable rules each time you backtrack. Step 508 is a step of recursively calling the present backward inference means 104 by inputting the goal sequence of the condition part and the goal series E for the obtained rule. Step 509 is a step that ends with success.

Step 510 is a step that ends with success.

〔Effect of the invention〕

As described above, the present invention is a structural representation inference method that can be used in a system that performs verification etc. by structurally expressing the input/output functional specifications of a program and inferring the properties of the representation, and restricts the rules used for inference. Since loops in the chain of rules are eliminated and inference is performed using a depth-first search strategy, there is no reduction in inference ability and there is no need for a large increase in memory capacity or calculation amount.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the principle of the structure representation inference method of the present invention, FIG. 2 is a diagram showing the representation of nodes and links in the structure description holding unit used in the embodiment of the present invention, and FIG. 3 is a block diagram showing the principle of the structure representation inference method of the present invention. FIG. 4 is a flowchart of the inference control means used in the embodiment of the present invention, and FIG. 5 is a flowchart of the inference control means used in the embodiment of the present invention. 3 is a flowchart of backward inference means. In the figure, 101 is a structure description holding section, 102 is an inference rule holding section, 103 is an inference control means, and 104 is a backward inference means.

Claims (1)

[Claims] A structural description holding unit that holds a structural description with structures and relationships between structures as descriptive elements, and a plurality of rules that associate inferable properties of descriptive elements in the structural description with the structural description. an inference rule holding unit for controlling the structure description held in the structural description holding unit; It is characterized by comprising a backward inference means that performs inference backwards while controlling the inference process with the inference control means using the rules held in the inference rule holding unit to determine the properties of the descriptive elements in the structural description. A structural representation inference method.