JP2847369B2 - Expert model construction equipment - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] The present invention relates to an expert model construction apparatus for determining an expert model for a conceptualized problem, and a rule set and a hierarchical data structure for the expert model for solving a problem prepared in advance as a knowledge base. , And a set of hypothesis sets, etc., with the aim of making it easy for even inexperienced technicians to construct expert models by modifying and adding them while interacting. Preliminarily prepare at least one or more of a set of rules such as a set, a generalized hierarchical data structure related to an expert model, and a set of hypotheses that are sets such as intermediate results generated in an execution process with respect to the expert model, For the input conceptualized problem, the corresponding rule from the above rule set, the corresponding data structure from the hierarchical data structure, or Is an expert model decision section that proposes an expert model proposal consisting of the hypothesis set from the hypothesis set, prompts input and selection of problem-specific structures as necessary, and determines the expert model framework. An editing unit for editing the framework of the expert model determined by the model determining unit by changing or adding a problem-specific structure while further talking, and generating a desired expert model by the editing unit It is configured to obtain.

[Industrial applications]

The present invention relates to an expert model construction device for determining an expert model for a conceptualized problem.

In recent years, the application of expert systems to industry has been remarkable, and has been applied in a very wide range of fields. At this time, when constructing an expert system in the framework of a knowledge-based system, specialized know-how on construction procedures and knowledge-based structural design is required, and it is desired that even inexperienced engineers can easily construct the system. I have.

[Problems to be solved by conventional technology and invention]

Conventionally, an expert system collects rules by trial and error for an object to be constructed (for example, a layout design of an apartment) and defines the rules one by one according to a knowledge expression framework defined by an expert shell. Has been taken. Therefore, when an inexperienced engineer constructs an expert system, there is a problem that it is difficult to efficiently construct an expert model for solving the problem.

The present invention proposes a rule set, a hierarchical data structure, a hypothesis set, and the like relating to an expert model for solving a problem prepared in advance as a knowledge base, and inexperienced by modifying and adding it while interacting. The aim is to make it easy for engineers to build expert models.

[Means to solve the problem]

 Means for solving the problem will be described with reference to FIG.

In FIG. 1, an expert model determination unit 1 presents a corresponding expert model draft among a set of rules, a hierarchical data structure, and a set of hypotheses prepared in advance, and inputs / selects a problem-specific structure as necessary. This will promote the expert model framework.

The editing unit 2 edits the framework of the determined expert model by changing or adding a problem-specific structure while talking.

The rule set is composed of generation rules and inspection rules for solving a problem.

The hierarchical data structure is a generalized data structure that expresses an object to be considered.

The hypothesis set is an intermediate result or the like generated in the execution process to solve the problem.

[Action]

As shown in FIG. 1, according to the present invention, an expert model determination unit 1 determines an expert model proposal corresponding to an input conceptualized problem from a rule set, a hierarchical data structure, and a hypothesis set. And prompting the input / selection of the problem-specific structure as necessary to determine the framework of the expert model, and the editorial unit 2 changes and adds the problem-specific structure while further talking to create the expert model. It is generated.

Therefore, even an inexperienced engineer can easily construct a user-specific expert model by inputting, selecting, changing, and adding a problem-specific structure to an expert model proposal presented by the system while interacting. Becomes possible.

〔Example〕

Next, the configuration and operation of one embodiment of the present invention will be sequentially described in detail with reference to FIGS.

In FIG. 1, an expert model decision unit 1 decides a framework of an expert model for solving a problem. The expert model decision unit 1 decides a problem solving process, a target model, and a hypothesis set. Do one of them.

The decision of the problem solving process is to present the one with the corresponding pattern from the rule set prepared by the system in advance,
The framework of the expert model is determined by inputting and selecting as necessary. (Described later with reference to FIG. 2).
Here, the rule set describes the process of solving the problem as “IF
~ THEN "format and the like (for example, the generation rule set shown in Fig. 2A) and the inspection rule set for inspecting the process of solving the problem (for example, the inspection shown in Fig. 2A) Rule set).

The determination of the target model is to determine the framework of the expert model by selecting a model having a corresponding pattern from a hierarchical data structure prepared in advance by the system (described later with reference to FIG. 3). Where the hierarchical data structure is
The object (for example, a floor plan of an apartment) to be solved is expressed in a hierarchical structure.

The determination of the hypothesis set is to determine the framework of the expert model by selecting an appropriate result from the intermediate results generated in the execution process prepared in advance by the system (described later with reference to FIG. 4). .

The editing unit 2 presents the framework of the expert model determined by the expert model determining unit 1 and edits it into a user-specific expert model that matches the user's intention by changing / adding it while talking. Things.

 Next, the operation of the configuration in FIG. 1 will be described.

In FIG. 1, means that the expert model is determined. The determination of the expert model means that the problem solving process is determined by and the target model is determined by, or the set of hypotheses is determined by and.
This will be specifically described later with reference to FIGS. 2 to 4.

Performs structure presentation from the system. This means that the framework (structure) of the expert model determined by or is displayed on the screen.

Performs editing with the editor. This means that the framework of the expert model displayed on the screen is changed / added while having a conversation according to the intention of the user. If YES (OK), refine the expert model. If NO, repeat.

Prompts the user to edit the parts required for detailing. For example, as shown on the right side, this means that a menu is displayed on the screen, "Define an arrangement element."

Performs problem-specific data definitions. This means that "Western room data [size (34), etc.] is input and defined, as shown on the right side corresponding to the menu display.

Means that the data structure is reorganized focusing on the common data. This means that the data of the overlapping part of the data related to the expert model that has been detailed in the processing up to this point is collected (to be described later with reference to FIG. 5).

As described above, when constructing an expert model for solving a conceptualized problem, the system presents a proposal (pattern) from among a set of rules, a hierarchical data structure, and a set of hypotheses prepared in advance. By modifying / changing / adding this while talking, the structure of the expert model is determined, and further speaking, the expert model is determined by refinement, so that even inexperienced engineers can follow the presentation of the system. An expert system can be easily constructed.

The determination of the problem solving process will be described with reference to FIG. FIG. 2 (a) shows an example of a rule, FIG. 2 (b) shows an example of a constraint, and FIG. 2 (c) shows a schematic representation of FIG. 2 (b).

In FIG. 2 (a), the generation rule is a rule describing a process of solving a problem (for example, determining a layout of an apartment). If all the conditions are satisfied, the arrangement is allowed. " Here, as shown in the absolute condition in “Western room” in FIG.
It is a condition of "touching the corner".

The inspection rule is a rule for inspecting a process of solving a problem (for example, deciding a layout of an apartment), and is a rule of “Is the arrangement completed? → (if YES) registered” as shown in the figure.

Fig. 2 (c) is Fig. 2 (b) "Size" of "Western room"
Is (34) and “position” is (31) schematically by 2
It is a dimensionally represented one.

As described above, as a problem solving process of the input problem, for example, in response to the presentation of the generation rule and the inspection rule proposal, the necessary specialty is changed or added as necessary, for example. The house model can be easily determined.

 The determination of the target model will be described with reference to FIG.

In FIG. 3, a generalized data structure from the problem classification knowledge base is proposed. This means that, as shown on the right side, a generalized hierarchical data structure expressing the object to be considered (for example, the layout of an apartment) is proposed. In this proposal, a type (element) of a hierarchical data structure including a keyword (for example, an apartment) of a sentence input in Japanese is searched for and found by searching a problem classification knowledge base created in advance.

Presents and confirms the word extracted in Japanese and replaced with the concept. This is, as shown on the right,
(System): In response to the message “Is the name of the placement space an apartment?”, In the case of an affirmative response “Yes”, a data structure specific to the problem of the placement space, as shown by the arrow in the drawing. As "condominium".

Is re-entered in the case of NO, ie, if it is not appropriate.

Prompts for information required by each element. As shown on the right, S: "What is the fixed part?" Is displayed and U: "Entrance, there is a window" The “window” is extracted and associated with the fixed part as a problem-specific data structure as indicated by an arrow.

Performs editing with an editor. This means that, as shown on the right side, the user further adds a “bright window” as a data structure specific to the problem to the fixed part. And
The processing shown in FIG.

As described above, by presenting a generalized hierarchical data structure extracted from the problem classification knowledge base created in advance, and editing it into a problem-specific data structure while talking,
Even inexperienced technicians can easily determine the expert model.

The determination of the hypothesis set will be described with reference to FIG. This hypothesis set is an intermediate result or the like generated in the execution process. For example, as shown in FIG. 4 (a), the arrangement state (management information of what and where exists in the arrangement process),
And placement plans (individual plans showing what and where they exist). Here, as shown in FIG. 4 (b), in the case of a condominium layout,
For example, it shows a state of which arrangement plan is “Western room”. The layout plan is a plan for elements (for example, Western-style rooms) and positions, as shown in FIG. in this way,
The arrangement state and the arrangement plan of the hypothesis set are dynamically generated so as to satisfy the constraint condition at the time of execution, and a target solution can be derived.

FIG. 5 is an explanatory diagram of reorganization of common data. This explains the operation of FIG.

In FIG. 5, in a state where the data structure of the target model is displayed on the screen, re-editing is started by clicking re-edit with the mouse, and the following processing is automatically executed. The “size (21)” of the elements “push-in” and “box” having common data in the data structure are searched for and found, and these common data “size (2)
1) "is newly stored as one data structure (named as a storage space) and the original common data is deleted. Thereby, duplicate data is aggregated, and the memory capacity of the data structure of the expert model is reduced. .

〔The invention's effect〕

As described above, according to the present invention, the expert model determination unit 1 determines the expert model corresponding to the input conceptualized problem from among a rule set, a hierarchical data structure, and a hypothesis set. It proposes a proposal, prompts the input and selection of the problem-specific structure as necessary, determines the framework of the expert model, and further edits and modifies the problem-specific structure while the editorial unit 2 further talks. , It is easy for even inexperienced technicians to input, select, change, or add problem-specific structures to the expert model proposal presented by the system while interacting. A unique expert model can be built. This makes the problem-dependent expert model (program parts)
By reusing, the part that the developer must newly create is dramatically reduced, and an expert system can be efficiently constructed.

[Brief description of the drawings]

FIG. 1 is a block diagram of one embodiment of the present invention, FIG. 2 is a diagram for explaining the problem solving process, FIG. 3 is a diagram for explaining the structure of the target model, FIG. 4 is a diagram for explaining the hypothesis set, FIG. The figure shows a re-editing diagram of the common data. In the figure, an expert model determination unit, an editing unit, a rule set, a hierarchical data structure, and a hypothesis set are shown.

Continuation of front page (58) Field surveyed (Int.Cl. ⁶ , DB name) G06F 9/44

Claims (1)

(57) [Claims] An expert model construction apparatus for constructing an expert model for an input conceptualized problem, comprising: a hierarchical data structure storage unit for storing a generalized hierarchical data structure for an expert model; In addition to presenting the data structure corresponding to the input conceptualized problem among the hierarchical data structures stored in the hierarchical data structure storage unit, the presented data is used to embody the generalized hierarchical data structure. A target model determining means for prompting input of data of a target model to be associated with the data of the lowest layer of the structure, and determining the target model by associating the input data accordingly; and determining by the target model determining means Editing means for constructing the final target model by changing the data structure of the target model Construction equipment.