Data model generation device, data model generation system, and data model generation method

Abstract

The purpose of the present invention is to provide a technology capable of generating appropriate data models without detailed classification conditions or other domain information. Provided is a data model generation device comprising: a candidate attribute generation unit that generates a candidate attribute from a plurality of selected attribute information on the basis of evaluation values for the plurality of selected attribute information obtained by repeated selection for the selected attribute information and calculation of evaluation values for the selected attribute information; and a data model editing unit that updates a data model on the basis of label information and a candidate attribute, or on the basis of a candidate attribute.

Description

Data model generation device, data model generation system, and data model generation method

[0001] The present disclosure relates to a data model generation device, a data model generation system, and a data model generation method.

[0002] Data models (hereafter sometimes abbreviated as "models") that use graph structures to represent relationships between data have been proposed for the purposes of improving information searchability and visualizing relationships between data. These data models are used in, for example, social graphs, recommendation systems, geospatial information, and master data management.

[0003] In large-scale facilities such as water treatment and power plants, the number of plant components to be monitored is enormous, resulting in large data models for managing the data and alarms collected from the plant components. Creating, managing, and maintaining such data models manually requires a great deal of effort. Furthermore, as the scale of the model increases, important data relationships become hidden for users who view the model, compromising the visibility of the model.

[0004] Regarding the labor cost, methods have been proposed to generate models based on existing domain information such as engineering diagrams and knowledge graphs, as well as to extend existing small-scale models. Regarding model visibility, methods have been proposed to improve the visibility of large-scale models by sorting model components into predetermined categories or by linking and aggregating related elements.

[0005] For example, Patent Document 1 discloses a technology for generating a data model by applying image recognition to engineering diagrams and other engineering sources to extract plant components and associating them based on predetermined classification conditions.

[0006] Special Publication No. 2022-524642

[0007] However, in the conventional technology, since it is necessary to design the classification conditions, i.e., the domain information in detail, there is a problem that the number of steps required to design the domain information increases when, for example, a hierarchical structure is created according to multiple types of classification. On the other hand, if the domain information is not designed in detail, there is a problem that it becomes difficult to classify the information in a way that reflects the user's preferences.

[0008] Therefore, the present disclosure has been made in consideration of the above-mentioned problems, and aims to provide a technology that can generate an appropriate data model even without domain information such as detailed classification conditions.

[0009] A data model generation device according to the present disclosure includes an attribute information extraction management unit that extracts a set of attribute information from a data model of a monitoring target; a selection unit that selects a subset of the attribute information from the set of attribute information as selected attribute information; an evaluation unit that calculates an evaluation value of the selected attribute information; an attribute candidate generation unit that generates attribute candidates from a plurality of selected attribute information based on the evaluation values ​​of the plurality of selected attribute information obtained by repeatedly selecting for the selected attribute information and calculating the evaluation values ​​of the selected attribute information; and a data model editing unit that updates the data model based on the attribute candidates and label information associated with classification labels that represent the attribute candidates, or based on the attribute candidates.

[0010] According to the present disclosure, attribute candidates are generated from a plurality of selected attribute information based on evaluation values ​​of the plurality of selected attribute information obtained by repeatedly selecting the selected attribute information and calculating evaluation values ​​of the selected attribute information, and a data model is updated based on label information and the attribute candidates or based on the attribute candidates. With this configuration, an appropriate data model can be generated even without domain information such as detailed classification conditions.

[0011] The objects, features, aspects and advantages of the present disclosure will become more apparent from the following detailed description and the accompanying drawings.

[0012] 1 is a diagram showing a configuration of a data model generating device according to a first embodiment. FIG. 2 is a flowchart showing an outline of a processing procedure of the data model generating device according to the first embodiment. FIG. 3 is a diagram showing an example of attribute information extraction information and attribute information selection evaluation information according to the first embodiment. FIG. 4 is a diagram showing an example of data held by a data model information accumulation unit according to the first embodiment. FIG. 5 is a flowchart showing a processing procedure of an attribute information extraction management unit according to the first embodiment. FIG. 6 is a diagram for explaining the operation of the attribute information extraction management unit according to the first embodiment. FIG. 7 is a diagram for explaining the operation of the attribute information extraction management unit according to the first embodiment. FIG. 8 is a diagram for explaining the operation of the attribute information extraction management unit according to the first embodiment. FIG. 9 is a diagram showing an example of information managed by the attribute information extraction management unit according to the first embodiment. FIG. 10 is a flowchart showing a processing procedure of an attribute candidate generation unit according to the first embodiment. FIG. 11 is a diagram for explaining the operation of the attribute candidate generation unit according to the first embodiment. FIG. 12 is a diagram for explaining the operation of the attribute candidate generation unit according to the first embodiment. FIG. 13 is a diagram for explaining the operation of the attribute candidate generation unit according to the first embodiment. FIG. 14 is a flowchart showing a processing procedure of a selection unit according to the first embodiment. FIG. 15 is a diagram for explaining the operation of the selection unit according to the first embodiment. FIG. 16 is a flowchart showing a processing procedure of an evaluation unit according to the first embodiment. FIG. 17 is a diagram for explaining the operation of the evaluation unit according to the first embodiment. FIG. 18 is a flowchart showing a processing procedure of a data model editing unit according to the first embodiment. FIG. 19 is a diagram for explaining the operation of the data model editing unit according to the first embodiment. FIG. 1 is a diagram showing a configuration of a data model generation device according to embodiment 2. FIG. 2 is a flowchart showing a processing procedure of an attribute candidate generation unit according to embodiment 2. FIG. 3 is a diagram for explaining the operation of the attribute candidate generation unit according to embodiment 2. FIG. 4 is a flowchart showing a processing procedure of a selection unit according to embodiment 2. FIG. 1 is a diagram showing a configuration of a data model generation device according to embodiment 3. FIG. 2 is a flowchart showing a processing procedure of an attribute candidate generation unit according to embodiment 3. FIG. 3 is a flowchart showing a processing procedure of an evaluation learning unit according to embodiment 3. FIG. 4 is a diagram showing a configuration of a data model generation device according to embodiment 4. FIG. 4 is a diagram for explaining the operation of the attribute candidate generation unit according to embodiment 4. FIG. 5 is a flowchart showing a processing procedure of the evaluation learning unit according to embodiment 4.Fig. 10 is a flowchart showing the processing procedure of a selection learning unit according to embodiment 4. Fig. 11 is a diagram showing the configuration of a data model generation device according to embodiment 5. Fig. 12 is a flowchart showing the processing procedure of a selection learning unit according to embodiment 5. Fig. 13 is a block diagram showing the hardware configuration of a data model generation device according to another modified example. Fig. 14 is a block diagram showing the hardware configuration of a data model generation device according to another modified example.

[0013] <First Embodiment> Fig. 1 is a diagram showing the configuration of a data model generation device according to the first embodiment. The data model generation device in Fig. 1 includes a data model information storage unit 101, an attribute information extraction management unit 102, an extraction setting storage unit 103, an attribute candidate generation unit 104, a setting reception unit 107, and a data model editing unit 108. The attribute candidate generation unit 104 includes a selection unit 105 and an evaluation unit 106.

[0014] As described below, the data model generation device extracts a set of attribute information from the data model of the monitored object, aggregates the set of attribute information to generate attribute candidates, and updates the data model by associating the attribute candidates with classification labels that represent the attribute candidates. This makes it possible to generate a new data model that is visible and searchable, even without domain information such as detailed classification conditions.

[0015] FIG. 2 is a flowchart showing an outline of the processing procedure of the data model generating device according to the first embodiment.

[0016] In step S201, the setting receiving unit 107 receives information including attribute information extraction information and attribute information selection evaluation information from an external user and records the information in the extraction setting accumulation unit 103. As will be described in detail later, the attribute information extraction information is information for acquiring a data model and extracting a set of attribute information from the data model, and the attribute information selection evaluation information is information for selecting attribute information and evaluating the selected attribute information.

[0017] In step S202, the attribute information extraction management unit 102 acquires attribute information extraction information from the extraction setting accumulation unit 103. The attribute information extraction management unit 102 acquires the corresponding data model from the data model information accumulation unit 101 based on the designation of the data model name included in the attribute information extraction information.

[0018] In step S203, the attribute information extraction management unit 102 extracts and manages a set of attribute information from the data model acquired in step S202 based on the attribute information extraction information acquired in step S202.

[0019] In step S204 , the attribute candidate generating unit 104 acquires a set of attribute information managed by the attribute information extraction managing unit 102 , and acquires attribute information selection evaluation information from the extraction setting accumulating unit 103 .

[0020] In step S205, the attribute candidate generation unit 104 selects a subset of attribute information to be selected attribute information from the set of attribute information based on the attribute selection evaluation information acquired in step S204. That is, the attribute candidate generation unit 104 selects a subset of attribute information from the set of attribute information as selected attribute information based on the attribute selection evaluation information. Then, the attribute candidate generation unit 104 evaluates the selected attribute information based on the attribute selection evaluation information acquired in step S204, and generates attribute candidates based on the evaluation results.

[0021] In step S206, the data model editing unit 108 acquires the attribute candidates generated by the attribute candidate generation unit 104, and the attribute information extraction information and attribute information selection evaluation information held by the extraction setting accumulation unit 103. Furthermore, the data model editing unit 108 acquires a corresponding data model from the data model information accumulation unit 101 based on the designation of a data model name included in the acquired attribute information extraction information.

[0022] In step S207, the data model editing unit 108 generates new classification labels representing the attribute candidates based on the attribute information selection evaluation information or the attribute candidates acquired in step S206. Then, the data model editing unit 108 associates the generated classification labels with each piece of attribute information included in the attribute candidates acquired in step S206, and updates the data model acquired in step S206, thereby generating a new data model.

[0023] In step S208, the data model editing unit 108 saves the data model generated in step S207 in the data model information storage unit 101, or overwrites the original data model held in the data model information storage unit 101 with the data model.

[0024] The outline of the data model generation device has been explained above. Next, each component of the data model generation device will be explained in detail.

[0025] <Settings Receiving Unit> The settings receiving unit 107 receives, from the user, information on attribute information for which searchability and visibility are desired to be improved from the vast amount of data held by the data model, as extraction settings, and formats the extraction settings to store them in the extraction settings storage unit 103. As shown in Fig. 3, the extraction settings are broadly divided into two pieces of information: attribute information extraction information and attribute information selection evaluation information.

[0026] The attribute information extraction information in FIG. 3 is information for acquiring a data model and extracting a set of attribute information from the data model.

[0027] The "data model name" is the name of the data model from which attribute information is to be extracted. This specifies the data model from which attribute information is to be extracted, among the data models existing in the data model information storage unit 101.

[0028] The "attribute information target" is information indicating the criteria for data to be treated as attribute information among the data of the data model.

[0029] An "entity target" is information that indicates the criteria for treating data in a data model as an entity associated with attribute information. The entity is used to update the data model for selected attribute information.

[0030] "Classification label target" is information that indicates the criteria for data in a data model that is treated as a classification label that can be used as a viewpoint or perspective when searching the data model. A classification label is data that can be used as a viewpoint or perspective when searching the data model, and several pieces of attribute information are grouped together under one classification label.

[0031] The attribute information selection evaluation information in FIG. 3 is information for selecting a subset of attribute information from a set of attribute information as selected attribute information and for evaluating the selected attribute information.

[0032] The "selection method" is an index for determining attribute information to be preferentially selected. The attribute candidate generation unit 104 acquires the selection method and passes it to the selection unit 105, whereby the selection unit 105 uses the selection method as a guideline for selecting a subset of attribute information from a set of attribute information.

[0033] An "evaluation method" is an index for evaluating a subset of selected attribute information, i.e., selected attribute information. The attribute candidate generation unit 104 acquires the evaluation method and passes it to the evaluation unit 106, and the evaluation unit 106 uses the evaluation method as a guideline for calculating the evaluation value of the selected attribute information.

[0034] The "reference value" is a standard for the evaluation value used to determine whether the evaluation value calculated based on the "evaluation method" is high or low. When the attribute candidate generation unit 104 acquires the reference value, the attribute candidate generation unit 104 determines whether the evaluation value calculated by the evaluation unit 106 is high or low using the reference value.

[0035] A "classification label" is a classification label to be newly generated for an existing data model. The classification label is used when the data model editing unit 108 generates a new classification label for an existing data model. The classification label may also be used when the selection unit 105 selects attribute information or when the evaluation unit 106 evaluates the selected attribute information. For example, as described in the second embodiment, when "power optimization" is specified as the classification label, the selection unit 105 may actively select attribute information that includes "power" or "optimization" included in "power optimization."

[0036] "Classification label configuration information" is information associated with a "classification label" and is several pieces of information that explain the classification label in detail. The classification label configuration information may be used when the selection unit 105 selects a subset of attribute information as selected attribute information or when the evaluation unit 106 evaluates the selected attribute information. For example, in FIG. 3, "EV," "optimized," "Smart Grid," and the like are specified as classification label configuration information, so the selection unit 105 may actively select attribute information related to these.

[0037] The "aggregation upper limit number" is the upper limit of attribute information that can be selected as selected attribute information by the selection unit 105. When evaluating the selected attribute information, the aggregation upper limit number is used by the evaluation unit 106 when evaluating the selected attribute information based on the amount of violation of the constraint of the selection upper limit number, etc.

[0038] The setting reception unit 107 may be, for example, an input device that allows a user to input text or the like, or a communication device that accepts forms, applications, etc. input from a web browser. Furthermore, the setting reception unit 107 may exchange information with an external device or the like using a text or other file.

[0039] In the first embodiment described above, the attribute information selection evaluation information received by the setting receiving unit 107 includes classification labels and classification label configuration information, and therefore corresponds to label information related to the classification labels. In the first and second embodiments, the attribute information selection evaluation information may include at least one of classification labels and classification label configuration information. In this specification, for example, "at least one of A, B, C, ..., and Z" means any one of all combinations of one or more items selected from the group of A, B, C, ..., and Z.

[0040] <Extraction Setting Storage Unit> The extraction setting storage unit 103 is an information storage unit that stores information received by the setting reception unit 107, and stores information such as that shown in Fig. 3. The extraction setting storage unit 103 outputs the information it stores when it receives an output instruction from the attribute information extraction management unit 102, the attribute candidate generation unit 104, or the data model editing unit 108.

[0041] <Data Model Information Storage Unit> The data model information storage unit 101 stores multiple data models to be monitored, and manages each data model by data model name. Fig. 4 is a diagram showing an example of data stored in the data model information storage unit 101. In the example shown in Fig. 4, the data model information storage unit 101 stores data models with data model names such as "academic paper," "magazine," and "web article." The data format of each data model is not limited, and may be, for example, text, a relational database, a graph database, or the like.

[0042] <Attribute Information Extraction Management Unit> The attribute information extraction management unit 102 receives attribute information extraction information stored in the extraction setting storage unit 103 and a data model to be extracted stored in the data model information storage unit 101. The attribute information extraction management unit 102 extracts and manages a set of attribute information from the data model to be extracted based on the attribute information extraction information.

[0043] FIG. 5 is a flowchart showing the processing procedure of the attribute information extraction management unit 102 according to the first embodiment.

[0044] In step S501, the attribute information extraction management unit 102 reads the attribute information extraction information shown in FIG. 3 from the extraction setting storage unit 103, and acquires the name of the data model to be extracted from the attribute information extraction information.

[0045] In step S502, the attribute information extraction management unit 102 acquires the data model specified by the data model name acquired in step S501 from the data model information accumulation unit 101. In the attribute information extraction information in Fig. 3, "academic paper" is specified as the data model name. Therefore, when the information in Fig. 4 is held in the data model information accumulation unit 101, the attribute information extraction management unit 102 acquires a data model whose data model name is "academic paper" from the data model information accumulation unit 101, as shown in Fig. 6. The data model shown in the example of Fig. 6 is expressed in a graph database format, and has a hierarchical structure made up of nodes and edges as shown in the legend.

[0046] In step S503, the attribute information extraction management unit 102 extracts a set of attribute information from the data model acquired in step S502. For example, the attribute information extraction management unit 102 refers to the attribute information target, entity target, and classification label target included in the attribute information extraction information acquired in step S501, and assigns each element of the data model to an attribute, entity, and classification label as shown in FIG. 7. In the example shown in FIG. 7, the attribute is the terminal node of the graph, the entity is the parent node of the attribute, and the classification label is the parent node of the entity. Then, the attribute information extraction management unit 102 extracts information held by the node classified as an attribute as attribute information. FIG. 8 shows an example of a set of attribute information extracted from the name of the terminal node of the data model in FIG. 6.

[0047] In step S504, the attribute information extraction management unit 102 manages the set of attribute information extracted in step S503 as shown in Fig. 8 together with the entities related to each piece of attribute information as shown in Fig. 9. Through the above operations, the attribute information extraction management unit 102 can manage the set of attribute information and the entities related to each piece of attribute information regardless of the data format of the data model to be extracted. A memory or the like may be used for management, or a method of persisting the data in a database or the like may be used.

[0048] 1 , the attribute candidate generation unit 104 includes a selection unit 105 and an evaluation unit 106. However, the present invention is not limited to this, and the selection unit 105 and the evaluation unit 106 may be provided separately from the attribute candidate generation unit 104, rather than being included in the attribute candidate generation unit 104.

[0049] The attribute candidate generation unit 104 receives as input a set of attribute information managed by the attribute information extraction management unit 102 and attribute information selection evaluation information acquired from the extraction setting accumulation unit 103. The attribute information selection evaluation information is used as a guideline for selecting selected attribute information and a guideline for evaluating the selected attribute information. The attribute candidate generation unit 104 is capable of generating attribute candidates from multiple pieces of selected attribute information based on evaluation values ​​of multiple pieces of selected attribute information obtained by repeatedly selecting for the selected attribute information and calculating evaluation values ​​of the selected attribute information, and outputs the attribute candidates to the data model editing unit 108.

[0050] FIG. 10 is a flowchart showing the processing procedure of the attribute candidate generating unit 104 according to the first embodiment.

[0051] In step S1001, the attribute candidate generation unit 104 acquires a set of attribute information as shown in Fig. 9 managed by the attribute information extraction management unit 102. The attribute candidate generation unit 104 also acquires attribute information selection evaluation information as shown in Fig. 3 from the extraction setting accumulation unit 103.

[0052] In step S1002, the attribute candidate generation unit 104 creates and stores parameters such as those shown in Fig. 11 using the set of attribute information acquired in step S1001 and the attribute information selection evaluation information acquired in step S1001. The parameters include a "list of attribute information" in which a set of attribute information is set, excluding entities from the set of attribute information such as that shown in Fig. 9. The parameters also include initialized "selected attribute information" and "evaluation values ​​of attribute candidates" from the attribute information selection evaluation information, and "attribute information selection evaluation information" such as that shown in Fig. 3.

[0053] In step S1003, the attribute candidate generation unit 104 passes the parameters created in step S1002 to the selection unit 105, causing the selection unit 105 to select one piece of attribute information from the set of attribute information. To this end, the selection unit 105 selects one piece of attribute information in accordance with the "selection method" included in the parameters. Details of the selection process of the selection unit 105 will be described later.

[0054] In step S1004, the attribute candidate generation unit 104 acquires one piece of attribute information selected by the selection unit 105 and adds it to the "selected attribute information" included in the parameters. If the one piece of attribute information selected by the selection unit 105 already exists in the "selected attribute information," the attribute candidate generation unit 104 deletes the one piece of attribute information selected by the selection unit 105 from the "selected attribute information."

[0055] In step S1005, the attribute candidate generation unit 104 passes the parameters to the evaluation unit 106, causing the evaluation unit 106 to calculate an evaluation value for a subset of attribute information included in the "selected attribute information," i.e., an evaluation value for the selected attribute information. To this end, the evaluation unit 106 calculates the evaluation value for the selected attribute information in accordance with the "evaluation method" included in the parameters.

[0056] In step S1006, the attribute candidate generating unit 104 acquires the evaluation value calculated by the evaluating unit 106, and updates the "evaluation value of selected attribute information" included in the parameters with the acquired evaluation value.

[0057] In step S1007, the attribute candidate generation unit 104 calculates a reference value as appropriate based on the "reference value" included in the parameter. Then, the attribute candidate generation unit 104 determines whether the evaluation value acquired in step S1006 is equal to or greater than the reference value. If the evaluation value is equal to or greater than the reference value, the process proceeds to step S1008. If the evaluation value is not equal to or greater than the reference value, the process returns to step S1003.

[0058] In step S1008, the attribute candidate generation unit 104 generates attribute candidates by associating a subset of attribute information included in the parameter "selected attribute information" with the corresponding entity, and outputs the generated attribute candidates to the data model editing unit 108. For example, if the set of attribute information and the entity have a correspondence relationship as shown in FIG. 9 and the selected attribute information that is equal to or greater than the reference value is "EV," "optimization," and "Smart Grid," the attribute candidate generation unit 104 outputs attribute candidates as shown in FIG. 12.

[0059] The attribute candidate generation unit 104 repeats the processes of steps S1003 to S1007 to appropriately change the combination of attribute information included in the selected attribute information, and calculates evaluation values ​​for multiple pieces of selected attribute information over a certain span of time. Thereafter, the attribute candidate generation unit 104 performs the process of step S1008 to generate, as attribute candidates, selected attribute information that is equal to or greater than a threshold value among the multiple pieces of selected attribute information over a certain span of time.

[0060] <Selection Unit> The selection unit 105 receives parameters provided by the attribute candidate generation unit 104. The selection unit 105 selects one piece of attribute information from the set of attribute information in the "list of attribute information" included in the parameters, and outputs the selected piece of attribute information to the attribute candidate generation unit 104. The attribute candidate generation unit 104 performs the processing procedure in Fig. 10 described above, whereby the selection unit 105 selects a subset of attribute information from the set of attribute information as selected attribute information.

[0061] FIG. 13 is a flowchart showing the processing procedure of the selection unit 105 according to the first embodiment.

[0062] In step S1301, the selection unit 105 acquires parameters such as those shown in FIG.

[0063] In step S1302, the selection unit 105 selects one piece of attribute information from the set of attribute information set in the "list of attribute information" in accordance with the "selection method" included in the parameters acquired in step S1301. In the selection method, at least one piece of information, a classification label and / or classification label constituent information, is specified, and attribute information related to the specified information is preferentially selected. For example, in the parameters shown in FIG. 11 , the degree of association with the classification label constituent information is specified in the "selection method," so attribute information related to the classification label constituent information is preferentially selected.

[0064] In the example shown in Fig. 14, "EV" exists in both the set of attribute information set in the "list of attribute information" and the classification label configuration information, so when the "selection method" is specified as shown in Fig. 11, the selection unit 105 preferentially selects "EV." Although not shown, when the degree of association with the classification label is specified in the "selection method," the selection unit 105 preferentially selects attribute information associated with the classification label.

[0065] In step S1303, the selection unit 105 outputs the one piece of attribute information selected in step S1302 to the attribute candidate generation unit 104. Note that in the example shown in Fig. 14, "EV," "Optimized," and "Smart Grid" exist in both the set of attribute information set in the "List of Attribute Information" and the classification label configuration information. Therefore, when the processes of steps S1003 to S1007 in Fig. 10 are repeated, the subset of attribute information, "EV," "Optimized," and "Smart Grid," is preferentially selected as selected attribute information.

[0066] <Evaluation Unit> The evaluation unit 106 receives parameters provided by the attribute candidate generation unit 104. The evaluation unit 106 calculates an evaluation value of the “selected attribute information” included in the parameters and outputs the evaluation value to the attribute candidate generation unit 104.

[0067] FIG. 15 is a flowchart showing the processing procedure of the evaluation unit 106 according to the first embodiment.

[0068] In step S1501, the evaluation unit 106 acquires, from the attribute candidate generation unit 104, parameters that reflect the selection made by the selection unit 105.

[0069] In step S1502, the evaluation unit 106 calculates an evaluation value for the "selected attribute information" in accordance with the "evaluation method" included in the parameters acquired in step S1501. The evaluation method may be a method of calculating the evaluation value using an index such as the degree of association between the attribute information included in the selected attribute information and the attribute information selection evaluation information included in the parameters, or a method of calculating the evaluation value using a weighted sum of several indexes.

[0070] For example, in the parameters shown in Fig. 11, "degree of relevance with classification label configuration information + amount of violation of constraint on aggregation upper limit number" is specified as an index in the "evaluation method", and the evaluation unit 106 calculates the evaluation value of the selected attribute information according to this index. In the example of calculating the evaluation value shown in Fig. 16, the evaluation unit 106 calculates "degree of relevance with classification label configuration information" and "amount of violation of constraint on aggregation upper limit number" specified as indexes in the "evaluation method", and calculates the sum of these as the evaluation value.

[0071] For the former "degree of relevance with classification label configuration information," the evaluation unit 106 calculates a degree of relevance of "3" because "EV," "optimized," and "Smart Grid" specified in the classification label configuration information are included in the "selected attribute information." For the latter "amount of violations of constraint on aggregation upper limit number," the number of attribute information included in the "selected attribute information" is "3," which does not violate the aggregation upper limit number of "3." Therefore, the evaluation unit 106 calculates a constraint violation amount of "1." Although not shown, if the number of attribute information included in the "selected attribute information" violates the aggregation upper limit number, the evaluation unit 106 calculates a constraint violation amount of "0" or a negative value. Then, the evaluation unit 106 calculates the sum ("4" in the example of FIG. 16 ) of the value calculated for the former "degree of relevance with classification label configuration information" and the value calculated for the latter "amount of violations of constraint on aggregation upper limit number" as the evaluation value.

[0072] In the above description, the evaluation unit 106 calculated the evaluation value based on the degree of association corresponding to the degree of match between the selected attribute information and the classification label configuration information, but this is not limited to this. For example, the evaluation unit 106 may calculate the evaluation value based on a difference that is inversely related to the degree of association between the selected attribute information and the classification label configuration information. Also, here, the evaluation unit 106 calculated the evaluation value based on the degree of association between the selected attribute information and the classification label configuration information, but it may also calculate the evaluation value based on the degree of association between the selected attribute information and the classification label, or based on the difference between the selected attribute information and the classification label. In other words, the evaluation unit 106 may calculate the evaluation value based on the degree of association or difference between the selected attribute information and the attribute information selection evaluation information.

[0073] In step S1503, the evaluation unit 106 outputs the evaluation value calculated in step S1502 to the attribute candidate generation unit 104.

[0074] <Data Model Editing Unit> The data model editing unit 108 receives information from the data model information storage unit 101, the attribute candidate generation unit 104, and the extraction setting storage unit 103. The data model editing unit 108 updates the data structure of the data model based on the received information. In the first embodiment, the data model editing unit 108 generates new classification labels based on the attribute information selection evaluation information or the attribute candidates. The data model editing unit 108 then updates the data model by associating the generated classification labels with the attribute information included in the attribute candidates. In other words, the data model editing unit 108 updates the data model based on the attribute information selection evaluation information and the attribute candidates, or based on the attribute candidates. The data model editing unit 108 outputs the updated data model to the data model information storage unit 101.

[0075] FIG. 17 is a flowchart showing the processing procedure of the data model editing unit 108 according to the first embodiment.

[0076] In step S1701, the data model editing unit 108 acquires the attribute information extraction information and attribute information selection evaluation information that are received by the setting receiving unit 107 from the user and are held by the extraction setting accumulation unit 103.

[0077] In step S 1702 , the data model editing unit 108 acquires the data model name included in the attribute information extraction information acquired in step S 1701 , and acquires the data model specified by the data model name from the data model information accumulation unit 101 .

[0078] In step S1703 , the data model editing unit 108 acquires attribute candidates from the attribute candidate generating unit 104 .

[0079] In step S1704, the data model editing unit 108 generates a new classification label based on the attribute information selection evaluation information acquired in step S1701. For example, if the attribute information selection evaluation information includes a "classification label," the data model editing unit 108 generates a new classification label using the "classification label." For example, if the attribute information selection evaluation information does not include a "classification label," the data model editing unit 108 generates a new classification label using attribute information selection evaluation information other than the "classification label" or attribute information included in the attribute candidates. Attribute information selection evaluation information other than the "classification label" is, for example, classification label configuration information.

[0080] In step S1705, the data model editing unit 108 updates the data model by associating each piece of attribute information included in the attribute candidates with the classification label generated in step S1704 via the entities. FIG. 18 is a diagram showing an example of updating the data model. The data model shown in FIG. 18 is the graph database format data model shown in FIG. 6. The data model editing unit 108 connects the attribute information "EV," "Optimization," and "Smart Grid" included in the attribute candidates to the newly generated classification labels via edges via the entities "Document A" and "Document X" of the attribute information. As a result, as shown in FIG. 19, the data model editing unit 108 generates a new data model in which the new classification labels are associated with the attribute information included in the attribute candidates. Note that the entities and the newly generated classification labels may be in different hierarchies, such as a parent-child relationship, or may be in the same hierarchical level.

[0081] In step S 1706 , the data model editing unit 108 stores the data model generated in step S 1705 in the data model information storage unit 101 .

[0082] Summary of First Embodiment According to the data model generation device of the first embodiment, a set of attribute information is extracted from a data model, relevant attribute candidates are generated by repeatedly selecting selected attribute information and calculating evaluation values ​​for the selected attribute information, and the data model is updated based on information such as the attribute candidates. With this configuration, a hierarchical data model with visibility and searchability can be generated even without domain information such as detailed classification conditions. Note that when the data model editing unit 108 generates new classification labels based on the attribute candidates generated by the attribute candidate generation unit 104 without using attribute information selection evaluation information from an external source such as a user, a data model with unknown classification labels not anticipated by the user can be generated.

[0083] In the first embodiment, the selection unit 105 selects the selected attribute information based on the attribute information selection evaluation information including at least one of the classification label and the classification label configuration information. With this configuration, the selection unit 105 can select the selected attribute information in accordance with the user's preferences.

[0084] In the first embodiment, the evaluation unit 106 calculates an evaluation value of the selected attribute information based on the degree of association or difference between the selected attribute information and the classification label configuration information. With this configuration, the selected attribute information can be evaluated, and attribute candidates can be generated, in accordance with the user's preferences.

[0085] 20 is a diagram showing the configuration of a data model generation device according to Embodiment 2. Among the components according to Embodiment 2, components that are the same as or similar to the components described above are given the same or similar reference numerals, and different components will be mainly described below.

[0086] The configuration in FIG. 20 is the same as that in FIG. 1, except that the attribute candidate generating unit 104 and the selecting unit 105 are changed to an attribute candidate generating unit 104A and a selecting unit 105A, respectively.

[0087] As a first example, assume that the attribute information selection evaluation information received by the setting receiving unit 107 and held in the extraction setting accumulation unit 103 contains a "classification label" but is missing "classification label configuration information." In such a case, the attribute candidate generation unit 104A generates "classification label configuration information" based on the "classification label" and provides parameters including the "classification label configuration information" to the selection unit 105A. The selection unit 105A selects a subset of attribute information from the set of attribute information as selected attribute information, based on the "classification label configuration information" included in the parameters. Note that the selection unit 105A, rather than the attribute candidate generation unit 104A, may generate the "classification label configuration information" based on the "classification label."

[0088] As a second example, assume that the attribute information selection evaluation information received by the setting receiving unit 107 and held in the extraction setting storage unit 103 is missing a "classification label" and "classification label configuration information." In such a case, the selection unit 105A selects a subset of attribute information from the set of attribute information as selected attribute information in accordance with a predetermined procedure. For example, the selection unit 105A may select selected attribute information randomly or may select selected attribute information using heuristics as the predetermined procedure.

[0089] Fig. 21 is a flowchart showing the processing procedure of the attribute candidate generation unit 104A according to the embodiment 2. The processing of steps S2101 and S2103 to S2108 in Fig. 21 is similar to the processing of steps S1001 and S1003 to S1008 in Fig. 10, respectively, and therefore the processing of step S2102 will be mainly described below.

[0090] In step S2102, the attribute candidate generation unit 104A creates and stores parameters such as those shown in FIG. 11 using the set of attribute information acquired in step S2101 and the attribute information selection evaluation information acquired in step S2101. If the attribute information selection evaluation information contains a "classification label" but lacks "classification label configuration information," the attribute candidate generation unit 104A generates "classification label configuration information" using the "classification label" or other attribute information selection evaluation information. In this way, the attribute candidate generation unit 104A complements the "classification label configuration information."

[0091] FIG. 22 is a diagram showing an example of the complementation process for "classification label configuration information." For example, if "processing classification" is set as the classification label, the attribute candidate generation unit 104A generates the word "processing" or "classification" from "processing classification." The attribute candidate generation unit 104A then generates several words containing "processing" or "classification" as the parameter "classification label configuration information." The generated classification label configuration information may be in the form of a word, a regular expression, or the like.

[0092] In step S2103, the attribute candidate generation unit 104A passes the parameters created in step S2102 to the selection unit 105A, causing the selection unit 105A to select one piece of attribute information from the set of attribute information. To this end, the selection unit 105A makes the selection in accordance with the "selection method" included in the parameters.

[0093] Fig. 23 is a flowchart showing the processing procedure of the selection unit 105A according to the embodiment 2. The processing of steps S2301 and S2303 in Fig. 23 is similar to the processing of steps S1301 and S1303 in Fig. 13, respectively, and therefore the processing of step S2302 will be mainly described below.

[0094] In step S2302, the selection unit 105A selects one piece of attribute information from the set of attribute information set in the "list of attribute information" in accordance with the "selection method" included in the parameters acquired in step S2102. In the selection method, at least one piece of information, a classification label and classification label configuration information, is specified, and attribute information related to the specified information is preferentially selected. Depending on the "selection method" specified, the selection unit 105A may select one piece of attribute information randomly, or may select one piece of attribute information using heuristics or the like.

[0095] Summary of Second Embodiment According to the data model generation device of the second embodiment, the selection unit 105A makes a selection for selected attribute information based on classification label configuration information generated from classification labels. With this configuration, the selection unit 105A can make a selection for selected attribute information even if the "classification label configuration information" is missing. Furthermore, it is possible to obtain unknown classifications that the user has not anticipated, i.e., unknown combinations of attribute information.

[0096] Furthermore, in the second embodiment, the selection unit 105A performs selection for selected attribute information according to a predetermined procedure, such as random or heuristic. With this configuration, the selection unit 105A can perform selection for selected attribute information even when the "classification label" and "classification label configuration information" are missing. Furthermore, it is possible to obtain unknown classifications that the user has not anticipated, i.e., unknown combinations of attribute information. Note that when heuristics are used, it is expected that the number of trials required to generate a data model that matches the user's preferences can be reduced compared to when random selection is used.

[0097] 24 is a diagram showing the configuration of a data model generation device according to this embodiment 3. Among the components according to this embodiment 3, components that are the same as or similar to the components described above are given the same or similar reference numerals, and different components will be mainly described below.

[0098] The configuration of Figure 24 is the same as the configuration of Figure 20, except that the attribute candidate generation unit 104A, the evaluation unit 106, and the setting reception unit 107 are changed to an attribute candidate generation unit 104B, an evaluation learning unit 106B, and a setting reception unit 107B, respectively.

[0099] The setting receiving unit 107B has the functions of the setting receiving unit 107 described in the first embodiment, a function of presenting at least a part of parameters including selected attribute information to a user, and a function of receiving feedback on the selected attribute information from the user. The feedback on the selected attribute information includes, for example, designation of selected attribute information to be used as attribute candidates.

[0100] The evaluation learning unit 106B learns an evaluation value using the selected attribute information and feedback on the selected attribute information presented to the user and accepted by the setting accepting unit 107B as learning data, and calculates an evaluation value based on the selected attribute information and the learning result. The learning is, for example, machine learning (training) such as deep learning, the learning result is, for example, a trained model, and the learning data is, for example, teacher data.

[0101] Fig. 25 is a flowchart showing the processing procedure of the attribute candidate generation unit 104B according to Embodiment 3. The processing of steps S2501 to S2504 and S2506 to S2508 in Fig. 25 is similar to the processing of steps S2101 to S2104 and S2106 to S2108 in Fig. 21, respectively, and therefore the processing of step S2505 will be mainly described below.

[0102] In step S2505, the attribute candidate generation unit 104B passes the parameters to the evaluation learning unit 106B, and causes the evaluation learning unit 106B to calculate an evaluation value for a subset of attribute information included in the "selected attribute information." To this end, the evaluation learning unit 106B calculates the evaluation value as follows.

[0103] First, the evaluation learning unit 106B passes at least a portion of the parameters including the selected attribute information to the setting receiving unit 107B, and the setting receiving unit 107B presents the at least a portion to the user and receives feedback on the selected attribute information from the user.

[0104] The evaluation learning unit 106B uses the feedback for the selected attribute information received by the setting receiving unit 107B and the selected attribute information as learning data to learn an evaluation value corresponding to the relationship between the feedback and the selected attribute information. Then, the evaluation learning unit 106B calculates an evaluation value for the selected attribute information based on the selected attribute information and the learning result.

[0105] FIG. 26 is a flowchart showing the processing procedure of the evaluation learning unit 106B according to the third embodiment.

[0106] In step S2601, the evaluation learning unit 106B acquires parameters including the "selected attribute information" from the attribute candidate generating unit 104B.

[0107] In step S2602, the evaluation learning unit 106B passes at least a portion of the parameters acquired in step S2601 to the setting receiving unit 107B. At least a portion of the parameters includes "selected attribute information" and at least a portion of "attribute information selection evaluation information." At least a portion of the "attribute information selection evaluation information" is information related to the index specified in the "evaluation method" of the parameter, and includes, for example, "classification label" and "classification label configuration information."

[0108] In step S2603, evaluation learning unit 106B acquires, from setting reception unit 107B, feedback on the "selected attribute information" received from the user by setting reception unit 107B.

[0109] In step S2604, the evaluation learning unit 106B performs learning on the evaluation value of the selected attribute information using the "selected attribute information" and the feedback acquired in step S2603 as learning data. The learning in the evaluation learning unit 106B may be learning that performs weighting so that the evaluation value reflects the time at which feedback is acquired from the setting accepting unit 107B. The time at which feedback is acquired may be the time it takes for the user to input the feedback into the setting accepting unit 107B. In this case, for example, the weighting can be changed depending on whether the user made an immediate decision on the feedback or took a long time to consider it. The time at which feedback is acquired may also be the time from when the data model generation device is put into operation to when the user inputs the feedback into the setting accepting unit 107B. In this case, for example, feedback input after a long time has passed since the data model generation device was put into operation may be weighted more heavily than feedback input after a short time has passed since the data model generation device was put into operation.

[0110] Alternatively, in step S2604, the learning in the evaluation learning unit 106B may be learning in which the setting receiving unit 107B performs weighting to reflect the attributes of the user whose feedback has been received in the evaluation value. The user attributes include, for example, at least one of the user's job title, experience in the assigned job, aptitude for the assigned job, and whether or not the user is an expert. This makes it possible to weight the feedback of an experienced user more heavily than the feedback of a new user, for example.

[0111] Furthermore, the evaluation learning unit 106B may perform pre-learning using history information that pairs the "selected attribute information" with feedback for that information. This allows the evaluation learning unit 106B to appropriately calculate the evaluation value for the "selected attribute information" after learning, without any feedback from the user.

[0112] In step S2605, the evaluation learning unit 106B calculates an evaluation value for the "selected attribute information" included in at least a portion of the parameters based on at least a portion of the parameters including the "selected attribute information" and the learning results obtained in step S2604.

[0113] In step S2606, the evaluation learning unit 106B outputs the evaluation value calculated in step S2605 to the attribute candidate generating unit 104B.

[0114] Summary of Third Embodiment According to the data model generation device of the third embodiment, an evaluation value is learned using feedback on selected attribute information presented externally, such as by a user, and the selected attribute information as learning data, and an evaluation value is calculated based on the selected attribute information and the learning result. With this configuration, after learning, it is possible to evaluate selected attribute information that matches the user's preferences without any feedback from the user.

[0115] <Modification> The learning and calculation of the evaluation value by the evaluation learning unit 106B are not limited to the learning and calculation described in the third embodiment. For example, the evaluation learning unit 106B may learn the evaluation value using previously associated classification labels and attribute candidates as learning data. The evaluation learning unit 106B may then calculate the degree of correspondence between the classification labels and the selected attribute information as an evaluation value based on the classification labels included in the attribute information selection evaluation information received by the setting receiving unit 107B, the selected attribute information, and the learning results. Even with this configuration, after learning, it is possible to evaluate selected attribute information that matches the user's preferences without feedback from the user.

[0116] Alternatively, the evaluation learning unit 106B may learn evaluation values ​​using previously associated classification labels and selected attribute candidates as learning data, rather than previously associated classification labels and attribute candidates. For example, the evaluation learning unit 106B may learn evaluation values ​​by converting each piece of attribute information in the selected attribute information into a feature vector and learning a nonlinear approximation function that associates a relationship between the distance between the vectors and the level of the evaluation value.

[0117] In the third embodiment, the evaluation learning unit 106B calculates the evaluation value by learning based on the feedback on the selected attribute information presented externally, but this is not limited to this. For example, if the feedback on the selected attribute information presented externally is an evaluation value from a user, the evaluation unit 106 described in the first embodiment may use the evaluation value from the user as is.

[0118] 27 is a diagram showing the configuration of a data model generation device according to this embodiment 4. Among the components according to this embodiment 4, components that are the same as or similar to the components described above are given the same or similar reference numerals, and different components will be mainly described below.

[0119] The configuration of Figure 27 is the same as that of Figure 24, except that the attribute candidate generation unit 104B, selection unit 105A, evaluation learning unit 106B, and setting reception unit 107B are changed to an attribute candidate generation unit 104C, selection learning unit 105C, evaluation learning unit 106C, and setting reception unit 107C, respectively.

[0120] Evaluation learning unit 106C has the function of evaluation learning unit 106B according to the third embodiment and a loop function of feeding back the evaluation value of selected attribute information to selection learning unit 105C.

[0121] The setting reception unit 107C has the functions of the setting reception unit 107B described in embodiment 3, a function of presenting a monitoring and control screen represented by attribute information to a user, a function of accepting user operations on the monitoring and control screen, and a function of accepting plant-specific information including the context of the processing steps of the plant represented by attribute information.

[0122] The selective learning unit 105C performs at least one of the following first learning, second learning, and third learning. In the following, an example will be described in which the selective learning unit 105C performs all of the first learning, second learning, and third learning.

[0123] <First Learning> The selection learning unit 105C learns selection for the selected attribute information using the past selected attribute information fed back from the evaluation learning unit 106C and the evaluation value of the selected attribute information as learning data, and makes selection for the selected attribute information based on the set of attribute information and the learning result.

[0124] <Second Learning> The selection learning unit 105C learns selections for selected attribute information using the user's operation history on the monitoring and control screen as learning data, and makes selections for the selected attribute information based on the set of attribute information and the learning results.

[0125] <Third Learning> The selection learning unit 105C learns selection for selected attribute information using the plant-specific information as learning data, and makes selection for the selected attribute information based on the set of attribute information and the learning result.

[0126] After sufficient learning, the selection learning unit 105C and the evaluation learning unit 106C automatically select and evaluate selected attribute information. As a result, after learning, it is possible to select and evaluate selected attribute information that matches the user's preferences, and ultimately generate attribute candidates that match the user's preferences, without any feedback from the user.

[0127] Fig. 28 is a flowchart showing the processing procedure of the attribute candidate generation unit 104C according to Embodiment 4. The processing of steps S2801, S2802, and S2804 to S2808 in Fig. 28 is the same as the processing of steps S2501, S2502, and S2504 to S2508 in Fig. 25, and therefore the processing of step S2803 will be mainly described below.

[0128] In step S2803, the attribute candidate generation unit 104C passes the parameters created in step S2802 to the selection learning unit 105C, and causes the selection learning unit 105C to select one piece of attribute information from the set of attribute information. The attribute candidate generation unit 104C includes information specifying whether or not to learn in the parameters, which enables the selection learning unit 105C and evaluation learning unit 106C to selectively perform learning in steps S2803 and S2805.

[0129] Fig. 29 is a flowchart showing the processing procedure of evaluation learning unit 106C according to embodiment 4. The processing of steps S2901 and S2903 to S2907 in Fig. 29 is similar to the processing of steps S2601 and S2602 to S2606 in Fig. 26, respectively, and therefore the processing of step S2902 will be mainly described below.

[0130] In step S2902, evaluation learning unit 106C determines whether to execute learning based on the information specifying whether to execute learning included in the parameters acquired in step S2901. If it is determined that learning is to be executed, the process proceeds to step S2903, and if it is determined that learning is not to be executed, the process proceeds to step S2906.

[0131] After step S2906, in step S2907, the evaluation learning unit 106C outputs the evaluation value calculated in step S2906. The evaluation value output from the evaluation learning unit 106C is used to update the parameter "evaluation value of selected attribute information" in the attribute candidate generation unit 104C, and is also output to the selection learning unit 105C.

[0132] Fig. 30 is a flowchart showing the processing procedure of the selective learning unit 105C according to embodiment 4. The processing of steps S3001 and S3006 in Fig. 30 is similar to the processing of steps S2301 and S2303 in Fig. 23, respectively, and therefore the following description will mainly focus on the processing of steps S3002 to S3005.

[0133] In step S3002, the selective learning unit 105C determines whether to execute learning based on the information specifying whether to execute learning included in the parameters acquired in step S3001. If it is determined that learning is to be executed, the process proceeds to step S3003, and if it is determined that learning is not to be executed, the process proceeds to step S3005.

[0134] In step S3003, the selection learning unit 105C acquires, as learning data, the evaluation value of the "evaluation value of selected attribute information" included in the parameter as feedback from the evaluation learning unit 106C. In addition, if the setting receiving unit 107C has accepted the user's operation history on the monitoring and control screen and the plant-specific information, the selection learning unit 105C acquires the operation history and the plant-specific information from the setting receiving unit 107C as learning data.

[0135] In step S3004, the selection learning unit 105C learns about the selection for the selected attribute information using the learning data acquired in step S3003. Through this learning, the tendency of a subset of attribute information that is thought to be likely to increase the evaluation value of the "selected attribute information" is learned.

[0136] In step S3005, the selection learning unit 105C selects one piece of attribute information from the set of attribute information that is thought to be likely to increase the evaluation value, based on the set of attribute information set in the parameter ``List of attribute information'' and the learning result obtained in step S3004.

[0137] Summary of Fourth Embodiment As described above, the data model generation device according to the fourth embodiment learns selection for selected attribute information using previously selected attribute information that has been fed back and the evaluation value of that selected attribute information as learning data, and makes selection for the selected attribute information based on the set of attribute information and the learning result. With this configuration, after learning, it is possible to preferentially select attribute information that is thought to be likely to increase the evaluation value, without any feedback from the user.

[0138] In addition, in the fourth embodiment, the selection for the selected attribute information is learned using the operation history of the user on the monitoring and control screen represented by the attribute information as learning data, and the selection for the selected attribute information is made based on the set of attribute information and the learning result. With this configuration, learning is possible using the operation history, which is a type of domain information, so that know-how in operating the monitoring and control system can be reflected in the selection for the selected attribute information.

[0139] In addition, in the fourth embodiment, the plant-specific information represented by the attribute information is used as learning data to learn the selection for the selected attribute information, and the selection for the selected attribute information is made based on the set of attribute information and the learning result. With this configuration, learning can be performed using the plant-specific information, which is a type of domain information, so that know-how in operating the monitoring and control system can be reflected in the selection for the selected attribute information.

[0140] <Modification> In the fourth embodiment, the selection learning unit 105C automatically performs selection for selected attribute information, but this is not limited to this. For example, similar to the above-described <first learning unit>, the selection learning unit 105C first learns selection for selected attribute information using past selected attribute information and the evaluation value of the selected attribute information as learning data, and selects candidates for selected attribute information based on the set of attribute information and the learning result.

[0141] The setting receiving unit 107C may then present the candidates for selected attribute information selected by the selection learning unit 105C to the user and receive feedback on the candidates for selected attribute information from the user. The selection learning unit 105C may then select selected attribute information based on the feedback on the candidates for selected attribute information received by the setting receiving unit 107C. With this configuration, selection of selected attribute information that matches the user's preferences can be performed semi-automatically.

[0142] 31 is a diagram showing the configuration of a data model generation device according to this embodiment 5. Among the components according to this embodiment 5, components that are the same as or similar to the components described above are given the same or similar reference numerals, and different components will be mainly described below.

[0143] The configuration of FIG. 31 is the same as the configuration of FIG. 27, except that selective learning unit 105C is changed to selective learning unit 105D.

[0144] The selection learning unit 105D has the functions of the selection units 105 and 105A according to the first and second embodiments, and the functions of the selection learning unit 105C according to the third embodiment. The selection learning unit 105D also has the function of making selections for selected attribute information, taking into consideration an increase or decrease in the evaluation value of the newly generated "selected attribute information." In other words, the selection learning unit 105D is able to preferentially select attribute information so as not to lower the evaluation value of the "selected attribute information."

[0145] When the selection learning unit 105D selects attribute information taking into consideration an increase or decrease in the evaluation value of the "selected attribute information," the evaluation learning unit 106C does not execute learning, i.e., proceeds from the processing of step S2902 to the processing of step S2906 in Fig. 29, and calculates an evaluation value using the learning result. Then, in step S2907, the evaluation learning unit 106C outputs the evaluation value calculated in step S2906 to the selection learning unit 105D, not to the attribute candidate generation unit 104C.

[0146] The selection learning unit 105D receives as input the parameters provided by the attribute candidate generation unit 104C and the evaluation value fed back from the evaluation learning unit 106C. Based on the parameters and the evaluation value, the selection learning unit 105D selects one piece of attribute information from the "list of attribute information" included in the parameters. As in the fourth embodiment, when the setting reception unit 107C acquires domain information such as a user's operation history or plant-specific information, the domain information may be input to the selection learning unit 105D.

[0147] Fig. 32 is a flowchart showing the processing procedure of the selective learning unit 105D according to embodiment 5. The processing of steps S3201 to S3204, S3207, and S3208 in Fig. 32 is similar to the processing of steps S3001 to S3004, S3005, and S3006 in Fig. 30, respectively, and therefore the processing of steps S3205 and S3206 will be mainly described below.

[0148] In step S3205, the selection learning unit 105D refers to the parameters acquired in step S3201 and determines whether or not to perform collaborative processing to select one piece of attribute information in collaboration with the evaluation learning unit 106C. Collaborative processing may be determined with a certain probability. If it is determined that collaborative processing will be performed, the process proceeds to step S3206; if it is determined that collaborative processing will not be performed, the process proceeds to step S3207.

[0149] In step S3206, the selection learning unit 105D selects several pieces of attribute information from the set of attribute information included in the "list of attribute information" parameter and inputs the selected pieces of attribute information to the evaluation learning unit 106C. Note that the selection learning unit 105D may select several pieces of attribute information using heuristics or the like. The evaluation learning unit 106C calculates evaluation values ​​for the selected attribute information using several pieces of attribute information as selected attribute information without performing learning in the evaluation learning unit 106C. Then, the selection learning unit 105D selects one piece of attribute information based on the calculated evaluation value. For example, the selection learning unit 105D checks whether the evaluation value of the selected attribute information has increased or decreased, and selects one piece of attribute information that maximizes the evaluation value of the selected attribute information.

[0150] Summary of Embodiment 5 In learning such as that in Embodiment 4, there is a tendency for one of multiple pieces of selected attribute information that are similar to each other within a certain range to be selected. Therefore, if the evaluation value of selected attribute information outside the certain range is the largest, appropriate selected attribute information may not be obtained. In contrast, in Embodiment 5, the selection learning unit 105D appropriately uses functions similar to those of the selection units 105 and 105A, and therefore can use selected attribute information outside the certain range, i.e., selected attribute information that cannot be obtained by learning, when that selected attribute information is appropriate. In other words, the selection learning unit 105D can use selected attribute information such as mutations obtained by a genetic algorithm, thereby increasing the possibility of using truly optimal selected attribute information.

[0151] 1 described above will be referred to as the "attribute information extraction management unit 102, etc." The attribute information extraction management unit 102, etc. are realized by a processing circuit 81 shown in FIG. 33. That is, the processing circuit 81 includes the attribute information extraction management unit 102 that extracts a set of attribute information from a data model of a monitoring target, a selection unit 105 that selects a subset of attribute information from the set of attribute information as selected attribute information, an evaluation unit 106 that calculates an evaluation value of the selected attribute information, an attribute candidate generation unit 104 that generates attribute candidates from a plurality of selected attribute information based on evaluation values ​​of the plurality of selected attribute information obtained by repeatedly selecting for the selected attribute information and calculating the evaluation values ​​of the selected attribute information, and a data model editing unit 108 that updates the data model based on the attribute candidates and label information associated with classification labels that represent the attribute candidates, or based on the attribute candidates. Dedicated hardware or a processor that executes a program stored in memory may be applied to the processing circuit 81. Examples of the processor include a central processing unit, a processing unit, an arithmetic unit, a microprocessor, a microcomputer, and a DSP (Digital Signal Processor).

[0152] When the processing circuitry 81 is dedicated hardware, the processing circuitry 81 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination of these. The functions of each unit such as the attribute information extraction management unit 102 may be realized by a circuit in which processing circuits are distributed, or the functions of each unit may be realized together by a single processing circuit.

[0153] When the processing circuit 81 is a processor, the functions of the attribute information extraction management unit 102 and the like are realized in combination with software or the like. The software or the like includes, for example, software, firmware, or both software and firmware. The software or the like is written as a program and stored in a memory. As shown in FIG. 34 , the processor 82 applied to the processing circuit 81 realizes the functions of each unit by reading and executing a program stored in a memory 83. That is, the data model generation device includes a memory 83 for storing a program that, when executed by the processing circuit 81, results in the following steps: extracting a set of attribute information from a data model of a monitoring target; selecting a subset of attribute information from the set of attribute information as selected attribute information; calculating an evaluation value for the selected attribute information; generating attribute candidates from a plurality of selected attribute information based on evaluation values ​​for the plurality of selected attribute information obtained by repeatedly selecting the selected attribute information and calculating the evaluation value for the selected attribute information; and updating the data model based on the attribute candidates and label information associated with classification labels representing the attribute candidates, or based on the attribute candidates. In other words, this program can be said to cause a computer to execute the procedures and methods of the attribute information extraction management unit 102, etc. Here, the memory 83 may be, for example, a non-volatile or volatile semiconductor memory such as a random access memory (RAM), a read-only memory (ROM), a flash memory, an erasable programmable read-only memory (EPROM), or an electrically erasable programmable read-only memory (EEPROM), a hard disk drive (HDD), a magnetic disk, a flexible disk, an optical disk, a compact disk, a minidisk, a digital versatile disk (DVD), a drive device for any of these, or any storage medium that will be used in the future.

[0154] The above describes a configuration in which each function of the attribute information extraction management unit 102 and the like is realized either by hardware or software, etc. However, the present invention is not limited to this, and a configuration in which part of the attribute information extraction management unit 102 and the like is realized by dedicated hardware and another part is realized by software, etc. For example, the function of the attribute information extraction management unit 102 can be realized by a processing circuit 81 as dedicated hardware, and the other functions can be realized by the processing circuit 81 as a processor 82 reading and executing programs stored in a memory 83.

[0155] As described above, the processing circuitry 81 can realize the above-described functions by hardware, software, or a combination of these.

[0156] The data model generation device may be configured as a single device or as a combination of multiple devices. The elements of the data model generation device may be configured by a computer executing a program, or may be configured as hardware that does not execute a program. In other words, processing may be performed by software, or signal processing may be performed by hardware.

[0157] Furthermore, the data model generation device described above can also be applied to a data model generation system that includes a data model generation unit that is a data model generation device other than setting reception units 107, 107B, and 107C, and a setting reception unit that has the same functions as setting reception units 107, 107B, and 107C and is capable of communicating with the data model generation unit via the Internet or the like.

[0158] It should be noted that the embodiments and modifications may be freely combined, and the embodiments and modifications may be modified or omitted as appropriate.

[0159] The above description is illustrative in all respects and is not restrictive. It is understood that countless variations not illustrated can be envisioned.

[0160] 102 attribute information extraction management unit, 104, 104A, 104B, 104C attribute candidate generation unit, 105, 105A selection unit, 105C, 105D selection learning unit, 106 evaluation unit, 106B, 106C evaluation learning unit, 107, 107B, 107C setting reception unit, 108 data model editing unit.

Claims

1. A data model generation device comprising: an attribute information extraction management unit that extracts a set of attribute information from a data model of a monitored object; a selection unit that selects a subset of the attribute information from the set of attribute information as selected attribute information; an evaluation unit that calculates an evaluation value of the selected attribute information; an attribute candidate generation unit that generates attribute candidates from a plurality of selected attribute information based on the evaluation values ​​of the plurality of selected attribute information obtained by repeating selection for the selected attribute information and calculation of the evaluation value of the selected attribute information; and a data model editing unit that updates the data model based on the attribute candidates and label information related to a classification label that represents the attribute candidates.

2. A data model generating device as described in claim 1, further comprising a setting receiving unit that receives, from an external device, the label information including at least one of the classification label and classification label configuration information associated with the classification label, and the selection unit makes a selection for the selected attribute information based on the received label information.

3. A data model generating device as described in claim 1, further comprising a setting receiving unit that receives the label information including the classification label from outside, wherein classification label configuration information associated with the classification label is generated based on the classification label included in the received label information, and the selection unit makes a selection for the selected attribute information based on the classification label configuration information.

4. A data model generating device according to claim 1, wherein the selection unit makes a selection for the selected attribute information in accordance with a predetermined procedure.

5. A data model generating device as described in claim 1, further comprising a setting receiving unit, wherein the selection unit is a selection learning unit that learns selections for the selected attribute information using the past selected attribute information and the evaluation value of the selected attribute information as learning data, selects candidates for the selected attribute information based on the set of attribute information and learning results, and makes selections for the selected attribute information based on feedback received by the setting receiving unit for the candidates for selected attribute information presented externally.

6. A data model generating device as described in claim 1, wherein the selection unit is a selection learning unit that learns selections for the selected attribute information using the past selected attribute information and the evaluation value of the selected attribute information as learning data, and makes selections for the selected attribute information based on the set of attribute information and the learning results.

7. A data model generating device as described in claim 1, further comprising a setting receiving unit that receives a user's operation history on a monitoring control screen represented by the attribute information, wherein the selection unit is a selection learning unit that uses the operation history as learning data to learn a selection for the selected attribute information and makes a selection for the selected attribute information based on the set of attribute information and a learning result.

8. A data model generation device as described in claim 1, further comprising a setting receiving unit that receives plant-specific information including a sequence of processing steps of the plant represented by the attribute information, wherein the selection unit is a selection learning unit that uses the plant-specific information as learning data to learn a selection for the selected attribute information and makes a selection for the selected attribute information based on the set of attribute information and a learning result.

9. A data model generating device as described in claim 1, further comprising a setting receiving unit that receives, from an external device, the label information including at least one of the classification label and classification label configuration information associated with the classification label, and the evaluation unit calculates the evaluation value based on the relevance or difference between the selected attribute information and the label information.

10. A data model generating device as described in claim 1, further comprising a setting receiving unit that receives feedback from the outside regarding the selected attribute information presented to the outside, and the evaluation unit calculates the evaluation value based on the feedback.

11. A data model generation device as claimed in claim 1, further comprising a setting reception unit that receives the label information including the classification label from outside, wherein the evaluation unit is an evaluation learning unit that learns the evaluation value using the classification label and the attribute candidate that have been previously associated, or the classification label and the selected attribute information that have been previously associated, as learning data, and calculates the evaluation value based on the classification label included in the received label information, the selected attribute information, and a learning result.

12. A data model generation device as described in claim 10, wherein the evaluation unit is an evaluation learning unit that learns the evaluation value using the feedback and the selected attribute information as learning data, calculates the evaluation value based on the selected attribute information and a learning result, and performs weighting on the evaluation value to reflect the time at which the feedback was obtained from the setting receiving unit.

13. A data model generating device as described in claim 10, wherein the evaluation unit is an evaluation learning unit that learns the evaluation value using the feedback and the selected attribute information as learning data and calculates the evaluation value based on the selected attribute information and a learning result, and the setting receiving unit weights the evaluation value to reflect the attributes of the external user that received the feedback.

14. A data model generating device as claimed in any one of claims 1 to 13, wherein the data model editing unit generates a new classification label based on the label information or the attribute candidate, and updates the data model by associating the generated classification label with the attribute information included in the attribute candidate.

15. A data model generating device as claimed in any one of claims 1 to 14, further comprising: a data model information storage unit that holds the data model used in the attribute information extraction management unit and the data model updated in the data model editing unit; and an extraction setting storage unit that holds the label information used in the data model editing unit.

16. A data model generation system comprising: a data model generation unit; and a setting reception unit capable of communicating with the data model generation unit via the Internet, wherein the data model generation unit includes: an attribute information extraction management unit that extracts a set of attribute information from a data model of a monitoring target; a selection unit that selects a subset of the attribute information from the set of attribute information as selected attribute information; an evaluation unit that calculates an evaluation value of the selected attribute information; an attribute candidate generation unit that generates attribute candidates from the multiple selected attribute information based on the evaluation values ​​of the multiple selected attribute information obtained by repeating selection for the selected attribute information and calculation of the evaluation value of the selected attribute information; and a data model editing unit that updates the data model based on label information associated with a classification label representative of the attribute candidates and the attribute candidates or based on the attribute candidates, and the data model generation unit receives the label information.

17. A data model generation method comprising: extracting a set of attribute information from a data model of a monitored object; selecting a subset of the attribute information from the set of attribute information as selected attribute information; calculating an evaluation value of the selected attribute information; generating attribute candidates from the multiple selected attribute information based on the evaluation values ​​of the multiple selected attribute information obtained by repeating selection for the selected attribute information and calculation of the evaluation value of the selected attribute information; and updating the data model based on the attribute candidates and label information related to a classification label representing the attribute candidates, or based on the attribute candidates.

Images