JP6639749B1 - Search device, search method, and machine learning device - Google Patents

Abstract

A retrieval device (1) for retrieving specification data similar to a retrieval condition from a specification data group including at least one specification item value indicating details of an item indicating a product specification, A similarity learning unit that learns, for all combinations of specification item values, how similar two specification item values of the same item are based on the specification data group held in the product design database (13). (14), an input unit (12) for receiving an input of a specification item value to be used as a search condition, and specification data similar to the input specification data constituted by the specification item value received by the input unit (12). A similarity specification search unit (16) for searching based on the learning result by the similarity learning unit (14).

Description

  The present invention relates to a search device, a search method, and a machine learning device for searching, from data on past product specifications, data on specifications similar to the specifications indicated by input data.

  At the design site of a product with a wide variety of customer requirements, when designing a new product, by reusing part of the data of past product design cases with similar specifications, the efficiency of design work and design can be improved. The validation of the validity is simplified. However, in order to efficiently search the data of design examples of past products having similar specifications, deep knowledge of the product is required, and this largely depends on the experience of the designer.

  In view of this, the invention described in Patent Document 1 has been proposed as a technique for efficiently retrieving past design data of a product without depending on the experience of a designer. The invention described in Patent Literature 1 should be applied using a design case base storing past design case data, a knowledge base storing knowledge for determining a design specification to be applied from a required specification, and a knowledge base. An inference engine for determining design specifications is provided, and similar design cases are searched for in consideration of design constraints and interference relations between specification items constituting required specifications.

JP-A-8-222737

  However, in the invention described in Patent Literature 1, it is necessary to manually describe in a knowledge base a rule for determining a design specification to be applied from a required specification as a rule, which is extremely time-consuming.

  The present invention has been made in view of the above, and an object of the present invention is to provide a search device capable of reducing a user's work load.

In order to solve the above-described problem and achieve the object, the present invention provides a method for searching a specification data group including at least one specification item value indicating details of an item indicating a product specification. This is a search device for searching for specification data similar to conditions. The search device learns, for all combinations of specification item values, how similar two specification item values of the same item are based on the specification data group held in the product design database. And an input unit for receiving an input of a specification item value used as a search condition. Further, the search device includes a similar specification search unit that searches for specification data similar to input specification data composed of specification item values received by the input unit based on a learning result by the similarity learning unit. The similarity learning unit calculates a similarity between item values indicating how similar two specification item values are based on the specification data and data of parts constituting each product corresponding to each of the specification data. Learning is performed by calculating, and the similar specification search unit performs a search based on the similarity between item values.

  The search device according to the present invention has an effect that the work load on the user can be reduced.

FIG. 2 is a diagram illustrating a configuration example of a search device according to the first embodiment; FIG. 2 is a diagram illustrating an example of a hardware configuration that implements a search device according to the first embodiment. FIG. 4 is a diagram showing another example of a hardware configuration for realizing the search device according to the first embodiment. FIG. 3 is a diagram illustrating an example of a detailed configuration of a similarity learning unit included in the search device according to the first embodiment; FIG. 3 is a diagram showing a configuration example of a product design database provided in the search device according to the first embodiment; FIG. 3 is a diagram showing an example of a specification list held by a product design database according to the first embodiment; FIG. 3 is a diagram showing an example of a component configuration table held by a product design database according to the first embodiment; 5 is a flowchart illustrating an example of the operation of the similarity learning unit included in the search device according to the first embodiment; FIG. 4 is a diagram illustrating a concept of a method of calculating a similarity between item values in a similarity learning unit included in the search device according to the first embodiment; FIG. 3 is a diagram illustrating an example of similarity data generated by a similarity learning unit included in the search device according to the first embodiment; FIG. 9 is a diagram showing another example of the similarity data generated by the similarity learning unit included in the search device according to the first embodiment. 5 is a flowchart illustrating an example of a search operation performed by the search device according to the first embodiment. FIG. 4 is a diagram showing an example of a specification input screen displayed by the search device according to the first embodiment. FIG. 3 is a diagram illustrating an example of a detailed configuration of a similar specification search unit included in the search device according to the first embodiment; FIG. 3 is a diagram illustrating an example of similarity weight data used by a similar specification search unit of the search device according to the first embodiment; 5 is a flowchart illustrating an example of the operation of the similar specification search unit included in the search device according to the first embodiment; FIG. 6 is a diagram illustrating an example of a calculation result by a specification similarity calculation unit of the search device according to the first embodiment. FIG. 4 is a diagram showing an example of a similar specification search result displayed by a search result display unit of the search device according to the first embodiment. FIG. 4 is a diagram showing a first modification of the search device according to the first embodiment. 11 is a flowchart illustrating an example of a search operation performed by the search device according to the first modification of the first embodiment; FIG. 8 is a diagram showing an example of a similar specification search result displayed by a search result detail display unit of the search device of the first modification example according to the first embodiment. FIG. 5 is a diagram showing a second modification of the search device according to the first embodiment. 11 is a flowchart illustrating an example of a search operation by the search device according to the second modification of the first embodiment; FIG. 14 is a diagram showing an example of a similarity weight data change screen displayed by the search device according to the second modification of the first embodiment; FIG. 14 is a diagram showing an example of a similar specification search result displayed by a search result detail display unit of the search device of the second modification example according to the first embodiment. FIG. 4 is a diagram illustrating a configuration example of a search device according to a second embodiment; 9 is a flowchart illustrating an example of a specification similarity learning process in the search device according to the second embodiment. FIG. 14 is a diagram showing an example of a similarity adjustment screen between item values displayed by the search device according to the second embodiment; The figure which shows an example of the adjusted similarity data which the similarity adjustment part of the search device concerning Embodiment 2 outputs. FIG. 9 is a diagram illustrating a configuration example of a similar specification search unit included in the search device according to the second embodiment; The figure which shows the example of a structure of the similarity learning part with which the search device concerning Embodiment 3 is provided. 11 is a flowchart illustrating an example of the operation of the similarity learning unit included in the search device according to the third embodiment. The figure which shows an example of the use component selection definition which the similarity learning part of the search device concerning Embodiment 3 uses. The figure which shows the concept of the specification similarity learning process which the similarity learning part of the search device concerning Embodiment 3 performs. FIG. 9 is a diagram illustrating a configuration example of a search device according to a fourth embodiment. 11 is a flowchart illustrating an example of a search operation performed by the search device according to the fourth embodiment. The figure which shows an example of the specification input screen which the search device concerning Embodiment 4 displays. FIG. 14 is a diagram illustrating a configuration example of a similar specification search unit included in a search device according to a fourth embodiment. The figure which shows an example of the extended similarity weight data which the similar specification search part of the search device concerning Embodiment 4 uses. 11 is a flowchart illustrating an example of the operation of the similar specification search unit included in the search device according to the fourth embodiment. The figure which shows the concept of the special specification similarity calculation performed by the special specification similarity calculation part of the similar specification search part with which the search device concerning Embodiment 4 is provided. The figure which shows an example of the similar specification search result which the search result detailed display part of the search device concerning Embodiment 4 displays. The figure which shows the example of a structure of the similar specification search part with which the search device concerning Embodiment 5 is provided. 15 is a flowchart illustrating an example of the operation of the similar specification search unit included in the search device according to the fifth embodiment. The figure which shows an example of the keyword for every item which the similar specification search part of the search device concerning Embodiment 5 uses. FIG. 17 is a diagram illustrating an example of a search result obtained by the search device according to the fifth embodiment.

  Hereinafter, a search device, a search method, and a machine learning device according to an embodiment of the present invention will be described in detail with reference to the drawings. The present invention is not limited by the embodiment.

Embodiment 1 FIG.
FIG. 1 is a diagram illustrating a configuration example of the search device according to the first embodiment of the present invention. The search device 1 according to the first embodiment includes a control unit 11, an input unit 12, a product design database 13, a similarity learning unit 14, a similar specification search unit 16, and a search result display unit 17.

  The control unit 11 controls the entire processing executed by the search device 1. In the present embodiment, the control unit 11 controls the input unit 12, the similarity learning unit 14, the similar specification search unit 16, and the search result display unit 17. I do.

  The input unit 12 is provided to allow a user to input information on product specifications as a condition for searching for necessary data from data held in the product design database 13. The input unit 12 acquires information on product specifications input by a user. The information on the product specifications acquired by the input unit 12 is assumed to be information representing the specifications of a newly designed product.

  The product design database 13 holds data on specifications of products designed in the past. Details of the data held by the product design database 13 will be described later.

  The similarity learning unit 14 analyzes the data held in the product design database 13 to learn the similarity between the item values of the product specification, and uses the similarity used in the search processing by the similar specification search unit 16. The data 15 is generated. The item value of the product specification is a value of each item representing the specification of the product. The similarity learning unit 14 constitutes a machine learning device.

  The similar specification search unit 16 generates specification data indicating a specification similar to the specification indicated by the information acquired by the input unit 12 based on the similarity data 15 which is a learning result by the similarity learning unit 14, and outputs the specification data to the product design database 13. Search from the stored specification data group.

  The search result display unit 17 displays a search result by the similar specification search unit 16.

  In the present embodiment, the description will be made on the assumption that the product design database 13 exists inside the search device 1, but the product design database 13 may exist outside the search device 1. For example, the search device 1 and the product design database 13 may be connected via a communication network, and the search device 1 may acquire necessary data from the product design database 13 via the communication network.

  FIG. 2 is a diagram illustrating an example of a hardware configuration that implements the search device according to the first embodiment. The search device 1 is realized by the processing circuit 101, the storage device 102, the input device 103, and the display device 104 illustrated in FIG.

  The processing circuit 101 is a circuit that executes various types of processing in the search device 1, and includes, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), and an FPGA ( Field-Programmable Gate Array) or a combination of these. The storage device 102 is a hard disk drive, a flash memory, or the like. The input device 103 is a keyboard, a mouse, a touch panel, or the like, and may be any device that can input information necessary for the search device 1 to operate. The display device 104 is a display, a liquid crystal monitor device, or the like. The display device 104 may be any device as long as the user can refer to the information input using the input device 103, the search result by the similar specification search unit 16, and the like, and may be a projector or the like.

  The control unit 11, the similarity learning unit 14, and the similar specification search unit 16 of the search device 1 are realized by the processing circuit 101. The product design database 13 is realized by the storage device 102. The storage device 102 also stores the similarity data 15, information input by the user via the input unit 12, a search result by the similar specification search unit 16, and the like. The input unit 12 is realized by the input device 103, and the search result display unit 17 is realized by the display device 104.

  In addition, all or a part of the control unit 11, the similarity learning unit 14, and the similar specification search unit 16 of the search device 1 can be realized by a general-purpose processor and a memory. In this case, the hardware configuration of the search device 1 is as shown in FIG. FIG. 3 is a diagram illustrating another example of the hardware configuration that implements the search device according to the first embodiment.

  The processor 201 illustrated in FIG. 3 includes a CPU (Central Processing Unit, a central processing unit, a processing device, an arithmetic device, a microprocessor, a microcomputer, a DSP (Digital Signal Processor)), a system LSI (Large Scale Integration), and the like. is there. The memory 202 is a RAM (Random Access Memory), a ROM (Read Only Memory), an EPROM (Erasable Programmable Read Only Memory), an EEPROM (registered trademark) (Electrically Erasable Programmable Read Only Memory), or the like. The storage device 102, the input device 103, and the display device 104 illustrated in FIG. 3 are similar to the storage device 102, the input device 103, and the display device 104 illustrated in FIG.

  The hardware having the configuration shown in FIG. 3 may be a tablet PC (Personal Computer), a desktop PC, or a server-client server and client PC. When the control unit 11, the similarity learning unit 14, and the similar specification search unit 16 are realized by the processor 201 and the memory 202 illustrated in FIG. 3, the processor 201 is used as the control unit 11, the similarity learning unit 14, and the similar specification search unit 16. A program for functioning is stored in the memory 202 or the like. Then, the processor 201 reads out and executes the program stored in the memory 202, thereby realizing the control unit 11, the similarity learning unit 14, and the similar specification search unit 16.

  Note that the search device according to each of the second and subsequent embodiments described later is also implemented by hardware having the configuration illustrated in FIG. 2 or FIG.

  Hereinafter, the operation of the search device 1 according to the first embodiment will be described. First, the specification similarity learning process by the similarity learning unit 14 will be described, and then the search operation by the search device 1 will be described. In the following, an example of a search operation for designing a motor will be described. However, the present invention can be applied to any product having data corresponding to data held in a product design database 13 described later.

(Specification similarity learning processing by similarity learning unit 14)
FIG. 4 is a diagram illustrating an example of a detailed configuration of the similarity learning unit 14 included in the search device 1 according to the first embodiment. The similarity learning unit 14 configuring the machine learning device includes a used component extraction unit 41 and a similarity calculation unit 42. The used component extraction unit 41 extracts a used component list for each specification item value from the product design database 13. The similarity calculator 42 calculates the similarity between the item values based on the used component list for each specification item value extracted by the used component extractor 41.

  FIG. 5 is a diagram illustrating a configuration example of the product design database 13 included in the search device 1 according to the first embodiment. The product design database 13 includes a specification list 131 listing specifications of products designed in the past, and a component configuration table 132 recording the component configuration of each product. The specification list 131 corresponds to specification data. An example of the specification list 131 is shown in FIG. 6, and an example of the component configuration table 132 is shown in FIG.

  FIG. 6 is a diagram illustrating an example of a specification list stored in the product design database 13 according to the first embodiment. The specification list 131 includes at least a model ID (Identification) 601 for identifying each product, a series name 602 which is a name of a product group in which specific products having similar specifications are collected, and a rating of the product of each series. Includes an output 603, a rated rotation speed 604 of each series of products, a brake 605 indicating whether each series of products has an electromagnetic brake, and a special specification 606 indicating special specifications of each series of products. . The special specification 606 stores a natural sentence, that is, text data. A series in which the special specification 606 is "-" means that text data indicating the special specification is not stored. In an actual motor, there are a number of specification items such as a maximum rotation speed, presence or absence of an oil seal, and a shape of a shaft end. In the following, for simplicity of description, among the six types of items shown in FIG. Only five types of items except for the special specification 606 are used. The number of items used for the search may be any number as long as it is one or more. Hereinafter, products belonging to the same series, that is, a product group to which the same model ID is assigned may be collectively referred to as a model.

  FIG. 7 is a diagram illustrating an example of a component configuration table held by the product design database 13 according to the first embodiment. The component configuration table 132 includes at least a model ID 601 for identifying individual products, a component hierarchy 701, a component ID 702 for identifying individual components, a component name 703, and components used in one product. And the number 704 indicating the number. The parts configuration table 132 is information on parts constituting each product corresponding to the model ID 601. The model ID 601 included in the component configuration table 132 is the same as the model ID 601 included in the specification list 131. A code common to the specification list 131 and the parts configuration table 132 is used as the model ID 601. The component hierarchy 701 indicates to which hierarchy the component belongs when the components are hierarchized and managed. For example, when a product is composed of a plurality of types of first parts, and a certain first part is composed of a plurality of second parts, the part hierarchy 701 of the first part is “1”, The component hierarchy 701 of the second component is “2”. Further, when the second part is composed of the third part, the part hierarchy 701 of the third part is “3”. The actual motor includes other items such as material and dimensions, but only the five types of items shown in FIG. 7 will be used below to simplify the description.

  FIG. 8 is a flowchart illustrating an example of an operation of the similarity learning unit 14 included in the search device 1 according to the first embodiment, and illustrates a procedure of a specification similarity learning process. For example, when the search device 1 receives an operation for instructing the start of the specification similarity learning from the user, the similarity learning unit 14 starts the specification similarity learning process illustrated in FIG.

  When starting the specification similarity learning process, the similarity learning unit 14 first executes a specification list acquisition process (step S11). Specifically, the used component extraction unit 41 acquires a specification list 131 having the contents shown in FIG. 6 from the product design database 13.

  Next, the similarity learning unit 14 repeats the processing of steps S12 and S13 for each specification item included in the specification list 131. Specifically, the similarity learning unit 14 repeats the processing of steps S12 and S13 for each specification item of the series name 602, the rated output 603, and the rated rotation speed 604. At this time, the similarity learning unit 14 repeats the process of step S12 for each specification item value belonging to the specification item. For example, when the specification item A includes the specification item values a1, a2, and a3, the similarity learning unit 14 performs the processing of step S12 when performing the processing of steps S12 and S13 for the specification item A. It is repeatedly executed for each of the specification item values a1, a2 and a3. In step S12, the similarity learning unit 14 extracts a component corresponding to the specification item value from the component configuration table 132. In step S13, the similarity learning unit 14 calculates the similarity between the item values by using a method described later. Step S12 is a process executed by the used component extraction unit 41, and step S13 is a process executed by the similarity calculation unit 42.

  For example, the series name 602 of the specification list 131 shown in FIG. 6 includes A series, B series, C series, D series, and so on. Therefore, in the processing of steps S12 and S13 performed for the series name 602, the similarity learning unit 14 repeatedly executes the processing of step S12 for the A series, the B series, the C series, the D series, and so on. Then, the similarity between the item values is calculated using the extracted parts.

  Here, in the case of the processing in step S12 for the A series, the A series includes a product with a model ID of “1” and a product with a model ID of “2”. Therefore, the similarity learning unit 14 determines all the components used in the product with the model ID “1” and all the components used in the product with the model ID “2” in the component configuration shown in FIG. Obtained from the table 132. For example, the parts used in the product whose model ID is "1" correspond to the parts whose part ID is PA001, PA002, PA003, PA004,.

  The rated output 603 of the specification list 131 shown in FIG. 6 includes 3.5 kW, 2.0 kW, 1.5 kW, 400 W,... Therefore, the similarity learning unit 14 executes the processing of step S12 for 3.5 kW, 2.0 kW, 1.5 kW, 400 W,... In the processing of steps S12 and S13 for the rated output 603. The corresponding part is extracted, and then the similarity between the item values is calculated using the extracted part.

  Here, in the case of the processing of step S12 for 2.0 kW, the product with the rated output of 2.0 kW includes the product with the model ID of “2”, the product with the model ID of “3”, and the product with the model ID of “3”. However, there is a product with '4'. Therefore, the similarity learning unit 14 determines that all the components used in the product with the model ID “2”, all the components used in the product with the model ID “3”, and All parts used in the product are obtained from the parts configuration table 132 shown in FIG.

  In step S13, the similarity calculation unit 42 determines the item value based on the data of the used component acquired by the used component extraction unit 41 executing step S12 for each specification item value included in the specification item to be processed. A similarity between them, that is, a value indicating how similar the specification item values are, is calculated. Hereinafter, the value calculated by the similarity calculation unit 42 is referred to as similarity. The similarity calculation unit 42 calculates the value of the inter-item value similarity SimSpecValue (Va, Vb), which is the similarity between the specification item value Va and the specification item value Vb, according to the following equation (1).

  Here, Parts (V) is a set of parts used in the entire model having V as the specification item value. Count (P) is the number of types of components in the component set P. FIG. 9 is a diagram illustrating a concept of a method of calculating the similarity between item values in the similarity learning unit 14 included in the search device 1 according to the first embodiment. For example, there are 90 types of parts used in the model of the A series, PA001, PA002, ..., PA090, and 80 types of components used in the model of the B series, PB001, PB002, ..., PB080. I do. When 70 types of components are common among them, SimSpecValue ('A series', 'B series') = 70 / (90 + 80−70) = 0.7.

  The example in which the similarity between the item values is calculated for the A series and the B series has been described. However, when there is a series other than the A series and the B series, in step S13, the combination of all the series is Calculate the similarity between item values. For example, when the C series exists in addition to the A series and the B series, the similarity calculation unit 42 calculates the similarity between the item values of the A series and the B series, the similarity between the item values of the A series and the C series, and The similarity between the item values of the B series and the C series is calculated.

  In step S13, when the similarity calculation unit 42 finishes calculating the inter-item-value similarities for all the combinations of the specification item values, the similarity calculation unit 42 generates the similarity data 15 including all the calculated inter-item-value similarities. Output.

  As described above, the similarity learning unit 14 analyzes the specification list 131, which is the specification data group held by the product design database 13, and the component configuration table 132, and determines the same for each of the specification items representing the product specifications. The degree of similarity between the two specification item values of the specification item is learned for all combinations of the specification item values. That is, based on the specification data group held in the product design database 13, the similarity learning unit 14 determines how similar two specification item values of the same use item are to each other. Learn about combinations.

  FIG. 10 is a diagram illustrating an example of the similarity data 15 generated by the similarity learning unit 14 included in the search device 1 according to the first embodiment. The similarity data 15 includes at least a specification item, a specification item value # 1, a specification item value # 2, and a similarity. In the following example, the similarity between the specification item values is a one-to-one relationship, that is, the similarity between one specification item, but the similarity calculation unit 42 has a plurality of similarities as shown in FIG. The similarity data 15 may be generated by calculating the similarity between the combinations of the specification items. FIG. 11 shows an example of the similarity data 15 including the similarity for the combination of the series name and the rated output, and the similarity for the combination of the rated output and the rated rotation speed.

(Search operation by search device 1)
Subsequently, a search operation by the search device 1 will be described. FIG. 12 is a flowchart illustrating an example of a search operation performed by the search device 1 according to the first embodiment. For example, when the search device 1 receives an operation for instructing the start of the search process from the user, the search device 1 starts the operation according to the flowchart illustrated in FIG.

  When the search device 1 starts the search process, first, the input unit 12 receives a specification input, that is, receives a specification item value input by a designer who is a user as a search condition (step S21). In step S21, the search device 1 displays, for example, the specification input screen illustrated in FIG. 13 on a display unit whose description is omitted in FIG. 1, and waits for a user to input a specification item value. The display unit is realized by the display device 104 shown in FIGS. The specification input screen shown in FIG. 13 includes a series name input field 1301, a rated output input field 1302, a rated rotation speed input field 1303, and a brake input / output field 1304. In the example of the specification input screen shown in FIG. 13, the configuration is such that the specification item value is input by selecting a pull-down menu, but any configuration may be used. The specification item value may be automatically set based on the order data in cooperation with another system, for example, a sales system.

  When the input of the specification item value by the input unit 12 is completed, the similar specification search unit 16 searches for a similar specification (step S22). That is, the similar specification search unit 16 searches the product design database 13 based on the similarity data 15 and the specification item value input in step S21, and obtains a specification similar to the specification indicated by the specification item value input in step S21. Specify the model of. Hereinafter, details of the search processing by the similar specification search unit 16 will be described.

  FIG. 14 is a diagram illustrating an example of a detailed configuration of the similar specification search unit 16 included in the search device 1 according to the first embodiment. The similar specification search unit 16 includes a specification similarity calculation unit 61 and a data alignment unit 62.

  Based on the similarity data 15 generated by the similarity learning unit 14 and the similarity weight data 65, the specification similarity calculation unit 61 calculates the specification of the past model which is the model stored in the product design database 13. For each of them, the similarity with the specification indicated by the input specification data 63 output from the input unit 12 is calculated. The input specification data 63 is data including each specification item value that the input unit 12 has received an input in step S21. Details of the similarity weight data 65 will be described separately.

  The data sorting unit 62 sorts the calculation result of the specification similarity calculation unit 61, that is, the similarity of the specifications of each past model in descending order of similarity, and the similarity of the specifications of each past model after sorting. The similarity is output as the similar specification search result 64.

  FIG. 15 is a diagram illustrating an example of the similarity weight data used by the similar specification search unit 16 of the search device 1 according to the first embodiment. The similarity weight data 65 includes a specification item 1501 and a similarity weight 1502. The similarity weight data 65 in the example shown in FIG. 15 includes similarity which is a weight coefficient for each of the series name, the rated output, the rated rotation speed, and the brake, which are the specification items included in the specification list 131 of the product design database 13. Includes weights. These similarity weights are used when calculating the similarity of specifications described later. The weighting of each specification item shall be manually defined. For example, the search device 1 has a function of creating the similarity weight data 65, and the designer creates the similarity weight data 65 using this function. The designer may create the similarity weight data 65 using an external device, for example, a personal computer. When the operation of the search device 1 is started, the search result may be evaluated by trial operation using the temporarily set similarity weight, and the value of each similarity weight may be adjusted based on the evaluation result.

  FIG. 16 is a flowchart illustrating an example of an operation of the similar specification search unit 16 included in the search device 1 according to the first embodiment, and illustrates a procedure of a similar specification search process. For example, when the search device 1 receives from the user an operation instructing to start searching for similar specifications, the similar specification search unit 16 starts the similar specification search process illustrated in FIG.

  When starting the similar specification search process, the similar specification search unit 16 first executes a specification list acquisition process (step S31). Specifically, the specification similarity calculation unit 61 acquires the specification list 131 from the product design database 13.

  Next, the similar specification search unit 16 repeats the process of step S32 for each past model included in the specification list 131. For example, when the specification list 131 includes models whose model IDs are 1 to 5, the similar specification search unit 16 executes the process of step S32 for each of the five past models whose model IDs are 1 to 5. Then, an item-specific specification similarity described later is calculated.

  In step S32, the specification similarity calculation unit 61 of the similar specification search unit 16 compares the specification of the past model and the input specification data 63 stored in the product design database 13 based on the similarity data 15 and the similarity weight data 65. Calculate the similarity of Specifically, the specification similarity calculation unit 61 calculates the similarity SimSpec (Sinp, Smx) between the input specification Sinp and the specification Smx of the model mx according to the following equation (2).

  Here, Sim (Sinp (s), Smx (s)) indicates whether the s-th specification item Sinp (s) matches Smx (s) among the four types of specification items included in the specification list 131. It is a value determined by Sim (Sinp (s), Smx (s)) is included in the similarity data 15 shown in FIG. 10 when the Sinp (s) and Smx (s) match, and when the Sin (s) and Smx (s) do not match. Is the value of the similarity. Wdef (s) is the value of the similarity weight corresponding to the s-th specification item defined in the similarity weight data 65.

  FIG. 17 is a diagram illustrating an example of a calculation result by the specification similarity calculation unit 61 of the similar specification search unit 16 included in the search device 1 according to the first embodiment. In the example shown in FIG. 17, when the specifications having the contents shown in FIG. 13 are input in step S21, the model IDs of the specification list 131 shown in FIG. 6 are “1”, “2”, and “3”. Indicates the result of the specification similarity calculation unit 61 calculating the degree of similarity SimSpec to each model.

  In the specifications of the contents shown in FIG. 13, the series name is “A series”, the rated output is “2.0 kW”, the rated rotation speed is “2000 r / min”, and the brake is “none”. Therefore, for example, the past model whose model ID is “1” has the same series name, rated rotational speed, and brake specifications from the contents shown in FIG. 6, and the similarity of these specification items is “1”. . On the other hand, the specifications of the rated outputs are inconsistent, and the similarity of the rated outputs is similar to the similarity “0.05” of the rated outputs 3.5 kW and 2.0 Kw included in the similarity data 15 shown in FIG. Become.

  In the example of the calculation result illustrated in FIG. 17, the similarity SimSpec (Sinp, Sm1) of the past model whose model ID is “1” is 0.83, and the similarity SimSpec (Sinp, Simp, Sm2) is 0.75, and the similarity SimSpec (Sinp, Sm3) of the past model whose model ID is “3” is 0.89. The past model whose model ID is "3" has a series name different from the series name input in step S21 described above. However, due to the effect of the similarity data 15, the similarity is the highest among the three models. Become.

  Returning to the description of FIG. 16, after the specification similarity calculation unit 61 has performed the processing of step S32 for all the past models, the data alignment unit 62 determines the similarity calculation result SimSpec (Sinp , Smx) are rearranged (step S33). That is, the data sorting unit 62 sorts the similarity calculation results SimSpec (Sinp, Smx) so as to sort them in descending order of similarity.

  Returning to the description of FIG. 12, when the similar specification search process of step S22 is completed by the similar specification search unit 16, the search result display unit 17 displays the search result of the similar specification and presents the search result to the designer (step S22). S23). FIG. 18 is a diagram illustrating an example of the similar specification search result displayed by the search result display unit 17 of the search device 1 according to the first embodiment. FIG. 18 shows an example in which the similar specification search result is displayed together with the specification input in step S21. In the example illustrated in FIG. 18, the model ID of the past model and the similarity score that is the above-described similarity SimSpec (Sinp, Smx) are arranged and displayed in descending order of the similarity score. When the number of past models is large, the search result display unit 17 may extract and display the similarity score that is larger than a predetermined value, or may display the similarity score in descending order of the highest similarity score. Alternatively, a predetermined number of past models may be extracted and displayed. The designer selects a model similar to the required specification as appropriate based on the search result displayed on the search result display unit 17 and designs a new model that satisfies the required specification.

  Further, when displaying the search result by the similar specification search unit 16, the search device 1 may be configured to display in detail the search score that is the basis of the search result in order to clearly indicate why the model was searched. FIG. 19 is a diagram illustrating a first modification of the search device according to the first embodiment. The search device 1a shown in FIG. 19 is obtained by replacing the search result display unit 17 of the search device 1 shown in FIG. The configuration of the search device 1a other than the search result detail display section 18 is the same as that of the search device 1.

  FIG. 20 is a flowchart illustrating an example of a search operation by the search device 1a according to the first modification of the first embodiment. The flowchart shown in FIG. 20 is obtained by replacing step S23 of the flowchart shown in FIG. 12 with step S24. Step S23 is processing executed by the search result detail display unit 18. That is, in step S23, the search result detail display unit 18 performs display based on the search result output from the similar specification search unit 16, and presents the search result of the similar specification to the designer. The search result details display unit 18 displays, for example, search results of similar specifications having the configuration shown in FIG. The example shown in FIG. 21 displays details of the similarity score corresponding to the search score in addition to the contents of the similar specification search result shown in FIG. The search result detail display unit 18 corresponds to a search result display unit that displays a search result by the similar specification search unit 16 together with a similarity corresponding to the search result.

  Further, by providing the search device 1 with a search weight changing unit for changing the weight of the specification similarity, it is possible to easily change the search condition for obtaining the search result desired by the designer. FIG. 22 is a diagram illustrating a second modification of the search device according to the first embodiment. The search device 1b shown in FIG. 22 is obtained by adding a weight data changing unit 21 as search weight changing means to the search device 1a shown in FIG. The configuration of the search device 1b other than the weight data changing unit 21 is the same as that of the search device 1a.

  FIG. 23 is a flowchart illustrating an example of a search operation performed by the search device 1b according to the second modification of the first embodiment. The flowchart shown in FIG. 23 is obtained by adding step S25 after step S24 of the flowchart shown in FIG.

  In step S25, the weight data changing unit 21 receives a similarity weight changing operation from the user. At this time, the weight data changing unit 21 displays, for example, the similarity weight data changing screen shown in FIG. 24 on a display unit not shown in FIG. 22 and performs an operation for the user to correct the similarity weight data. Wait to do. FIG. 24 shows an example in which the weight is changed to the similarity weight that emphasizes the matching of the series names. Specifically, FIG. 24 shows an example in which the similarity weight of the series name is changed from 2.0 to 5.0 and the similarity weight of the rated rotation speed is changed from 1.5 to 1.0. ing.

  When the operation of changing the similarity weight data is received in step S25, the search device 1b performs step S22 using the changed similarity weight data to perform a search again, and displays the result of the search again in step S23. . FIG. 25 is a diagram illustrating an example of a result of the re-search performed by changing the similarity weight data. By changing the similarity weight in the above step S25, the past models having the same series name are the top candidates as intended by the designer.

  As described above, the search device according to this embodiment learns the similarity between the item values of the specifications of the past model, which is the product designed in the past, based on the data on the product designed in the past. The learning unit includes a learning unit and a similar specification search unit that searches for specification data indicating a specification similar to the specification specified by the designer, based on the similarity data obtained as a result of learning by the similarity learning unit. This eliminates the need for the designer to create a complicated rule such as the knowledge base described in Patent Literature 1 in advance, and can reduce the workload of the designer in searching for data to be reused. . In addition, an efficient search that does not depend on the experience of the designer can be realized.

  In addition, when the search device according to the present embodiment displays details of the similarity score in addition to the similar specification search result as the configuration illustrated in FIG. 19, the search device notifies the designer of the basis of the search result. be able to.

  In addition, the search device according to the present embodiment displays the details of the similarity score in addition to the similar specification search result as the configuration illustrated in FIG. 22, and further receives a change in the similarity weight for each specification item. In this case, the search conditions can be changed so that the search result desired by the designer can be obtained.

Embodiment 2 FIG.
FIG. 26 is a diagram illustrating a configuration example of a search device according to the second embodiment. The search device 1c according to the second embodiment replaces the similar specification search unit 16 of the search device 1 (see FIG. 1) according to the first embodiment with a similar specification search unit 16c, and further adds a similarity adjustment unit 31. Things. In the search device 1c, the similar specification search unit 16c performs a search process using the adjusted similarity data generated by the similarity adjustment unit 31. The configuration of the search device 1c is the same as that of the search device 1 except for the configuration of the similarity adjustment unit 31 and the similar specification search unit 16c.

  The basic operation when the search device 1c searches for a past model having a specification similar to the specification specified by the designer is the same as that of the search device 1 (see FIG. 12). The search device 1c differs from the search device 1 in part of the content of the specification similarity learning process.

  FIG. 27 is a flowchart illustrating an example of the specification similarity learning process in the search device 1c according to the second embodiment. The flowchart shown in FIG. 27 is obtained by adding step S41 to the flowchart shown in FIG. Except for step S41, the process is the same as the specification similarity learning process described in the first embodiment, and thus the same step numbers as those in the corresponding processes in FIG.

  In step S41, the similarity between the item values is adjusted. This adjustment is adjustment of the similarity data 15 which is a learning result by the similarity learning unit 14, and is performed by the designer using the similarity adjustment unit 31. When executing the step S41, the similarity adjustment unit 31 displays, for example, the similarity adjustment screen between item values shown in FIG. 28 on a display unit whose description is omitted in FIG. The similarity adjustment unit 31 waits for an operation performed by the designer while the similarity adjustment screen between item values is displayed on the display unit. If the similarity needs to be adjusted, the designer inputs the adjusted similarity for the similarity to be adjusted, and when the input is completed, presses the save button to end the operation. If it is not necessary to adjust the similarity, the designer presses the save button or the cancel button without inputting the adjusted similarity, and ends the operation. FIG. 28 shows an example in which the similarity between the A series and the B series and the similarity between the A series and the C series are adjusted by the user.

  When the operation of adjusting the similarity by the designer is completed, the similarity adjustment unit 31 outputs all the similarities between the item values including the adjusted similarity as the adjusted similarity data 32. FIG. 29 is a diagram illustrating an example of the adjusted similarity data 32. The adjusted similarity data 32 has a configuration in which the adjusted similarity is added to the similarity data 15 (see FIG. 10) described in the first embodiment.

  FIG. 30 is a diagram illustrating a configuration example of the similar specification search unit 16c included in the search device 1c according to the second embodiment. The similar specification search unit 16c is configured by replacing the specification similarity calculation unit 61 of the similar specification search unit 16 (see FIG. 14) described in the first embodiment with a specification similarity calculation unit 61c. The specification similarity calculation unit 61c inputs the specifications of the past model stored in the product design database 13 based on the adjusted similarity data 32 generated by the similarity adjustment unit 31 and the similarity weight data 65. The similarity with the input specification data 63 output from the unit 12 is calculated. At this time, the specification similarity calculation unit 61c performs the calculation using the adjusted similarity included in the adjusted similarity data 32. The method of calculating similarity by the specification similarity calculation unit 61c is the same as the method of calculation of similarity by the specification similarity calculation unit 61, and a description thereof will be omitted.

  As described above, the search device 1c according to the present embodiment includes the similarity adjustment unit 31 that adjusts the similarity data 15 that is the learning result of the similarity learning unit 14. As a result, the designer can specify the similarity to be strengthened and the similarity to be suppressed, and adjust the value of the similarity appropriately. Further, according to the search device 1c, even if the similarity data is not adjusted, the designer can individually check the value of the similarity data 15 to determine whether appropriate learning has been performed. it can.

Embodiment 3 FIG.
FIG. 31 is a diagram illustrating a configuration example of the similarity learning unit 14d included in the search device according to the third embodiment. The configuration of the search device according to the third embodiment other than the similarity learning unit 14d is the same as that of the search device 1 according to the first embodiment. Therefore, description of components other than the similarity learning unit 14d will be omitted.

  The similarity learning unit 14d is obtained by adding a used component selection unit 43 to the similarity learning unit 14 described in the first embodiment.

  FIG. 32 is a flowchart illustrating an example of an operation of the similarity learning unit 14d included in the search device according to the third embodiment. The flowchart shown in FIG. 32 is obtained by adding step S51 to the flowchart shown in FIG. Since the processes other than step S51 are the same as the processes included in the flowchart illustrated in FIG. 8, the same step numbers as the corresponding processes in FIG. 8 are assigned and the description is omitted.

  Step S51 is used item selection processing for each item value. In this step S51, when the similarity calculator 42 calculates the similarity between the item values from the extraction result by the used component extractor 41 based on the used component selection definition 44, Select the extraction result to use. FIG. 33 is a diagram illustrating an example of the used component selection definition 44. As shown in FIG. 33, the used component selection definition 44 includes at least a specification item 3301 and a related component number list 3302 corresponding to the specification item 3301. The used component selection definition 44 is created in advance by, for example, a designer. The use component selection definition 44 may be created using a device external to the search device, for example, a personal computer, or the search device has a function of creating the use component selection definition 44. May be created by the designer.

  In step S <b> 51, the used component selection unit 43 writes the related component number list 3302 associated with the corresponding specification item of the used component selection definition 44 among the components corresponding to the specification item value extracted in step S <b> 12. Select the part with the part number. The used component selection unit 43 outputs the selection result to the similarity calculation unit 42.

  In step S13, the similarity learning unit 14d does not calculate the similarity between the item values for all the item values extracted by the used component extraction unit 41, but targets the item values selected by the used component selection unit 43. And calculate the similarity between the item values.

  FIG. 34 is a diagram illustrating the concept of the specification similarity learning process executed by the similarity learning unit 14d of the search device according to the third embodiment. In the example shown in FIG. 34, in calculating the similarity between the specifications of the model X and the model Y, the similarity changes depending on which of the specification item “rated output” and the specification item “rated rotation speed” is focused on. That is, focusing on the “rated output”, it can be said that the similarity is high because the component PA010 and the component PA011 are common. On the other hand, focusing on the “rated rotation speed”, since the model X is composed of the components PB020 and PB021, and the model Y is composed of the components PC020 and PC021, the two have no commonality and low similarity. It can be said that. The similarity learning unit 14d improves the calculation accuracy of the similarity data 15 by selecting the characteristic parts group for each specification item and calculating the similarity using the used component selection definition 44. be able to.

  As described above, in the search device according to the present embodiment, the similarity learning unit 14d includes the used component selection unit 43, and the used component selection unit 43 selects a characteristic component group for each specification item, and Calculate the similarity between values. Thereby, the calculation accuracy of the similarity data can be improved.

Embodiment 4 FIG.
FIG. 35 is a diagram illustrating a configuration example of a search device according to the fourth embodiment. The search device 1e according to the fourth embodiment is obtained by replacing the similar specification search unit 16 of the search device 1a shown in FIG. 19 with a similar specification search unit 16e, and further adding a special specification input unit 51. The configuration of the search device 1e other than the special specification input unit 51 and the similar specification search unit 16e is the same as that of the search device 1a, so the same reference numerals as those in FIG.

  FIG. 36 is a flowchart illustrating an example of a search operation performed by the search device 1e according to the fourth embodiment. The flowchart shown in FIG. 36 is obtained by adding step S61 to the flowchart shown in FIG. Since the processes other than step S61 are the same as the processes included in the flowchart shown in FIG. 12, the same step numbers as those of the corresponding processes in FIG.

  In step S61, the special specification input unit 51 receives an input of a special specification from a designer. At this time, the search device 1e displays the specification input screen shown in FIG. 37 on a display unit not shown in FIG. The special specification input unit 51 waits for an operation performed by the designer with the display unit displaying the specification input screen. The specification input screen shown in FIG. 37 has a configuration in which an input column 3701 for special specifications is added to the specification input screen shown in FIG. The specification input screen of FIG. 37 is displayed at the start of step S21.

  The designer inputs a series name, a rated output, and the like to the search device 1e, and also inputs special specifications. When the operation of inputting the special specification by the designer is completed, the special specification input unit 51 generates data indicating the content input in the special specification input field 3701 and outputs the data to the similar specification search unit 16e.

  FIG. 38 is a diagram illustrating a configuration example of the similar specification search unit 16e included in the search device 1e according to the fourth embodiment. The similar specification search unit 16e replaces the specification similarity calculation unit 61 of the similar specification search unit 16 (see FIG. 14) described in the first embodiment with a special specification similarity calculation unit 81 and an extended specification similarity calculation unit 82. It is.

  The special specification similarity calculation unit 81 calculates the similarity between the input special specification data 83 output from the special specification input unit 51 and the special specification of the past model stored in the product design database 13. The extended specification similarity calculation unit 82 calculates the specifications of the past model stored in the product design database 13 based on the similarity of the special specification calculated by the special specification similarity calculation unit 81 and the extended similarity weight data 84. And the input specification data 63 output from the input unit 12 are calculated.

  FIG. 39 is a diagram illustrating an example of the extended similarity weight data 84. The extended similarity weight data 84 includes a specification item 3901 and a similarity weight 3902. The extended similarity weight data in the example shown in FIG. 39 is a weight coefficient for each of a series name, a rated output, a rated rotation speed, a brake, and a special specification, which are specification items included in the specification list 131 of the product design database 13. Contains some similarity weights. The weighting of each specification item shall be manually defined. When the operation of the search device 1e is started, the search result may be evaluated by trial operation using the temporarily set similarity weight, and the value of each similarity weight may be adjusted based on the evaluation result.

  FIG. 40 is a flowchart illustrating an example of an operation of the similar specification search unit 16e included in the search device according to the fourth embodiment. The flowchart shown in FIG. 40 is obtained by replacing step S32 of the flowchart shown in FIG. 16 with steps S71 and S72. Since the processes other than steps S71 and S72 are the same as the processes included in the flowchart shown in FIG. 16, the same step numbers as those of the corresponding processes in FIG. 16 are assigned and the description is omitted.

  Step S71 is a special specification similarity calculation process. In step S71, the special specification similarity calculation unit 81 calculates the similarity between the special specification of the past model stored in the product design database 13 and the input special specification data 83. Specifically, the special specification similarity calculation unit 81 calculates the similarity SpecialSimSpec (SSinp, SSmx) between the input special specification SSinp and the special specification SSmx of the model mx according to the following equation (3).

  Here, SSinp (t1) and SSmx (t2) are the t1st and t2th texts when each special specification is separated by a reading point (,), and SSim (SSinp (t1), SSmx (t2)) is , Is the edit distance between these texts. MAX (1-SSim (SSinp (t1), SSmx (t2)) / L) indicates the maximum value and the value at t2 of SSinp (t1). L is the total number of characters in the texts SSinp (t) and SSmx (t), and n indicates the number of texts when the input special specification is delimited by the reading point (,). Note that the calculation of the edit distance between texts is performed using a known method, for example, a method disclosed in Patent Document “International Publication No. WO 2015/040793”.

  FIG. 41 is a diagram illustrating the concept of the special specification similarity calculation performed by the special specification similarity calculation unit 81 of the similar specification search unit 16e included in the search device according to the fourth embodiment. As shown in FIG. 41, the special specification similarity calculation unit 81 divides each of the text indicated by the input special specification data 83 and the text corresponding to the special specification of the past model mx by a reading point, and , The similarity is calculated. In the example shown in FIG. 41, the special specification similarity calculation unit 81 divides each text into two, and calculates the similarity of the divided text. The special specification similarity calculation unit 81 outputs the calculated average value of the similarities as the special specification similarity.

  Step S72 is an extended item-specification similarity calculation process. In step S72, the extended specification similarity calculation unit 82 stores in the product design database 13 based on the similarity of the special specification calculated by the special specification similarity calculation unit 81 in step S71 and the extended similarity weight data 84. The similarity between the specification of the past model and the input specification data 63 is calculated. Specifically, the extended specification similarity calculation unit 82 calculates the similarity ESimSpec (Sinp, SSinp) between the input specification Sinp and the input special specification SSinp, and the specification Smx and the special specification SSmx of the model mx according to the following equation (4). , Smx, SSmx).

  Here, Sim (Sinp (s), Smx (s)) is the s-th specification item Sinp (s) and Smx of the four types of specification items included in the specification list 131 as described in the first embodiment. This value is determined depending on whether or not (s) matches. That is, Sim (Sinp (s), Smx (s)) is “1” when Sinp (s) and Smx (s) match, and similarity data 15 shown in FIG. Is the value of the similarity included in. EWdef (s) is a similarity weight value corresponding to the s-th specification item defined in the extended similarity weight data 84.

  In step S33 in FIG. 40, the data sorting unit 62 sorts the similarity calculation results ESimSpec (Sinp, SSinp, Smx, SSmx) by the extended specification similarity calculation unit 82. That is, the data sorting unit 62 sorts the similarity calculation results ESimSpec (Sinp, SSinp, Smx, SSmx) so that they are sorted in descending order of similarity.

  The search result detail display unit 18 displays, for example, a search result of a similar specification having the configuration shown in FIG. 42 on the display unit, and presents the search result to the designer (FIG. 36, step S24).

  As described above, the similar specification search unit 16e of the search device according to the present embodiment analyzes the special specification described in a natural sentence as a text and calculates the similarity of the special specification by the special specification similarity calculation unit 81. And searches for a past model having a specification similar to the specification specified by the designer in consideration of the similarity of the special specification. As a result, it is possible to search for past models having high similarity in consideration of special specifications, and search accuracy can be improved.

Embodiment 5 FIG.
FIG. 43 is a diagram illustrating a configuration example of the similar specification search unit 16f included in the search device according to the fifth embodiment. The configuration of the search device according to the fifth embodiment other than the similar specification search unit 16f is the same as that of the search device 1e according to the fourth embodiment. Therefore, description of the components other than the similar specification search unit 16f will be omitted.

  The similar specification search unit 16f is obtained by adding a weight adjustment unit 85 to the similar specification search unit 16e according to the fourth embodiment. The weight adjustment unit 85 adjusts the extended similarity weight data 84 based on the input special specification data 83 and the item-specific keywords 86.

  FIG. 44 is a flowchart illustrating an example of the operation of the similar specification search unit 16f included in the search device according to the fifth embodiment. The flowchart shown in FIG. 44 is obtained by adding step S81 to the flowchart shown in FIG. Since the processes other than step S81 are the same as the processes included in the flowchart shown in FIG. 40, the same step numbers as those of the corresponding processes in FIG. 40 are assigned and the description is omitted.

  Step S81 is a similarity weight adjustment process. In step S81, the weight adjustment unit 85 checks whether or not the input special specification data 83 includes the keyword registered as the item-specific keyword 86, and based on the check result, changes the value of the extended similarity weight data 84. adjust. The extended specification similarity calculator 82 receives the extended similarity weight data 84 adjusted by the weight adjuster 85. The weight adjustment unit 85 adjusts the extended similarity weight data 84 according to a procedure defined in advance. For example, when the keyword registered in the item-specific keyword 86 is included in the input special specification data 83, the weight adjusting unit 85 determines, among the similarity weights included in the extended similarity weight data 84, the input special specification data 83. The similarity weight of the specification item corresponding to the keyword included in is changed to a large value. One example of a process in which the weight adjustment unit 85 changes the similarity weight to a large value is a process of multiplying the similarity weight by a coefficient having a value larger than one.

  FIG. 45 is a diagram illustrating an example of an item-specific keyword 86 used by the similar specification search unit 16f of the search device according to the fifth embodiment. The item-specific keyword 86 includes a specification item 4501 and a keyword list 4502.

  FIG. 46 is a diagram illustrating an example of a search result obtained by the search device according to the fifth embodiment. In the example shown in FIG. 46, the series name included in the information input using the input unit 12 and the special specification input unit 51 is “B series”, the rated output is “2.0 kW”, and the rated rotation speed is “1000r”. / Min ”, the brake is“ present ”, the special specification is“ brake torque special ”, and the keyword list 4502 is the search result shown in FIG.

  In the above example, the special specification is “brake torque special”, and “brake torque special” exists in the keyword list of the specification item “brake” of the item-specific keyword 86 shown in FIG. Therefore, the search result shown in FIG. 46 is obtained by adjusting the similarity weight 3902 corresponding to the brake of the specification item 3901 of the extended similarity weight data 84 shown in FIG. Has been done. The reason why the similarity weight corresponding to the brake is adjusted twice is to simplify the explanation. It may be adjusted to another value.

  The item-specific keywords 86 are created in advance by, for example, a designer. The creation of the item-specific keywords 86 may be performed using a device external to the search device, for example, a personal computer, or the search device has a function of creating the item-specific keywords 86, and is designed using this function May be created by the user.

  As described above, the similar specification search unit 16f of the search device according to the present embodiment includes the weight adjustment unit 85 that adjusts the similarity weight for each specification item based on the keyword included in each item and the text included in the special specification. . This makes it possible to calculate the similarity after adjusting the similarity weight according to the description content of the special specification, and to search for a past model having high similarity.

  The configurations described in the above embodiments are merely examples of the contents of the present invention, and can be combined with another known technology, and can be combined with other known technologies without departing from the gist of the present invention. Parts can be omitted or changed.

  1, 1a, 1b, 1c, 1e search device, 11 control unit, 12 input unit, 13 product design database, 14, 14d similarity learning unit, 15 similarity data, 16, 16c, 16e, 16f similar specification search unit, 17 search result display section, 18 search result detail display section, 21 weight data change section, 31 similarity adjustment section, 32 adjusted similarity data, 41 used component extraction section, 42 similarity calculation section, 43 used component selection section, 44 Use part selection definition, 51 Special specification input unit, 61, 61c specification similarity calculation unit, 62 data alignment unit, 63 input specification data, 64 similar specification search result, 65 similarity weight data, 81 special specification similarity calculation unit , 82 extended specification similarity calculation unit, 83 input special specification data, 84 extended similarity weight data, 85 weight adjustment unit, 86 keyword-specific keywords De.

Claims (11)

A search device for searching specification data groups each including one or more specification item values indicating details of an item representing a product specification from specification data groups each of which is similar to a search condition,
A similarity learning unit that learns, for all combinations of specification item values, how similar two specification item values of the same item are based on the specification data group held by the product design database;
An input unit that receives an input of a specification item value used as the search condition;
A similar specification search unit that searches for specification data similar to input specification data composed of specification item values received by the input unit based on a learning result by the similarity learning unit;
Equipped with a,
The similarity learning unit is configured to determine a similarity between two specification item values based on the specification data and data of parts constituting each product corresponding to each of the specification data. The learning is performed by calculating the similarity,
The similar specification search unit performs the search based on the similarity between the item values.
A search device characterized by the above-mentioned. A search result display unit for displaying a search result by the similar specification search unit,
The search device according to claim 1, further comprising: The similar specification search unit, based on the learning result, how similar each specification item value received by the input unit and the corresponding specification item value of the specification data held in the product design database. Is calculated for all the specification data held by the product design database, the search is performed based on the calculated similarity,
The search result display unit displays a search result by the similar specification search unit together with the similarity corresponding to the search result.
The search device according to claim 2, wherein: The similar specification search unit weights the similarity between item values, and performs the search based on the similarity between item values after weighting.
The retrieval device according to claim 1 , wherein the retrieval device includes: A weight data changing unit that changes a weight coefficient used when the similar specification search unit weights the similarity between the item values;
The retrieval device according to claim 4 , comprising: A similarity adjustment unit that adjusts a learning result by the similarity learning unit;
With
The similar specification search unit performs the search based on the learning result after being adjusted by the similarity adjustment unit.
The search device according to any one of claims 1 to 5 , wherein: The similarity learning unit performs the learning with respect to a specified specification item value among specification item values included in the specification data held by the product design database.
Search device according to any one of claims 1 to 6, characterized in that. A special specification data acquisition unit that acquires text data representing the special specification of the product,
With
The similar specification search unit searches for specification data similar to the input specification data and the special specification,
The retrieval device according to any one of claims 1 to 7 , wherein: The similar specification search unit adjusts the learning result based on the text data, and performs the search based on the adjusted learning result.
9. The search device according to claim 8 , wherein: A search method executed by a search device that searches for specification data similar to a search condition from a specification data group including at least one specification item value indicating details of an item indicating a product specification. hand,
A similarity learning step of learning how similar two specification item values of the same item are based on the specification data group held by the product design database for all combinations of specification item values;
An input step of receiving an input of a specification item value used as the search condition;
A similar specification search step of searching for specification data similar to input specification data composed of specification item values received in the input step based on a learning result in the similarity learning step;
Only including,
In the similarity learning step, based on the specification data and data of parts constituting each product corresponding to each of the specification data, an item value indicating the degree of similarity between two specification item values is defined. The learning is performed by calculating the similarity,
In the similar specification search step, the search is performed based on the similarity between the item values.
A search method characterized in that: A machine learning device that configures a search device that searches for specification data similar to a search condition from a specification data group that includes one or more specification item values indicating details of an item representing a product specification. So,
A similarity learning unit that learns how similar two specification item values of the same item are based on the specification data group for all combinations of specification item values;
The similarity learning unit is configured to determine a similarity between two specification item values based on the specification data and data of parts constituting each product corresponding to each of the specification data. There line the learning by calculating the similarity, wherein the item value between the similarity obtained by the learning, the search based on, to be executed by the search device,
A machine learning device, characterized in that:

Images