Apparatus, method and program

Abstract

To provide an apparatus, method and program configured to select target features for a learning processing apparatus and causes the learning processing device to perform leaning.SOLUTION: In a system 1, an apparatus includes: a specifying unit which specifies a target parameter to be estimated, in accordance with the contents of interaction with a user; a selection control unit which causes a selection apparatus configured to select features to be used for estimating a value of a parameter, in accordance with the parameter, to select target features which are features according to the target parameter; and a learning control unit which causes a learning processing apparatus to perform learning processing of an estimation model configured to output an estimated value of the target parameter, in accordance with an input of a value of the target features, using learning data including a value indicated by the target parameter and a value indicated by the target features when the target parameter indicates the above value.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an apparatus, a method, and a program. [Background technology]

[0002] Patent Document 1 and other documents state, "When building a machine learning model, feature selection can have a significant impact on the accuracy and usefulness of the model. However, appropriately selecting features that improve the accuracy of the model without subject matter expertise and a deep understanding of the machine learning problem can be a challenge." (Patent Document 1, paragraph 0016) [Prior art document] [Patent documents] [Patent Document 1] Japanese Patent Publication No. 2023-534475 [Patent Document 2] Special Publication No. 2017-529583 [Patent Document 3] JP 2021-149138 A [Patent Document 4] JP 2009-30476 A [Patent Document 5] JP 2022-115643 A [Patent Document 6] JP ​​2015-230576 A [Patent Document 7] JP 2019-220226 A [Patent Document 8] JP 2022-169291 A Summary of the Invention

[0003] In a first aspect of the present invention, there is provided an apparatus including: an identification unit that identifies a target parameter to be estimated in accordance with the content of a dialogue with a user; a selection control unit that causes a selection device that selects a feature to be used to estimate a parameter value in accordance with the parameter to select a target feature that is a feature in accordance with the target parameter; and a learning control unit that causes a learning processing device to perform a learning process of an estimation model that outputs an estimated value of the target parameter in response to input of a value of the target feature, using learning data including a value indicated by the target parameter and a value indicated by the target feature when the target parameter indicates that value.

[0004] In the above device, the learning control unit may extract a portion of a data set from a plurality of data sets each including a value indicated by the target parameter and a value indicated by the target feature, and use the portion as the learning data.

[0005] In any of the above devices, the identification unit may include a dialogue acquisition unit that acquires an input sentence in natural language from a user, a generation unit that generates an output sentence in natural language corresponding to the input sentence, and a dialogue output unit that outputs the output sentence.

[0006] In any of the above devices, the identification unit may ask the user a question about an item, of a plurality of pre-registered items necessary for identifying the target parameter, about which information has not yet been obtained through dialogue.

[0007] In any of the above devices, the selection device has a user interface that identifies the target parameter in response to user operation, and the device may further include a first learning processing unit that performs learning processing of the identification unit using learning data that includes text displayed on the user interface of the selection device, guiding users in the content to be entered, and text entered into the user interface.

[0008] In the above device, the first learning processing unit may further perform reinforcement learning on the identification unit using learning data including the content of the dialogue by the identification unit and a reward value corresponding to an evaluation input by a user regarding the target feature or the estimation model.

[0009] Any of the above devices may further include a memory unit that, each time identification by the identification unit or selection by the selection device is completed, stores the content of a dialogue between the identification unit and a user in association with the target parameter identified in accordance with the content of the dialogue or at least one of the target feature quantities selected in accordance with the target parameter; a similarity calculation unit that calculates a similarity between each of a plurality of past dialogue contents stored in the memory unit and the content of the current dialogue; and a first output unit that outputs the at least one of the target parameter quantities or the target feature quantities stored in the memory unit in association with the past dialogue content that has the greatest similarity to the content of the current dialogue.

[0010] In any of the above devices, the target parameter may be a control parameter to be applied to any one of a plurality of devices included in the facility.

[0011] The above device may further include an operation result acquisition unit that acquires the results of operating the equipment using the value of the target parameter output from the estimation model in response to supplying the target feature to the estimation model by at least one of simulating the equipment or actually operating the equipment, an evaluation unit that calculates an evaluation value that evaluates the result acquired by the operation result acquisition unit, and a second output unit that outputs at least one of the target parameter, the target feature, or the estimation model learned by the learning processing device, on condition that the evaluation value exceeds a reference value.

[0012] In the above-described device, the operation result acquisition unit may acquire a result when the equipment is actually operated a reference number of times or for a reference period.

[0013] In any of the above-described devices, the target parameter may be an index value of the quality of a product produced by the equipment.

[0014] Any of the above devices may further include an extraction unit that extracts information shown using graphics in the image from image data related to the equipment, and a second learning processing unit that performs learning processing of the selection device using learning data including the information extracted by the extraction unit.

[0015] In a second aspect of the present invention, there is provided a method executed by a computer, comprising: an identification step of identifying a target parameter to be estimated in accordance with the content of a dialogue with a user; a selection control step of causing a selection device, which selects a feature to be used to estimate a parameter value in accordance with the parameter, to select a target feature that is a feature in accordance with the target parameter; and a learning control step of causing a learning processing device to perform a learning process of an estimation model that outputs an estimated value of the target parameter in response to input of a value of the target feature, using learning data including a value indicated by the target parameter and a value indicated by the target feature when the target parameter indicates that value.

[0016] In a third aspect of the present invention, there is provided a program that, when executed by a computer, causes the computer to function as: an identification unit that identifies a target parameter to be estimated in accordance with the content of a dialogue with a user; a selection control unit that causes a selection device that selects a feature to be used to estimate a parameter value in accordance with the parameter to select a target feature that is a feature in accordance with the target parameter; and a learning control unit that causes a learning processing device to perform a learning process for an estimation model that outputs an estimated value of the target parameter in response to input of a value of the target feature, using learning data including a value indicated by the target parameter and a value indicated by the target feature when the target parameter indicates that value.

[0017] The above summary of the invention does not list all of the necessary features of the present invention, and subcombinations of these features may also constitute inventions. [Brief explanation of the drawings]

[0018] [Figure 1]1 shows a system 1 according to a first embodiment. [Figure 2] The operation of device 3 is shown. [Figure 3] An example of a dialogue between the identification unit 301 and the user will be shown. [Figure 4] 10 shows another example of a dialogue between the identification unit 301 and a user. [Figure 5] 10 shows another example of a dialogue between the identification unit 301 and a user. [Figure 6] 10 shows another example of a dialogue between the identification unit 301 and a user. [Figure 7] 10 shows another example of a dialogue between the identification unit 301 and a user. [Figure 8] 10 shows another example of a dialogue between the identification unit 301 and a user. [Figure 9] 3 shows a device 3A according to a second embodiment. [Figure 10] The operation of device 3A is shown. [Figure 11] 12 illustrates an example computer 1200 in which aspects of the present invention may be embodied in whole or in part. DETAILED DESCRIPTION OF THE INVENTION

[0019] The present invention will be described below through embodiments of the invention, but the following embodiments do not limit the scope of the invention according to the claims. Furthermore, not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

[0020] First Embodiment 1 shows a system 1 according to this embodiment. The system 1 includes a facility 2 and a device 3.

[0021] (Equipment 2) The facility 2 may be a facility or device equipped with one or more devices 21. For example, the facility 2 may be a plant or a composite device incorporating multiple devices 21. Examples of plants include industrial plants such as chemical and biotech plants, plants that manage and control wellheads and surrounding areas of gas and oil fields, plants that manage and control hydroelectric, thermal, and nuclear power generation, plants that manage and control environmental power generation such as solar and wind power, and plants that manage and control water, sewage, and dams. The devices 21 installed in the facility 2 may be tools, machines, or devices, and may be so-called field devices. For example, the devices may be valve devices such as flow control valves and on-off valves, or actuator devices such as fans and motors. Each device 21 may be controllable by one or more control parameters that indicate control details. For example, the control parameters may indicate instruction values ​​for manipulated variables or target values ​​(also referred to as set values).

[0022] The facility 2 may be provided with one or more sensors 20. Each sensor 20 may measure a condition related to the facility 2, may measure a condition of the facility 2, or may measure a condition of a product produced by the facility 2. The measured condition may be a physical quantity, such as pressure, temperature, pH, speed, flow rate, weight, or volume. The condition of the product may be a physical quantity related to the quality of the product, such as the percentage of impurities contained in the product. The sensors 20 may be of different types, or at least some of the two or more sensors 20 may be of the same type. Each sensor 20 may supply a measurement value to the device 3. Communication between the sensor 20 and the device 3 may be performed using, for example, a wireless communication protocol of the International Society of Automation (ISA), such as ISA100, Highway Addressable Remote Transducer (HART) (registered trademark), BRAIN (registered trademark), FOUNDATION Fieldbus, or PROFIBUS.

[0023] (device 3) The device 3 supports the operation of the facility 2 and includes a memory unit 300, an identification unit 301, a selection device 302, a selection control unit 303, a learning processing device 304, a learning control unit 305, and an estimation unit 306.

[0024] ((Storage unit 300)) The storage unit 300 stores various types of information related to the facility 2. The storage unit 300 may store operation data 3001.

[0025] (((Operation Data 3001))) The operation data 3001 includes data related to the operation of the equipment 2. The operation data 3001 may include a plurality of data sets that associate states related to the equipment 2 (for example, measured values ​​measured by each sensor 20) with control parameters of each device 21 in the corresponding states.

[0026] ((Specific Section 301)) The identification unit 301 identifies a target parameter to be estimated from among a plurality of parameters related to the facility 2. The identification unit 301 identifies a target parameter to be estimated according to the content of a dialogue with a user. The identification unit 301 includes a dialogue acquisition unit 3011, a generation unit 3012, and a dialogue output unit 3013.

[0027] Here, the target parameter may be a control parameter to be applied to any one of the multiple devices 21 included in the equipment 2. The control parameter to be applied may be a control parameter to be applied to improve the quality of the product or a KPI (Key Performance Indicator) of the equipment 2. The KPI may be, for example, the yield of the product or production efficiency.

[0028] The target parameter may be a parameter indicating a state related to the equipment 2. For example, the target parameter may be an index value of the quality of the product produced by the equipment 2, or may be a KPI of the equipment 2. The target parameter may indicate whether the equipment 2 or the device 21 is normal or abnormal (or whether there is a sign of an abnormality), and may be a so-called health index. The target parameter may be a physical quantity such as temperature or pressure at a location within the equipment 2, and may be a value measured by the sensor 20 or a value not measured. The physical quantity not measured by the sensor 20 may be, for example, the temperature at a location where the sensor 20 is not installed, or may be a value calculated using measurements from one or more sensors 20 (for example, power consumption, energy consumption, or CO2 emissions).

[0029] (((Dialogue text acquisition unit 3011))) The dialogue acquisition unit 3011 acquires an input sentence in a natural language from a user. The dialogue acquisition unit 3011 according to this embodiment may acquire text input by the user via an input device (not shown) such as a keyboard as the input sentence. The dialogue acquisition unit 3011 may convert speech input by the user via a microphone into text by speech recognition and acquire the text as the input sentence. The dialogue acquisition unit 3011 may supply the acquired input sentence to the generation unit 3012.

[0030] (((Generation unit 3012))) The generation unit 3012 generates an output sentence in a natural language according to the input sentence. The generation unit 3012 may be a generation AI that generates output content according to a user input, or may be a dialogue AI that dialogues with a user via the dialogue acquisition unit 3011 and the dialogue output unit 3013.

[0031] The generation unit 3012 may generate a question for the user as an output sentence. The generation unit 3012 may supply the generated output sentence to the dialogue output unit 3013 for output, and may identify a target parameter according to the input sentence supplied from the dialogue acquisition unit 3011. The generation unit 3012 may supply identification information of the identified target parameter to the selection control unit 303.

[0032] The generation unit 3012 may store a plurality of pre-registered items (also referred to as identifying items) necessary for identifying the target parameter, and may output an output statement asking a question about an identifying item of the plurality of identifying items for which information has not yet been acquired through dialogue, and may acquire information about the identifying item from a new input statement, thereby identifying the target parameter. The generation unit 3012 may narrow down a plurality of parameters related to the equipment 2 according to information about each identifying item included in the input statement, and identify any one parameter as the target parameter. The generation unit 3012 may estimate the content of the identifying item and generate an output statement asking whether the estimated content is correct, or may generate an output statement asking which of candidate content of the identifying item is correct.

[0033] ((((specific item)))) The identification items may include the type of target parameter. The type of target parameter may be a rough classification of the target parameter, or may be an item for identifying whether the target parameter is a control parameter, a product quality index value, a KPI, a health index, or a physical quantity within the equipment 2. The identification items may include at least one of a first identification item for directly identifying the target parameter and a second identification item for indirectly identifying the target parameter (in this embodiment, both are used as an example). The second identification item may be an item for identifying an object that shows a correlation with the target parameter.

[0034] Table 1 shows the first and second identification items for each type of target parameter. As shown in this table, the contents of the first and second identification items may differ depending on the type of target parameter.

[0035] [Table 1]

[0036] (((((When the target parameter is a control parameter))))) If the target parameter is a control parameter of the device 21, the first identification item may include an item for identifying the type of control parameter, or may include an item for identifying the device 21 to which the control parameter is applied (for example, at least one of the type, model number, individual identification number, or installation location of the device 21).

[0037] When the target parameter is a control parameter of the device 21, the second identification item may include an item for identifying other parameters (as an example in the present embodiment, other parameters affected by the target parameter) that show a correlation with the control parameter as the target parameter, among parameters related to the equipment 2. The other parameters may be parameters that the user wishes to improve in the dialogue.

[0038] For example, the second identification items may include items for specifying the type of physical quantity and the location where the physical quantity is generated (e.g., measurement location) as items for specifying a physical quantity affected by the target parameter (e.g., pressure affected by the valve opening). The second identification items may include items for specifying the type of KPI as items for specifying a KPI affected by the target parameter. The second identification items may include items for specifying the type of product as items for specifying the quality of the product affected by the target parameter. The second identification items may include items for specifying the type of health index (e.g., a type corresponding to the target whose state is indicated by the health index) as items for specifying a health index affected by the target parameter.

[0039] When the identification items include second identification items for identifying a physical quantity, KPI, product quality, or health index that is affected by the target parameter, the generation unit 3012 may be capable of identifying control parameters that show a correlation with these values ​​among the control parameters of each device 21. As an example in the present embodiment, the generation unit 3012 may be capable of identifying control parameters that have a large influence on these values. In this way, the generation unit 3012 may identify a physical quantity, KPI, or quality that the user desires to improve, and identify, as the target parameter, a control parameter that has a large influence on the identified physical quantity, KPI, or quality. The influence may be calculated, for example, by calculating a correlation coefficient or performing principal component analysis on the operation data 3001.

[0040] (((((When the target parameter is the quality index value))))) When the target parameter is a quality index value of a product, the first identification item may include an item for identifying the type of product as an item for identifying the quality index value. The second identification item may include an item for identifying other parameters related to the equipment 2 that show a correlation with the quality index value as the target parameter. In the present embodiment, as an example, the second identification item may include an item for identifying a control parameter that affects the quality index value (for example, at least one of the type of control parameter, the type of device 21, the model number, the individual identification number, and the installation location).

[0041] When the identification items include a second identification item for identifying a control parameter that affects quality, the generation unit 3012 may be able to identify the quality of the product that is significantly affected by each control parameter using a conventionally known method. Thus, the generation unit 3012 may identify the control parameter desired to be adjusted by the user, thereby identifying an index value of the product quality that is suitable for evaluating the adjustment results as the target parameter. Note that, when a quality affected by a control parameter is identified as the target parameter by identifying a control parameter, the generation unit 3012 may supply identification information of the control parameter to the selection control unit 303.

[0042] (((((When the target parameter is KPI))))) When the target parameter is a KPI of the equipment 2, the first identification item may include an item for identifying the type of KPI as an item for identifying the KPI. The second identification item may include an item for identifying other parameters that show a correlation with the KPI as the target parameter, among parameters related to the equipment 2. In the present embodiment, as an example, the second identification item may include an item for identifying a control parameter that affects the KPI (for example, at least one of the type of control parameter, the type of device 21, model number, individual identification number, and installation location).

[0043] When the identification items include a second identification item for identifying a control parameter that affects a KPI, the generation unit 3012 may be able to identify a KPI that is significantly affected by each control parameter using a conventionally known method. Thus, the generation unit 3012 may identify a control parameter that the user desires to adjust, and thereby identify a KPI that is suitable for evaluating the adjustment results as a target parameter. Note that, when a KPI that is affected by a control parameter is identified as a target parameter by identifying the control parameter, the generation unit 3012 may supply identification information of the control parameter to the selection control unit 303.

[0044] (((((When the target parameter is the health index))))) When the target parameter is a health index, the first identification item may include an item for identifying the type of health index (for example, a type corresponding to the target whose state is indicated by the health index) as an item for identifying the health index. The second identification item may include an item for identifying other parameters related to the equipment 2 that show a correlation with the health index as the target parameter. In the present embodiment, as an example, the second identification item may include an item for identifying a control parameter that affects the health index (for example, at least one of the type of control parameter, the type of device 21, model number, individual identification number, and installation location).

[0045] When the identification items include a second identification item for identifying a control parameter that affects the health index, the generation unit 3012 may be able to identify a health index that is significantly affected by each control parameter using a conventionally known method. Thus, the generation unit 3012 may identify a control parameter that the user desires to adjust, thereby identifying a health index suitable for evaluating the adjustment results as a target parameter. Note that when a health index affected by a control parameter is identified as a target parameter by identifying the control parameter, the generation unit 3012 may supply identification information for the control parameter to the selection control unit 303.

[0046] (((((When the target parameter is a physical quantity))))) When the target parameter is a physical quantity within the equipment 2, the first identification item may include an item for identifying the type of the physical quantity and an item for identifying the location where the physical quantity is generated, as an item for identifying the physical quantity. The second identification item may include an item for identifying another parameter that correlates with the physical quantity as the target parameter, among parameters related to the equipment 2. In the present embodiment, as an example, the second identification item may include an item for identifying a control parameter that affects the physical quantity (for example, at least one of the type of control parameter, the type, model number, individual identification number, and installation location of the device 21).

[0047] When the identification items include a second identification item for identifying a control parameter that affects a physical quantity, the generating unit 3012 may be able to identify a physical quantity that is significantly affected by each control parameter by a conventionally known method. Thus, the generating unit 3012 may identify a control parameter that the user desires to adjust, thereby identifying a physical quantity that is suitable for evaluating the adjustment result as a target parameter. Note that, when a physical quantity affected by a control parameter is identified as a target parameter by identifying the control parameter, the generating unit 3012 may supply identification information of the control parameter to the selection control unit 303.

[0048] (((Dialogue sentence output unit 3013))) The dialogue output unit 3013 outputs the output sentence generated by the generation unit 3012. The dialogue output unit 3013 may output the output sentence as text via a display device (not shown) such as a display. The dialogue output unit 3013 may output the output sentence by voice via a speaker. In response to the output sentence being output from the dialogue output unit 3013, a new input sentence may be input by the user to the dialogue acquisition unit 3011.

[0049] ((Selection device 302)) The selection device 302 selects a feature to be used for estimating the value of a parameter (in this embodiment, as an example, an arbitrary parameter related to the equipment 2) in accordance with the parameter. The selection device 302 may select a feature in accordance with the type of parameter. In response to a parameter being designated from among a plurality of parameters related to the equipment 2, the selection device 302 may select a feature to be used for estimating the value of the parameter. As an example, the selection device 302 may have a user interface (not shown) that identifies a target parameter in response to a user operation, and may select a feature in response to a parameter being designated by the user.

[0050] The feature quantity may be a value indicating a state of the facility 2, or may be a measurement value from any of the sensors 20. The feature quantity may be a control parameter of any of the devices 21 of the facility 2. The selection device 302 may select a single feature quantity or may select multiple feature quantities. When a measurement value or a control parameter to be used as a feature quantity is designated by the user, the selection device 302 may include the measurement value or the control parameter in the feature quantity.

[0051] The selection device 302 may select a feature to be used to estimate a control parameter to be applied to any device 21 in facility 2 by specifying the control parameter. The selection device 302 may select a feature to be used to estimate the quality or KPI of a product produced by facility 2 by specifying the quality or KPI. The selection device 302 may select a feature to be used to estimate a health index by specifying a parameter. The selection device 302 may select a feature to be used to estimate a physical quantity such as temperature or pressure at a location in facility 2 by specifying the physical quantity.

[0052] The selection device 302 may select, as a feature, any of the multiple parameters related to the equipment 2 that may affect the specified parameter. The selection device 302 may select the feature based on the results of a conventionally known analysis performed on the operation data 3001, such as calculation of a correlation coefficient or principal component analysis. As an example, the selection device 302 may calculate a correlation coefficient between the time-series data of each of the multiple parameters included in the operation data 3001 and the specified parameter, and select the parameter with the highest correlation coefficient as the feature, or may display the parameters in descending order of correlation coefficient and select the parameter selected by the user as the feature. Additionally or alternatively, the selection device 302 may be a device that applies, for example, "Sizca," a plant AI analysis tool manufactured by Yokogawa Electric Corporation.

[0053] The selection device 302 may select the feature quantities under the control of a selection control unit 303 described below. In response to identification information of the target parameters being supplied from the selection control unit 303, the selection device 302 may select target feature quantities that are feature quantities corresponding to the target parameters. The selection device 302 may supply identification information of the target feature quantities selected under the control of the selection control unit 303 to the learning control unit 305.

[0054] ((Selection control unit 303)) The selection control unit 303 causes the selection device 302 to select a target feature quantity according to the target parameter. The selection control unit 303 may supply identification information of the target parameter to the selection device 302 to cause it to select a target feature quantity. When the selection control unit 303 receives identification information of a control parameter in addition to the identification information of the target parameter from the identification unit 301 (that is, when the identification unit 301 identifies a control parameter of any device 21 and thereby identifies a quality, KPI, or physical quantity affected by the control parameter as a target parameter), the selection control unit 303 may instruct the selection device 302 to include the control parameter in the target feature quantities. The selection control unit 303 may cause the selection device 302 to supply identification information of the target feature quantity to the learning control unit 305.

[0055] ((Learning processing device 304)) The learning processing device 304 performs a learning process for the estimation model 3040. The estimation model 3040 may output estimated values ​​of parameters in response to input values ​​of feature quantities. The learning processing device 304 may perform the learning process for the estimation model 3040 using a plurality of learning data sets for a predetermined combination of a parameter and a feature quantity, the learning data sets including values ​​indicated by the parameters and values ​​indicated by the feature quantities when the parameters indicate the corresponding values. The learning processing device 304 may perform the learning process using an algorithm such as kernel dynamic policy programming, reinforcement learning, support vector machines, logistic regression, decision trees, or neural networks.

[0056] The learning processing device 304 may perform learning processing of an estimation model 3040 that outputs an estimated value of a target parameter in response to input of a value of a target feature amount, under the control of a learning control unit 305 described below.

[0057] ((Learning control unit 305)) The learning control unit 305 causes the learning processing device 304 to perform a learning process for an estimation model 3040 that outputs an estimated value of a target parameter in response to input of a value of the target feature. The learning control unit 305 may supply the learning processing device 304 with learning data including a value indicated by the target parameter and a value indicated by the target feature when the target parameter indicates that value, to cause the learning processing device 304 to perform a learning process for the estimation model 3040. The learning control unit 305 may cause the estimation unit 306 to supply the estimation model 3040 generated by the learning processing device 304.

[0058] When the target parameter is a control parameter of the device 21, the learning control unit 305 may cause the learning processing device 304 to perform a learning process using reinforcement learning. The learning control unit 305 may cause the learning process to be performed using the learning data and a reward value determined by a preset reward function. For example, the reward function may be a function in which the reward value increases as the value of a preset KPI increases. When the specifying unit 301 specifies a control parameter as the target parameter in response to a user's desire to improve any physical quantity, KPI, or product quality of the equipment 2, and identification information indicating the physical quantity, KPI, or quality is supplied from the specifying unit 301, the learning control unit 305 may cause the learning processing device 304 to perform reinforcement learning using an evaluation function in which the reward value increases as the value of the physical quantity, KPI, or quality increases.

[0059] The learning control unit 305 may extract some data sets from multiple data sets each containing a value indicated by a target parameter and a value indicated by a target feature, and use the extracted data as learning data. For example, the learning control unit 305 may extract some data sets from multiple data sets included in the driving data 3001 stored in the storage unit 300, and use the extracted data as learning data. The learning control unit 305 may extract a data set with a reference data amount or a data set for a reference period from the multiple data sets included in the driving data 3001. The reference data amount may be a data amount preset by a user. The reference data amount may be a data amount at which the learning processing device 304 completes the learning process within a time span preset by a user. When extracting a data set with a reference data amount, the learning control unit 305 may extract a data set that is continuous in time, or may extract a data set that is discrete in time (for example, a data set for each reference interval). The reference period may be a period preset by a user.

[0060] ((Estimation section 306)) The estimation unit 306 estimates the value of the target parameter using the estimation model 3040 supplied from the learning processing device 304. As an example, the estimation unit 306 may be equipped with the estimation model 3040 internally to perform the estimation.

[0061] The estimation unit 306 may supply the value of the target feature to the estimation model 3040, and in response thereto, output the value output from the estimation model 3040 as an estimated value of the target parameter. The estimation unit 306 may display the estimated value of the target parameter on a display device (not shown), such as a display. When the target parameter is a control parameter to be applied to any of the devices 21, the estimation unit 306 may supply the estimated value of the target parameter, i.e., the value estimated to be applied, to the corresponding device 21. In this case, the estimation unit 306 may function as a control unit for the facility 2. Note that FIG. 1 illustrates a case in which the estimation model 3040 outputs a control parameter in response to input of a measured value of a physical quantity as the target feature.

[0062] According to the above-described device 3, a target parameter to be estimated is identified in response to the content of a dialogue with a user, the selection device 302 is caused to select a target feature corresponding to the target parameter, and the learning processing device 304 is caused to perform a learning process for an estimation model 3040 that outputs an estimated value of the target parameter in response to an input value of the target feature. Therefore, by engaging in a dialogue with the device 3, a series of processes can be automatically performed, including identifying a target parameter to be estimated, selecting a target feature to be used for estimating the target parameter, and performing a learning process for an estimation model 3040 that outputs an estimated value of the target parameter. Furthermore, because the target parameter can be identified in response to the dialogue, even if the user himself does not recognize the target parameter, an appropriate target parameter can be identified and the estimation model 3040 can be generated.

[0063] Furthermore, some data sets are extracted from multiple data sets each containing a value indicated by a target parameter and a value indicated by a target feature, and are used as training data. Therefore, the load of the training process can be reduced compared to when all data sets are used for training. Furthermore, by narrowing down the data sets and using them for training, the estimation accuracy of the estimation model 3040 can be improved.

[0064] Furthermore, an input sentence in natural language is acquired from the user, and an output sentence is generated and output in natural language, so that a user can interact with the device 3 in natural language to specify target parameters.

[0065] Furthermore, since the target parameter is a control parameter to be applied to any of the devices included in the facility 2, it is possible to select a target feature to be used for estimating the value of the control parameter to be applied, and perform a learning process for the estimation model 3040. Furthermore, by applying the target parameter output from the learning-processed estimation model 3040 to the facility 2, it is possible to improve the operating state of the facility 2.

[0066] Furthermore, since the target parameter is an index value of the quality of the product produced by equipment 2, the target feature to be used to estimate the index value of the quality of the product can be selected and the learning process of the estimation model 3040 can be performed.

[0067] (operation) 2 shows the operation of the device 3. The device 3 supports the operation of the facility 2 by performing the processes of steps S11 to S15.

[0068] In step S11, the identification unit 301 identifies a target parameter to be estimated according to the content of the dialogue with the user. The identification unit 301 may ask the user about a pre-registered identification item necessary for identifying the target parameter, about which information has not yet been acquired through dialogue.

[0069] The identification unit 301 may further identify a period for which the value of the target parameter should be estimated, that is, a target period for estimation, from the content of the dialogue with the user. The target period for estimation may be a specific period of a day (for example, 9:00 to 17:00 or 21:00 to 5:00), a specific period of a week (for example, Monday to Friday), a specific period of a year (for example, June to August or summer), or a period determined by a combination thereof. The identification unit 301 may identify the target period by outputting an output sentence asking about the target period for estimation from the dialogue sentence output unit 3013 and acquiring a new input sentence.

[0070] In step S13, the selection control unit 303 causes the selection device 302 to select a target feature according to the target parameter. The selection control unit 303 may display the selected target feature to the user and allow the user to select some of the target feature. The selection control unit 303 may end the processing of step S13 in response to an approval operation from the user.

[0071] In step S15, the learning control unit 305 causes the learning processing device 304 to perform a learning process for the estimation model 3040 using learning data including a value indicated by the target parameter and a value indicated by the target feature when the target parameter indicates that value. This enables the estimation unit 306 to estimate the target parameter using the estimation model 3040. Note that the learning control unit 305 may extract a dataset for a reference period from multiple datasets included in the driving data 3001 and use it as learning data. If the target period for estimation is identified by the identification unit 301 in step S11, the learning control unit 305 may set the target period as the reference period. As a result, a dataset acquired during the target period is extracted and used for the learning process.

[0072] According to the above operation, a plurality of identification items necessary for identifying the target parameter are stored, and the user is asked about the items of the plurality of pieces of information for which information has not yet been acquired through dialogue. Therefore, information on each item necessary for identifying the target parameter can be acquired, and the target parameter can be reliably identified.

[0073] (Example of a dialogue) Fig. 3 shows an example of a dialogue between the identification unit 301 and a user. Note that in Fig. 3 through Fig. 7 described later, a dialogue output from the identification unit 301 is indicated by "AI:", and a dialogue input by the user is indicated by "User:".

[0074] In the example shown in the figure, the user is asked whether they wish to use the generated estimation model 3040 for operation improvement, quality prediction, anomaly (sign) detection, or numerical prediction. If the user replies that they wish to improve operation, the type of target parameter is determined to be a control parameter. Then, through a dialogue between the user and the identification unit 301, "control parameter a" of "device A" is identified as the target parameter.

[0075] 4 shows another example of a dialogue between the identification unit 301 and a user. In this example, the user is asked whether he or she desires operation improvement, quality prediction, anomaly (sign) detection, or numerical prediction using the generated estimation model 3040. If the user replies that he or she desires quality prediction, an index value of the quality of one of the products is designated as the target parameter. Then, through the dialogue between the user and the identification unit 301, the index value of the quality of "Product X" is designated as the target parameter.

[0076] FIG. 5 shows another example of a dialogue between the identification unit 301 and a user. In this example, the user is asked whether they wish to use the generated estimation model 3040 for operation improvement, quality prediction, anomaly (sign) detection, or numerical prediction. If the user replies that they wish to use anomaly detection, the type of target parameter is determined to be the health index. Then, through the dialogue between the user and the identification unit 301, the health index of "pump a" is identified as the target parameter. In this example, through the dialogue between the user and the identification unit 301, summer is identified as the target period for estimation.

[0077] 6 shows another example of a dialogue between the identifying unit 301 and a user. In the example shown in this figure, the user is asked whether they wish to have the generated estimation model 3040 perform operation improvement, quality prediction, anomaly (sign) detection, or numerical prediction. If the user replies that they wish to perform numerical prediction, the type of target parameter is determined to be a physical quantity within facility 2. Then, through the dialogue between the user and the identifying unit 301, power consumption in "area S" of facility 2 is identified as the target parameter. In the example shown in this figure, through the dialogue between the user and the identifying unit 301, summer is identified as the target period for estimation.

[0078] 7 shows another example of a dialogue between the identification unit 301 and a user. In the example shown in this figure, the dialogue between the user and the identification unit 301 identifies the health index of the pump with "individual identification number 001" in "Area S" of facility 2 as the target parameter. Also, in the example shown in this figure, the dialogue between the user and the identification unit 301 identifies the period from June to August as the target period for estimation.

[0079] 8 shows another example of a dialogue between the identification unit 301 and a user. In the example shown in this figure, the quality index value of "product a" is identified as the target parameter through the dialogue between the user and the identification unit 301. Note that in the example shown in this figure, the control parameter of "device X" may be selected as the target feature through the dialogue between the user and the identification unit 301.

[0080] Second Embodiment 9 shows a device 3A according to the second embodiment. The device 3A includes a driving result acquisition unit 311, an evaluation unit 312, a result output unit 313, a storage unit 300A, a first learning processing unit 314, a similarity calculation unit 315, a similarity result output unit 316, an extraction unit 317, and a second learning processing unit 318. In the device 3A according to the second embodiment, components that are substantially the same as those in the device 3 shown in FIG. 1 are designated by the same reference numerals, and descriptions thereof will be omitted.

[0081] (Driving result acquisition unit 311) When the target parameter is a control parameter, that is, when a control parameter to be applied to any of the devices 21 is output from the estimation model 3040, the operation result acquisition unit 311 acquires a result of operating the equipment 2 using the estimation model 3040. The operation result acquisition unit 311 according to this embodiment may be a simulator of the equipment 2 and may be capable of calculating at least one estimated value indicating a state of the equipment 2. The operation result acquisition unit 311 may acquire, by simulating the equipment 2, a result of operating the equipment 2 using the value of the target parameter (here, a control parameter to be applied to any of the devices 21) output from the estimation model 3040 in response to supplying the target feature to the estimation model 3040. The operation result acquisition unit 311 may acquire the operation result of the equipment 2 using the estimation model 3040 that has been subjected to a learning process by the learning processing device 304 and has not been output to the estimation unit 306. The operation result acquisition unit 311 may acquire the operation result by supplying a target feature amount corresponding to one state (for example, a target feature amount corresponding to the current state) to the estimation model 3040 and applying the value of the target parameter output from the estimation model 3040 to the simulated facility 2. The operation result acquisition unit 311 may supply the acquired operation result to the evaluation unit 312.

[0082] (Evaluation unit 312) The evaluation unit 312 calculates an evaluation value by evaluating the results acquired by the operation result acquisition unit 311 (i.e., the operation results of the equipment 2). The evaluation unit 312 may calculate the evaluation value by evaluating one or more estimated values ​​that indicate the state of the equipment 2. As an example, the evaluation unit 312 may calculate the evaluation value using an evaluation function that has each estimated value as a variable. The evaluation value may be a KPI of the equipment 2. The evaluation unit 312 may supply the evaluation value to the result output unit 313.

[0083] (Result output unit 313) The result output unit 313 outputs the results of the operation by the device 3A. The result output unit 313 may be interposed between the learning processing device 304 and the estimation unit 306, and may acquire the estimation model 3040 learned by the learning processing device 304 under the control of the learning control unit 305 and output it to the estimation unit 306. The result output unit 313 is an example of a second output unit, and may output the estimation model 3040 that outputs a control parameter on the condition that the evaluation value calculated by the evaluation unit 312 exceeds a reference value. The reference value for the evaluation value may be set to an arbitrary value in advance.

[0084] (Storage unit 300A) The storage unit 300A stores support data 3002 and facility data 3003.

[0085] ((Support Data 3002)) The assistance data 3002 is data related to assistance provided by the device 3A and may include a plurality of data sets in which the content of the dialogue between the identification unit 301 and the user, the target parameters identified by the identification unit 301 in accordance with the content of the dialogue, and the target feature amounts selected by the selection device 302 in accordance with the target parameters are associated with each other. Each data set may be stored each time the identification by the identification unit 301 and the selection by the selection device 302 are completed. These data sets may be associated with the target feature amounts selected by the selection device 302 or with a user's evaluation of the estimation model 3040 that has been subjected to a learning process by the learning processing device 304. The user's evaluation may be input to the device 3A via an input unit (not shown). The user's evaluation may be input by the user who has confirmed the target feature amounts, or may be input by the user who has confirmed the driving results using the estimation model 3040. The user's evaluation may be an evaluation of the target parameters identified by the identification unit 301.

[0086] The assistance data 3002 may further include a plurality of data sets associating input and output texts of the selection device 302. The output text of the selection device 302 may be text displayed on the user interface of the selection device 302 to guide the content to be input, or may be text pre-registered in the selection device 302. The input text of the selection device 302 may be text input into the user interface of the selection device 302, or may be text input by the user.

[0087] The input / output text of the selection device 302 may indicate input / output when the selection device 302 selects features in response to user input without being controlled by the selection control unit 303. The text guiding the content to be input in the selection device 302 may be text that prompts the user to input parameters to be estimated, or may be text that prompts the user to input information necessary for the selection device 302 to identify the parameters to be estimated. The text input to the user interface of the selection device 302 may be text indicating parameters to be estimated, text indicating information necessary for the selection device 302 to identify the parameters to be estimated, or text unrelated to these. A dataset associating these texts may be useful for having the identification unit 301 learn question content that is effective for identifying parameters to be estimated.

[0088] ((Facility Data 3003)) The equipment data 3003 is data relating to the structure and operation of the equipment 2, and includes the process flow, operation procedures, piping locations, maintenance information, and the like.

[0089] (First learning processing unit 314) The first learning processing unit 314 performs a learning process for the identification unit 301. The first learning processing unit 314 may perform a learning process for the identification unit 301 (in this embodiment, as an example, the generation unit 3012 of the identification unit 301) using learning data including a dataset in which input and output texts of the selection device 302 are associated, among the datasets included in the assistance data 3002. This may allow the identification unit 301 to learn to generate questions that are effective for identifying target parameters. The learning process using the learning data including the dataset of input and output of the selection device 302 may be executed in the generation stage of the identification unit 301.

[0090] The first learning processing unit 314 may further perform reinforcement learning on the identification unit 301 using learning data including the dialogue content of the identification unit 301 and a target feature or a reward value according to an evaluation input by the user for the estimation model 3040. In the present embodiment, as an example, the first learning processing unit 314 may perform reinforcement learning using a data set associated with the evaluation input by the user, among data sets included in the assistance data 3002. In this way, the first learning processing unit 314 may perform learning processing on the identification unit 301 so as to increase the evaluation from the user. The reinforcement learning of the identification unit 301 may be executed on the identification unit 301 that has identified a target parameter through dialogue with the user.

[0091] (Similarity calculation unit 315) The similarity calculation unit 315 calculates the similarity between each of the multiple past dialogue contents stored in the storage unit 300A and the current dialogue content. The similarity calculation unit 315 may acquire, from the assistance data 3002 in the storage unit 300A, the past dialogue content with the user each time the target parameter is identified by the identification unit 301, or may acquire the current dialogue content from the identification unit 301. The current dialogue content may be the content of a dialogue that started after the last time the target parameter was identified, or the content of a dialogue before the target parameter was identified. The similarity between the dialogue contents may be calculated from the similarity between the contents of the identification items acquired from the dialogue. As an example, the similarity between the dialogue contents may be the average (or weighted average) of the semantic similarity between the contents of the acquired identification items. The similarity between the dialogue contents may be the similarity between a feature vector calculated from the past dialogue content and a feature vector calculated from the current dialogue content. A feature vector may be calculated, for example, by assigning a word that may appear in a conversation (for example, each word in a dictionary) to each of the multiple bits contained in the vector, and setting the bit of a word that appears in the conversation to 1 and the bit of a word that does not appear to 0.

[0092] (Similar result output unit 316) The similarity result output unit 316 is an example of a first output unit and outputs one of the previous assistance results from the device 3A. The similarity result output unit 316 outputs target parameters and target features stored in the assistance data 3002 of the storage unit 300A in association with the previous dialogue content that has the highest similarity to the current dialogue content. The similarity result output unit 316 may identify the dialogue content with the highest similarity calculated by the similarity calculation unit 315 from among the dialogue contents in the assistance data 3002 stored in the storage unit 300A, and output the target parameters and target features associated with the identified dialogue content. This allows the user to understand what target parameters and target features are included in the training data used to train the estimation model 3040 that was previously trained using similar dialogue content. Furthermore, by knowing the target parameters and target features previously identified using similar dialogue content, the user can determine whether or not the same target parameters will be identified in the current dialogue.

[0093] The similar result output unit 316 may output the past target parameters and target feature amounts via a display device (not shown) such as a display. The similar result output unit 316 may output the past target parameters and target feature amounts separately from the output sentence of the dialogue by the identification unit 301. The similar result output unit 316 may output the past target parameters and target feature amounts as an output unit within the dialogue by the identification unit 301. In this case, the similar result output unit 316 may cause the generation unit 3012 to generate an output sentence indicating the past target parameters and target feature amounts, and cause the dialogue sentence output unit 3013 to output the output sentence.

[0094] (Extraction part 317) The extraction unit 317 extracts information from image data related to the facility 2. The image data related to the facility 2 may indicate, for example, a PFD (Process Flow Diagram) of a process executed in the facility 2, an operation manual for the facility 2 or the equipment 21, an inspection and maintenance record book for the facility 2 or the equipment 21, a P&ID (Piping & Instrumentation Diagram) of the facility 2, a design drawing of the facility 2, etc.

[0095] The extraction unit 317 may extract information represented by graphics in the image from the image data, and may further extract information represented by characters or symbols through character recognition. For example, the extraction unit 317 may perform pattern recognition on graphics in the image to extract the meaning of the graphics and the relationships between the graphics. For example, the extraction unit 317 may extract the content and sequence of a process from a PFD. The extraction unit 317 may extract the operating procedures of the devices 21, the relationships between the devices 21 (e.g., positional relationships), the positions of the sensors 20, and the physical quantities to be measured from an operation manual. The extraction unit 317 may extract the flow direction in the piping, the connection relationships between the piping, the connection relationships between the devices 21, and the positional relationships between the piping, the devices 21, and the sensors 20 within the facility 2 from a P&ID or the like. The extraction unit 317 may extract the maintenance history of the devices 21 from an inspection and maintenance logbook. The extraction unit 317 may extract, from the design drawing of the facility 2, the connection relationships of the pipes, the connection relationships between the devices 21, the positional relationships of the pipes, the devices 21, and the sensors 20 within the facility 2, the positional relationships of the air conditioning equipment, and the like. The extraction unit 317 may store the extracted information in the storage unit 300 as facility data 3003.

[0096] (Second learning processing unit 318) The second learning processing unit 318 performs a learning process for the selection device 302 using learning data including the information extracted by the extraction unit 317. The second learning processing unit 318 may perform a learning process for the selection device 302 using learning data including facility data 3003. This may allow the selection device 302 to learn so that, from among multiple target feature quantities corresponding to the target parameters, a target feature quantity corresponding to information indicated using a graphic in an image (for example, the positional relationship between the piping and the equipment 21 in the facility 2) is selected. The second learning processing unit 318 may perform a learning process for the selection device 302 using learning data further including the operation data 3001. The second learning processing unit 318 may perform a learning process for the selection device 302 by unsupervised learning, or may perform a learning process for the selection device 302 by reinforcement learning. When a learning process by reinforcement learning is performed, a value corresponding to a user evaluation of the target feature quantity selected by the selection device 302 or the estimation model 3040 trained using the target feature quantity may be used as a reward value. This allows the selection device 302 to learn to use information within the image to select target features that are appropriate for the user.

[0097] According to the above-described device 3A, the selection device 302 has a user interface that identifies a target parameter in response to a user operation, and the identification unit 301 performs a learning process using learning data that includes text displayed on the user interface that guides users to enter information and text entered into the user interface. Therefore, the identification unit 301 can be trained to effectively acquire information required to identify a target parameter from the user.

[0098] Furthermore, the identification unit 301 performs further reinforcement learning using learning data including the dialogue content by the identification unit 301 and a reward value according to the evaluation input by the user for the target feature or the estimation model 3040. Therefore, the identification unit 301 can be trained to identify target parameters from which target feature appropriate for the user is selected, or to identify target parameters from which an estimation model 3040 appropriate for the user is generated.

[0099] Furthermore, the content of the dialogue with the user by the identification unit 301, the target parameter identified according to the content of the dialogue, and the target feature selected according to the target parameter are stored in association with each other, and the target parameter and the target feature stored in association with the content of the previous dialogue that has the greatest similarity to the content of the current dialogue are output. Therefore, it is possible to understand what kind of learning data is being used to train the estimation model 3040 that has undergone training processing using similar content of the dialogue. Furthermore, by knowing the target parameter identified in the past, it is possible to identify that parameter as the current target parameter or not to identify it as the current target parameter.

[0100] Furthermore, an evaluation value of the result of operating the equipment 2 is calculated using the value of the target parameter output from the estimation model 3040 in response to supplying the target feature to the estimation model 3040, and the trained estimation model 3040 is output on the condition that the evaluation value exceeds a reference value. Therefore, when an estimation model 3040 with a low evaluation of the operation result is generated, it is possible to prevent the output of the estimation model 3040.

[0101] Furthermore, information shown using graphics in the image is extracted from the image data of the facility 2, and learning data including the extracted information is used to perform the learning process of the selection device 302. Therefore, the information shown using graphics in the image can be used to perform the learning process of the selection device 302 so that appropriate target feature quantities are selected.

[0102] (operation) 10 shows the operation of the device 3A. The device 3A supports the operation of the facility 2 by performing the processes of steps S11 to S21. Note that the processes of steps S11 to S15 are the same as those in the above embodiment, and therefore their explanation will be omitted. In addition, in this operation, a case will be described in which a learning process is performed on the estimation model 3040 that outputs control parameters in step S15.

[0103] In step S17, the operation result acquisition unit 311 acquires the results of operating the equipment 2 using the estimation model 3040. The operation result acquisition unit 311 may acquire the results of operating the equipment 2 by simulating the equipment 2 using the values ​​of the target parameters (in the present embodiment, as an example, control parameters to be applied to any of the devices 21) output from the estimation model 3040 in response to supplying the target feature quantities to the estimation model 3040.

[0104] In step S19, the evaluation unit 312 calculates an evaluation value by evaluating the results acquired by the operation result acquisition unit 311 (i.e., the operation results of the equipment 2). The evaluation unit 312 may calculate an evaluation value by comprehensively evaluating a plurality of estimated values ​​indicating the state of the equipment 2.

[0105] In step S21, the result output unit 313 outputs the estimation model 3040 trained by the learning processing device 304 on the condition that the calculated evaluation value exceeds a reference value.

[0106] <Modification> In the above embodiment, the devices 3, 3A have been described as including the selection device 302, the learning processing device 304, the estimation unit 306, and the storage units 300, 300A. However, any of these may be omitted. If the devices 3, 3A do not include the selection device 302, the selection device 302 may be externally connected to the devices 3, 3A. Similarly, if the devices 3, 3A do not include the learning processing device 304, the learning processing device 304 may be externally connected to the devices 3, 3A. If the devices 3, 3A do not include the estimation unit 306, the estimation model 3040 learned by the learning processing device 304 may be output to an external device of the devices 3, 3A and used for estimation by the external device. If the devices 3, 3A do not include the storage units 300, 300A, the operating data 3001, the equipment data 3003, and the assistance data 3002 may be stored in a storage device externally connected to the devices 3, 3A.

[0107] Furthermore, although the device 3A has been described as including the driving result acquisition unit 311, the evaluation unit 312, the result output unit 313, the storage unit 300A, the first learning processing unit 314, the similarity calculation unit 315, the similarity result output unit 316, the extraction unit 317, and the second learning processing unit 318, any of these may be omitted. If the device 3A does not include the driving result acquisition unit 311, the evaluation unit 312, and the result output unit 313, the estimation model 3040 that has undergone learning processing in the learning processing device 304 may be output to the estimation unit 306, as in the device 3 in the first embodiment. If the device 3A does not include the extraction unit 317, the second learning processing unit 318 may perform learning processing on the selection device 302 using learning data including the driving data 3001.

[0108] In addition, in the above description, the operation result acquisition unit 311 is a simulator, and the results of operating the equipment 2 are acquired by simulation using the values ​​of the target parameters (here, control parameters) output from the estimation model 3040 in response to supplying the target feature quantities to the estimation model 3040. However, in addition to or instead of this, the results may be acquired by actually operating the equipment 2. In this case, the operation result acquisition unit 311 may supply the control parameters output from the estimation model 3040 to the target device 21, actually operate the equipment 2, and acquire the operation results. As an example, the operation result acquisition unit 311 may acquire results when the equipment 2 is actually operated a reference number of times or for a reference period. This makes it possible to limit the actual operation of the equipment 2 using an estimation model 3040 whose evaluation has not been established. Here, the reference number of times and the reference period may be set arbitrarily, but are preferably set to a small number of times and a short period in consideration of safety. When the operation result acquisition unit 311 acquires operation results by each of a simulation and an actual operation of the facility 2, the operation result acquisition unit 311 may supply the value of the control parameter output from the estimation model 3040 within a first range as is to the device 21 in the simulation, while in the actual operation, the operation result acquisition unit 311 may convert the value of the control parameter within the first range into a value within a second range and supply the converted value to the device 21. The first range may be a numerical range that the value of the control parameter can take, and the second range may be a range narrower than the first range.

[0109] In addition, the result output unit 313 has been described as outputting the estimation model 3040 that outputs control parameters on the condition that the evaluation value calculated by the evaluation unit 312 exceeds a reference value. However, in addition to or instead of the estimation model 3040, other operation results by the device 3A may be output. For example, the result output unit 313 may output at least one of the target parameters identified by the identification unit 301 or the target feature quantities selected by the selection device 302 on the condition that the evaluation value calculated by the evaluation unit 312 exceeds a reference value. In this case, when an estimation model 3040 with a low evaluation of the driving result is generated, the target parameters and target feature quantities that are the basis of the estimation model 3040 can be prevented from being output. The result output unit 313 may output the target parameters and target feature quantities to a user via a display device (not shown) such as a display. The result output unit 313 may output the target parameters and target feature quantities after a learning process is performed on the estimation model 3040 and an evaluation value for the operation result using the estimation model 3040 is obtained.

[0110] Furthermore, in the assistance data 3002, although it has been described that the target parameters and the target features are associated with the content of the dialogue between the identification unit 301 and the user, only one of these may be associated. In this case, a data set in which the dialogue content and the target parameters are associated may be stored each time the identification unit 301 completes identification, or a data set in which the dialogue content and the target features are associated may be stored each time the selection device 302 completes selection. The similarity result output unit 316 may output the target parameters or the target features stored in association with the content of the past dialogue that has the highest similarity to the content of the current dialogue in the assistance data 3002 in the storage unit 300A.

[0111] Various embodiments of the present invention may be described with reference to flowcharts and block diagrams, where the blocks may represent (1) stages of a process in which operations are performed or (2) sections of an apparatus responsible for performing the operations. Particular stages and sections may be implemented by dedicated circuitry, programmable circuitry provided with computer-readable instructions stored on a computer-readable medium, and / or a processor provided with computer-readable instructions stored on a computer-readable medium. Dedicated circuitry may include digital and / or analog hardware circuitry, and may include integrated circuits (ICs) and / or discrete circuits. Programmable circuitry may include reconfigurable hardware circuitry, including logical AND, OR, XOR, NAND, NOR, and other logical operations, flip-flops, registers, memory elements such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), and the like.

[0112] A computer-readable medium may include any tangible device capable of storing instructions that are executed by an appropriate device, such that the computer-readable medium having instructions stored thereon comprises an article of manufacture containing instructions that can be executed to create means for performing the operations specified in the flowcharts or block diagrams. Examples of computer-readable media may include electronic, magnetic, optical, electromagnetic, and semiconductor storage media. More specific examples of computer-readable media may include floppy disks, diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), electrically erasable programmable read-only memory (EEPROM), static random access memory (SRAM), compact disc read-only memory (CD-ROM), digital versatile disc (DVD), Blu-ray disc, memory stick, integrated circuit card, and the like.

[0113] The computer readable instructions may include either assembler instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state-setting data, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk®, JAVA®, C++, etc., and conventional procedural programming languages ​​such as the “C” programming language or similar programming languages.

[0114] The computer-readable instructions may be provided to a processor or programmable circuit of a programmable data processing device, such as a computer, locally or over a wide area network (WAN) such as a local area network (LAN) or the Internet, and the computer-readable instructions may be executed to create means for performing the operations specified in the flowcharts or block diagrams. Here, the computer may be a personal computer (PC), a tablet computer, a smartphone, a workstation, a server computer, a general-purpose computer, a special-purpose computer, or the like, or may be a computer system in which multiple computers are connected. Such a computer system in which multiple computers are connected is also called a distributed computing system, and is a broad definition of computer. In a distributed computing system, the multiple computers collectively execute a program by each executing a portion of the program and passing data between the computers as needed during program execution.

[0115] Examples of processors include a computer processor, a central processing unit (CPU), a processing unit, a microprocessor, a digital signal processor, a controller, a microcontroller, etc. A computer may have one or more processors. In a multiprocessor system with multiple processors, each processor executes a portion of a program and passes data between processors as needed during program execution, allowing the multiple processors to collectively execute a program. For example, in multitasking, each of the multiple processors may execute a portion of each task in small chunks by switching tasks at time slice intervals. In this case, which portion of a program each processor executes changes dynamically. Which portion of a program each of the multiple processors executes may also be statically determined by multiprocessor-aware programming.

[0116] 11 illustrates an example of a computer 1200 in which aspects of the present invention may be embodied, in whole or in part. Programs installed on the computer 1200 may cause the computer 1200 to function as or perform operations associated with an apparatus or one or more sections of the apparatus according to embodiments of the present invention, and / or to perform a process or steps of a process according to embodiments of the present invention. Such programs may be executed by the CPU 1212 to cause the computer 1200 to perform specific operations associated with some or all of the blocks of the flowcharts and block diagrams described herein.

[0117] A computer 1200 according to this embodiment includes a CPU 1212, a RAM 1214, a graphics controller 1216, and a display device 1218, which are interconnected by a host controller 1210. The computer 1200 also includes input / output units such as a communication interface 1222, a storage device 1224 such as a hard disk drive, a DVD-ROM drive 1226, and an IC card drive, which are connected to the host controller 1210 via an input / output controller 1220. The computer also includes legacy input / output units such as a ROM 1230 and a keyboard 1242, which are connected to the input / output controller 1220 via an input / output chip 1240.

[0118] The CPU 1212 operates according to programs stored in the ROM 1230 and the RAM 1214, thereby controlling each unit. The graphics controller 1216 acquires image data generated by the CPU 1212 into a frame buffer or the like provided in the RAM 1214 or into the graphics controller 1216 itself, and causes the image data to be displayed on the display device 1218.

[0119] The communication interface 1222 communicates with other electronic devices via a network. The storage device 1224 stores programs and data used by the CPU 1212 in the computer 1200. The DVD-ROM drive 1226 reads programs or data from a DVD-ROM 1227 and provides the programs or data to the storage device 1224 via the RAM 1214. The IC card drive reads programs and data from an IC card and / or writes programs and data to an IC card.

[0120] The ROM 1230 stores therein a boot program or the like that is executed by the computer 1200 upon activation, and / or programs that depend on the hardware of the computer 1200. The input / output chip 1240 may also connect various input / output units to the input / output controller 1220 via a parallel port, a serial port, a keyboard port, a mouse port, etc.

[0121] The programs are provided by a computer-readable medium such as a DVD-ROM 1227 or an IC card. The programs are read from the computer-readable medium, installed in the storage device 1224, RAM 1214, or ROM 1230, which are also examples of computer-readable media, and executed by the CPU 1212. Information processing described in these programs is read by the computer 1200, and causes cooperation between the programs and the various types of hardware resources described above. An apparatus or a method may be configured by implementing information manipulation or processing in accordance with the use of the computer 1200.

[0122] For example, when communication is performed between the computer 1200 and an external device, the CPU 1212 may execute a communication program loaded into the RAM 1214 and instruct the communication interface 1222 to perform communication processing based on the processing described in the communication program. Under the control of the CPU 1212, the communication interface 1222 reads transmission data stored in a transmission buffer processing area provided in the RAM 1214, the storage device 1224, the DVD-ROM 1227, or a recording medium such as an IC card, and transmits the read transmission data to the network, or writes reception data received from the network to a reception buffer processing area or the like provided on the recording medium.

[0123] Furthermore, the CPU 1212 may cause all or a necessary portion of a file or database stored in an external recording medium such as the storage device 1224, the DVD-ROM drive 1226 (DVD-ROM 1227), an IC card, etc. to be read into the RAM 1214, and perform various types of processing on the data in the RAM 1214. The CPU 1212 then writes back the processed data to the external recording medium.

[0124] Various types of information, such as various types of programs, data, tables, and databases, may be stored on the recording medium and may undergo information processing. CPU 1212 may perform various types of processing on data read from RAM 1214, including various types of operations, information processing, conditional judgment, conditional branching, unconditional branching, information search / replacement, etc., as described throughout this disclosure and specified by the instruction sequences of the programs, and write the results back to RAM 1214. CPU 1212 may also search for information in a file, database, etc. on the recording medium. For example, if multiple entries each having an attribute value of a first attribute associated with an attribute value of a second attribute are stored on the recording medium, CPU 1212 may search for an entry that matches a condition specified by the attribute value of the first attribute from among the multiple entries, read the attribute value of the second attribute stored in the entry, and thereby obtain the attribute value of the second attribute associated with the first attribute that satisfies a predetermined condition.

[0125] The above-described programs or software modules may be stored in a computer-readable medium on or near the computer 1200. A recording medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet can also be used as a computer-readable medium, thereby providing the programs to the computer 1200 via the network.

[0126] Although the present invention has been described above using embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments. It will be apparent to those skilled in the art that various modifications and improvements can be made to the above embodiments. It is clear from the claims that such modifications and improvements can also be included within the technical scope of the present invention.

[0127] It should be noted that the execution order of each process, such as operations, procedures, steps, and stages, in the devices, systems, programs, and methods shown in the claims, specifications, and drawings is not specifically stated as "before," "prior to," etc., and that the processes can be performed in any order unless the output of a previous process is used in a subsequent process. Even if the operational flow in the claims, specifications, and drawings is described using "first," "next," etc. for convenience, this does not mean that the processes must be performed in this order. [Explanation of symbols]

[0128] 1 System 2 Equipment 3 equipment 20 sensors 21 Equipment 300 Storage section 301 Specific section 302 Selection device 303 Selection Control Unit 304 Learning Processing Device 305 Learning control unit 306 Estimation Department 311 Driving result acquisition unit 312 Evaluation Department 313 Result output section 314 First learning processing unit 315 Similarity calculation unit 316 Similarity result output unit 317 Extraction part 318 Second Learning Processing Unit 1200 Computer 1210 host controller 1212 CPU 1214 RAM 1216 Graphics Controller 1218 Display Devices 1220 Input / Output Controller 1222 communication interface 1224 Storage device 1226 DVD-ROM drive 1227 DVD-ROM 1230 ROM 1240 Input / Output Chip 1242 keyboard 3001 Operational Data 3002 Support Data 3003 Facility Data 3011 Dialogue Acquisition Unit 3012 Generation part 3013 Dialogue output unit 3040 Estimation Model

Claims

1. an identification unit that identifies a parameter to be estimated in accordance with the content of a dialogue with a user; a selection control unit that causes a selection device that selects a feature to be used for estimating a parameter value in accordance with the parameter to select a target feature that is a feature corresponding to the target parameter; a learning control unit that causes a learning processing device to perform a learning process of an estimation model that outputs an estimated value of the target parameter in response to input of a value of the target feature, using learning data including a value indicated by the target parameter and a value indicated by the target feature when the target parameter indicates the value; An apparatus comprising:

2. The device according to claim 1 , wherein the learning control unit extracts a portion of a data set from a plurality of data sets each including a value indicated by the target parameter and a value indicated by the target feature, and uses the extracted portion as the learning data.

3. The identification unit a dialogue acquisition unit that acquires an input sentence in a natural language from a user; a generation unit that generates an output sentence in a natural language in accordance with the input sentence; a dialogue output unit that outputs the output sentence; 10. The apparatus of claim 1, comprising:

4. The device according to claim 1 , wherein the specifying unit asks the user a question about an item for which information has not yet been acquired through dialogue, among a plurality of pre-registered items necessary for specifying the target parameter.

5. the selection device has a user interface for specifying the target parameter in response to a user operation; The device comprises: The device described in claim 1, further comprising a first learning processing unit that performs learning processing of the identification unit using learning data including text displayed on a user interface of the selection device, guiding users in entering content, and text entered into the user interface.

6. 6. The device according to claim 5, wherein the first learning processing unit further performs reinforcement learning on the identification unit using learning data including a dialogue content by the identification unit and a reward value according to an evaluation input by a user for the target feature or the estimation model.

7. a storage unit that stores, each time the identification by the identification unit or the selection by the selection device is completed, a dialogue content between the identification unit and the user and at least one of the target parameter identified in accordance with the dialogue content or the target feature amount selected in accordance with the target parameter, in association with each other; a similarity calculation unit that calculates a similarity between each of the plurality of past conversation contents stored in the storage unit and the current conversation content; a first output unit that outputs at least one of the object parameters or the object feature amounts stored in the storage unit in association with a past dialogue content that has the highest similarity to the current dialogue content; The apparatus of claim 1 further comprising:

8. The apparatus according to claim 1 , wherein the target parameter is a control parameter to be applied to any one of a plurality of devices included in a facility.

9. an operation result acquisition unit that acquires a result of operating the equipment using the value of the target parameter output from the estimation model in response to supplying the target feature quantity to the estimation model, by at least one of simulating the equipment or actually operating the equipment; an evaluation unit that calculates an evaluation value by evaluating the results acquired by the driving result acquisition unit; a second output unit that outputs at least one of the target parameters, the target feature amounts, or an estimation model trained by a learning processing device, on condition that the evaluation value exceeds a reference value; The apparatus of claim 8 further comprising:

10. The device according to claim 9 , wherein the operation result acquisition unit acquires a result when the equipment is actually operated a reference number of times or for a reference period.

11. The apparatus according to claim 1 , wherein the target parameter is an index value of the quality of a product produced by a facility.

12. an extraction unit that extracts information indicated using graphics in the image from the image data related to the facility; a second learning processing unit that performs learning processing of the selection device using learning data including the information extracted by the extraction unit; 12. The apparatus of claim 8 or 11, further comprising:

13. A computer-implemented method comprising: an identification step of identifying parameters to be estimated based on the content of the dialogue with the user; a selection control step of causing a selection device that selects a feature to be used for estimating a parameter value in accordance with the parameter to select a target feature that is a feature corresponding to the target parameter; a learning control step of causing a learning processing device to perform a learning process of an estimation model that outputs an estimated value of the target parameter in response to input of a value of the target feature, using learning data including a value indicated by the target parameter and a value indicated by the target feature when the target parameter indicates the value; A method for providing the above.

14. When executed by a computer, the computer an identification unit that identifies a parameter to be estimated in accordance with the content of a dialogue with a user; a selection control unit that causes a selection device that selects a feature to be used for estimating a parameter value in accordance with the parameter to select a target feature that is a feature corresponding to the target parameter; a learning control unit that causes a learning processing device to perform a learning process of an estimation model that outputs an estimated value of the target parameter in response to input of a value of the target feature, using learning data including a value indicated by the target parameter and a value indicated by the target feature when the target parameter indicates that value. A program that functions as a

Images