JP7535845B2 - Information processing method and information processing system - Google Patents

Description

The present disclosure relates to an information processing method and information processing system that control the training of a machine learning model using food and drink data.

A recipe provision system is disclosed that determines attributes for identifying a user's cooking preferences, collects information on the user's cooking preferences identified by the attributes, and provides recipes that match the user's preferences based on the determined user attributes and the collected information on the user's cooking preferences (for example, Patent Literature 1).

JP 2017-215687 A

Patent Document 1 discloses that preferences are learned from the user's impressions of suggested recipes, but the user's impressions change depending on the weather, physical condition, mood, schedule, etc. at the time, and can be ambiguous. For this reason, data that should not be learned may be used in the machine learning, resulting in inefficient learning.

Therefore, the present disclosure provides an information processing method etc. that can efficiently train a machine learning model using food and drink data.

The information processing method according to the present disclosure includes a process of acquiring food and beverage data, presenting the data via a presentation device, generating options for the food and beverage data, presenting the options via the presentation device, generating reasons for selection according to the selected options, presenting the reasons via the presentation device, and controlling the training of a machine learning model using the food and beverage data according to the input results of the reasons for selection.

These comprehensive or specific aspects may be realized as a system, a method, an integrated circuit, a computer program, or a computer-readable recording medium such as a CD-ROM, or may be realized as any combination of a system, a method, an integrated circuit, a computer program, and a recording medium.

According to an information processing method according to one aspect of the present disclosure, it is possible to efficiently train a machine learning model using food and drink data.

FIG. 2 is a block diagram showing an example of a server and a presentation device according to an embodiment. 10 is a flowchart showing an example of an operation of the server according to the embodiment when transmitting food and drink data, etc. 13 is a flowchart showing another example of the operation of the server according to the embodiment when transmitting food and drink data and the like. 10 is a flowchart illustrating an example of an operation of the presentation device according to the embodiment. 1A and 1B are diagrams illustrating an example of an image presented by a presentation device according to an embodiment. 10 is a flowchart showing an example of an operation performed when receiving an input result of the server according to the embodiment. 13 is a flowchart showing an example of an operation of a server according to a modified example of an embodiment when transmitting food and drink data and the like. 13 is a flowchart showing another example of the operation of the server according to the modified example of the embodiment when transmitting food and drink data and the like. 13 is a flowchart showing an example of an operation of a presentation device according to a modified example of an embodiment. 13A and 13B are diagrams illustrating an example of an image presented by a presentation device according to a modified example of the embodiment. FIG. 11 is a diagram showing another example of an image presented by a presentation device according to a modified example of the embodiment. 13 is a flowchart showing an example of an operation of a server according to a modified example of an embodiment when receiving an input result.

An information processing method according to one aspect of the present disclosure includes a process of acquiring food and beverage data, presenting the data via a presentation device, generating options for the food and beverage data, presenting the options via the presentation device, generating reasons for selection according to the selected options, presenting the reasons via the presentation device, and controlling the training of a machine learning model using the food and beverage data according to the input results of the reasons for selection.

For example, a user's opinions on food and drink data may vary and be ambiguous depending on the weather, physical condition, mood, schedule, etc. at the time, and such ambiguous opinions may result in data with poor learning efficiency for a machine learning model using food and drink data. Therefore, selection reasons (specifically, multiple candidate selection reasons) are generated according to options for food and drink data, and the generated selection reasons are presented to the presentation device to have the user input a selection reason from among the selection reasons presented. The input result of the selection reason is not an opinion that the user thinks of each time, but is one that the user adopts from the specifically generated and presented selection reasons, so that the training of the machine learning model can be controlled according to the input result of such a specific selection reason, thereby efficiently training the machine learning model using food and drink data.

In addition, in controlling the training, the parameters of the training may be determined based on the input result of the reason for selection, and the training may be controlled using the parameters.

This allows the optimal parameters for training the machine learning model to be determined based on the input results of the selection reason, and the machine learning model can be efficiently trained using those parameters.

The parameter may also be a weighting parameter for the food and drink data.

This allows the weighting parameters for food and drink data to be determined based on the input of the reasons for selection, making it possible to efficiently train a machine learning model using the weighting parameters.

The parameter may also be a reward for reinforcement learning.

This makes it possible to efficiently train a machine learning model using the reinforcement learning reward by determining the reinforcement learning reward based on the input result of the reason for selection.

In addition, when determining the parameters, the training parameters may also be determined based on the input results of the options.

This allows the training to be controlled using the input of options related to the input of the reason for selection, thereby improving the effectiveness of the training.

The input result of the reason for selection or the input result of the option may also include the direction, strength, or speed of the input for the reason for selection or the option.

In this way, machine learning models can be efficiently trained according to the direction, strength, or speed of input regarding the reason for selection or option.

In addition, when determining the parameters, if the reason for selection is input late, the parameters that will reduce the effectiveness of the training may be determined.

If the user is slow to input the reason for selection, it is likely that the user has hesitated before entering the reason for selection, and the reason entered at this time may not be applicable to the user. For this reason, by determining training parameters that reduce the effectiveness of training when the reason for selection is entered slowly, it is possible to train the machine learning model efficiently.

In addition, when determining the parameters, the parameters that increase the effectiveness of the training may be determined if the reason for selection is input quickly.

When the reason for selection is entered quickly, it is likely that the user has entered the reason for selection without hesitation, and the reason entered at this time may be the reason for selection that applies to the user. For this reason, by determining training parameters that will increase the effectiveness of training when the reason for selection is entered quickly, it is possible to efficiently train the machine learning model.

The presentation of the reason for selection may also be controlled according to the speed or strength of input of the option.

According to this, for example, if the user types the options quickly or with strong input, it is assumed that the user input the options based on the appearance of the images of the food and drink data presented, so it is possible to present a reason for selection that is related to appearance. Also, for example, if the user types the options slowly or with weak input, it is assumed that the user entered the option after hesitating, and the reason for selection in this case is likely to be one that does not apply to the user, so it is possible to present a different, more specific reason for selection.

In addition, the control of the training may determine whether or not to provide the training depending on the input result of the reason for selection.

This allows training of the machine learning model to be performed efficiently by performing training when the input result of the selection reason is data with good learning efficiency, and not performing training when the input result of the selection reason is data with poor learning efficiency. In addition, because data with poor learning efficiency is not collected, the memory capacity for storing data for training the machine learning model can be reduced.

The machine learning model may also output a predicted value of the selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model, obtain the predicted value from the machine learning model, and not present the selection reason for the food and drink data for which the predicted value is equal to or greater than a threshold value.

According to this, for food and drink data where the predicted value of the option selection is equal to or greater than the threshold, training on which option to select is advanced, and it is possible to not present the reason for selection for such food and drink data. In other words, it is possible to prevent the user from having to go to the trouble of inputting the reason for selection.

In addition, if the input result of the option is positive, other food and drink data associated with the food and drink data may be presented via the presentation device.

According to this, for example, when options for food and drink data are presented to suggest a menu, if the input result of the option for the presented food and drink data is positive, other food and drink data associated with the food and drink data can be presented that can be combined with the food and drink data to create a better menu (for example, a side dish that goes well with the main dish).

In addition, if the input result of the option is negative, the next food and drink data may be determined based on the input result of the reason for selection and presented via the presentation device.

With this, if the input result of the option for the presented food and drink data is negative, the next food and drink data different from the food and drink data can be presented.

Furthermore, attributes of the food and drink data may be presented via the presentation device, and when presenting the reason for selection, the reason for selection corresponding to the attributes may be generated and presented via the presentation device.

There are various reasons for selecting options for food and drink data, depending on the attributes of the food and drink data, such as ingredients, cooking methods, or seasonings. Therefore, by generating and presenting selection reasons corresponding to the attributes and controlling the training of the machine learning model according to the input results of the selection reasons for each attribute, it is possible to efficiently train the machine learning model.

The machine learning model may also output a predicted value of the adoption or rejection of the selection reason for each attribute of the food and drink data obtained by inputting the food and drink data into the machine learning model, obtain the predicted value from the machine learning model, and not present the selection reason for attributes whose predicted value is equal to or greater than a threshold value.

According to this, for attributes where the predicted value of the selection reason acceptance is equal to or greater than a threshold, training has progressed in which selection reason is selected for that attribute, and it is possible to not present selection reasons for such attributes. In other words, it is possible to prevent the user from having to go to the trouble of inputting a selection reason.

In addition, if the input result of the option is negative, food and drink data with different attributes may be determined based on the input result of the reason for selection, and presented via the presentation device.

With this, if the input result of the option for the presented food and drink data is negative, food and drink data with different attributes (e.g. different ingredients, cooking methods, seasonings, etc.) can be presented.

The reason for selection may also be generated based on the characteristics of the food and drink data.

The food and drink data may include, for example, images of food and drink or the names of food and drink dishes, and the reasons for selection may be generated from the characteristics of this information.

The options and the reasons for selection may be input by flick input.

This allows options and reasons for selection to be easily entered using flick input.

Also, the voice captured during the flick input may be captured as an input result of the reason for selection.

This allows you to input the reason for your selection by voice while using flick input.

An information processing system according to one aspect of the present disclosure includes an acquisition unit that acquires food and beverage data, an option generation unit that generates options for the food and beverage data, a selection reason generation unit that generates a selection reason corresponding to the selected option, a presentation control unit that presents the food and beverage data, the options, and the selection reason via a presentation device, and a control unit that controls training of a machine learning model using the food and beverage data in response to an input result of the selection reason.

This makes it possible to provide an information processing system that can efficiently train machine learning models using food and drink data.

The following describes the embodiment in detail with reference to the drawings.

The embodiments described below are all comprehensive or specific examples. The numerical values, shapes, materials, components, component placement and connection forms, steps, and order of steps shown in the following embodiments are merely examples and are not intended to limit the present disclosure.

(Embodiment)
Hereinafter, an embodiment will be described with reference to FIG. 1 to FIG.

Figure 1 is a block diagram showing an example of a server 100 and a presentation device 200 according to an embodiment.

The server 100 is a device that communicates with the presentation device 200 and provides content to the presentation device 200, and is an example of an information processing system. For example, the server 100 has a function of suggesting menus to a user of the presentation device 200. The server 100 also has a function of acquiring responses to suggestions made to the user from the presentation device 200 and learning the user's preferences for food and drink.

The server 100 includes a content suggestion unit 110, a training unit 120, a storage unit 130, and a communication unit 140. The server 100 is a computer including a processor, a memory, a communication interface, and the like. The memory is a ROM (Read Only Memory) and a RAM (Random Access Memory), and can store a program executed by the processor. The storage unit 130 is an example of a memory, but may be a memory separate from the memory that stores the program. The content suggestion unit 110 and the training unit 120 are realized by a processor or the like that executes a program stored in the memory. The communication unit 140 is realized by a communication interface that includes a connector or an antenna to which a communication circuit, a communication line, and the like are connected. The components that make up the server 100 may be distributed and arranged in multiple servers.

The memory unit 130 stores food and drink data and a machine learning model that uses the food and drink data.

The content suggestion unit 110 is an example of an acquisition unit that acquires food and drink data from the storage unit 130, an option generation unit that generates options for the food and drink data, a selection reason generation unit that generates a selection reason corresponding to the selected option, and a presentation control unit that presents the food and drink data, the options, and the selection reason via the presentation device 200. Details of the operation of the content suggestion unit 110 will be described later.

The communication unit 140 is a communication interface that communicates with the presentation device 200. The communication unit 140 transmits the food and drink data acquired by the content proposal unit 110, and the options and selection reasons generated by the content proposal unit 110 to the presentation device 200. Specifically, the communication unit 140 transmits UI (user interface) data for presenting the food and drink data, the options, and the selection reasons on the presentation device 200 to the presentation device 200. The communication unit 140 also receives input results from the presentation device 200, which are the results of inputting the options and selection reasons on the presentation device 200.

The training unit 120 is an example of a control unit that controls the training of a machine learning model using food and drink data in response to the input result of the reason for selection. The training unit 120 controls the training of the machine learning model stored in the memory unit 130. In this way, the user's preferences for food and drink are learned. The operation of the training unit 120 will be described in detail later.

The presentation device 200 is a device that communicates with the server 100 and presents the content provided by the server 100 to the user, and is, for example, a smartphone or a tablet terminal.

The presentation device 200 includes a presentation control unit 210, a presentation unit 220, an input unit 230, and a communication unit 240. The presentation device 200 is a computer including a processor, a memory, a user interface, a communication interface, and the like. The memory is a ROM, a RAM, and the like, and can store a program executed by the processor. The presentation control unit 210 is realized by a processor or the like that executes a program stored in the memory. The communication unit 240 is realized by a communication interface consisting of a communication circuit and an antenna, and the like. The presentation unit 220 and the input unit 230 are realized by a user interface consisting of a display and a touch panel, and the like. For example, the presentation unit 220 and the input unit 230 are shown as separate components, but may be a single component having a presentation function and an input function.

The communication unit 240 is a communication interface that communicates with the server 100. The communication unit 240 receives food and drink data, options, and reasons for selection from the server 100. Specifically, the communication unit 240 receives UI data for presenting the food and drink data, options, and reasons for selection on the presentation device 200 from the server 100. The communication unit 240 also transmits input results, which are the results of inputting the options and reasons for selection on the presentation device 200, to the server 100.

The presentation control unit 210 presents food and drink data, options, and reasons for selection to the presentation unit 220 based on the UI data received by the communication unit 240. The presentation control unit 210 also controls the presentation on the presentation unit 220 in response to input to the input unit 230. The operation of the presentation control unit 210 will be described in detail later.

The presentation unit 220 is, for example, a touch panel display that presents food and drink data, options, and reasons for selection. The content presented by the presentation unit 220 changes according to input to the input unit 230.

The input unit 230 is an input device that accepts user operations, such as a touch panel display. The user, for example, determines whether to select an option presented on the presentation unit 220 and whether to accept the reason for selection by flick input. The selected option becomes the input result of the option, and the adopted reason for selection becomes the input result of the reason for selection.

Next, the operation of the server 100 will be described using Figures 2A, 2B, and 5, and the operation of the presentation device 200 will be described using Figure 3. Note that the information processing method according to the embodiment is, for example, a method executed by a processor included in the server 100, and each step shown in Figures 2A, 2B, and 5 may be processing performed by the information processing method according to the embodiment.

Figure 2A is a flowchart showing an example of the operation of server 100 in accordance with an embodiment when transmitting food and drink data, etc.

The content suggestion unit 110 (acquisition unit) acquires food and drink data from the storage unit 130 (step S11). Various food and drink data is stored in the storage unit 130, and the food and drink data is acquired, for example, randomly or according to predetermined rules. For example, the food and drink data includes data and metadata such as the name of the food and drink dish, images, ingredients, cooking method, seasoning, genre, cooking difficulty, and cooking time.

Next, the content suggestion unit 110 (option generation unit) generates options for the acquired food and drink data (step S12). For example, the options are options as to whether or not the food and drink data should be included in today's menu, and are specifically presented as "Yes" and "No" on the presentation device 200. For example, if the "Yes" option is selected, the food and drink data is included in today's menu, and if the "No" option is selected, the food and drink data is not included in today's menu.

Next, the content suggestion unit 110 (selection reason generation unit) generates a selection reason corresponding to the selected option (step S13). For example, the selection reasons are potential reasons for including the acquired food and drink data in today's menu, and potential reasons for not including it in today's menu, and are specifically presented on the presentation device 200 as "like," "dislike," and "also had it yesterday."

Then, the communication unit 140 transmits the food and drink data, the options, and the reason for selection to the presentation device 200 (step S14). Specifically, the communication unit 140 transmits UI data including the food and drink data, the options, and the reason for selection to the presentation device 200. As a result, the options and the reason for selection are presented on the presentation device 200, and the content suggestion unit 110 (presentation control unit) presents the food and drink data, the options, and the reason for selection via the presentation device 200.

In addition, once the training has progressed to a certain extent, the machine learning model, whose training is controlled by the training unit 120, may output a predicted value of the selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model. In other words, the machine learning model may output a predicted value of whether or not the user will adopt the input food and drink data in a menu. For example, the generation of the selection reason may be controlled using such a machine learning model. This will be explained using FIG. 2B.

Figure 2B is a flowchart showing another example of the operation of the server 100 according to the embodiment when transmitting food and drink data, etc.

The content suggestion unit 110 acquires food and drink data from the storage unit 130 (step S21) and generates options for the acquired food and drink data (step S22).

Next, the content suggestion unit 110 obtains a predicted value of the selection of the option for the food and drink data output from the machine learning model (step S23). For example, if the training results in a prediction that the user will include the food and drink data input to the machine learning model in a menu, that is, the user is highly likely to select the "Yes" option on the presentation device 200, the predicted value for the selection of the "Yes" option will be high. Also, if the user will not include the food and drink data input to the machine learning model in a menu, that is, the user is highly likely to select the "No" option on the presentation device 200, the predicted value for the selection of the "No" option will be high. On the other hand, if it is unclear whether the user will select the "Yes" option or the "No" option, the predicted value will be low.

Next, the content suggestion unit 110 determines whether the obtained predicted value is less than the threshold value (step S24).

If the acquired predicted value is less than the threshold (Yes in step S24), the content suggestion unit 110 generates a selection reason (step S25). If the predicted value is less than the threshold, it is not clear whether the user will select the "Yes" option or the "No" option, and it is necessary to present the selection reason to the presentation device 200 and have the user input the selection reason, and to further train the machine learning model for food and drink data whose predicted value is less than the threshold using the input result of the selection reason.

Then, the communication unit 140 transmits the food and drink data, the options, and the reason for selection to the presentation device 200 (step S26).

If the acquired predicted value is equal to or greater than the threshold, the content suggestion unit 110 does not generate a selection reason, and the communication unit 140 transmits the food and drink data and the options to the presentation device 200 (step S27). That is, the content suggestion unit 110 does not present a selection reason for food and drink data whose predicted value is equal to or greater than the threshold. When the predicted value is equal to or greater than the threshold, it is clear whether the user will select the "Yes" option or the "No" option, and there is no need to present a selection reason to the presentation device 200 to have the user input the selection reason, and there is no need to further train the machine learning model for food and drink data whose predicted value is equal to or greater than the threshold using the input result. In this way, for food and drink data whose predicted value for the selection of an option is equal to or greater than the threshold, training on which option will be selected has progressed, and it is possible not to present a selection reason for such food and drink data. That is, it is possible to avoid giving the user the trouble of having to input a selection reason.

Figure 3 is a flowchart showing an example of the operation of the presentation device 200 according to the embodiment.

The presentation control unit 210 determines whether the communication unit 240 has received food and drink data, options, and reasons for selection from the server 100 (step S31). Specifically, the presentation control unit 210 determines whether the communication unit 240 has received UI data including food and drink data, options, and reasons for selection from the server 100.

If the presentation control unit 210 determines that the communication unit 240 has not received the food and drink data, the options, and the reason for selection from the server 100 (No in step S31), it repeats the processing in step S31 until the communication unit 240 receives the food and drink data, the options, and the reason for selection.

When the presentation control unit 210 determines that the communication unit 240 has received the food and drink data, the options, and the reason for selection from the server 100 (Yes in step S31), the presentation control unit 210 presents the food and drink data and the options to the presentation unit 220 (step S32). Specifically, the presentation control unit 210 generates images of the food and drink data and images of the options using UI data including the food and drink data, the options, and the reason for selection, and presents these images to the presentation unit 220.

Next, the presentation control unit 210 determines whether or not an input has been made for the option presented on the presentation unit 220 (step S33). Specifically, the presentation control unit 210 determines whether or not a user operation has been made for the option presented on the presentation unit 220 via the input unit 230.

If there is no input for the options presented on the presentation unit 220 (No in step S33), the presentation control unit 210 repeats the process in step S33 until there is an input for the options presented on the presentation unit 220.

When an input is received for an option presented to the presentation unit 220 (Yes in step S33), the presentation control unit 210 presents a reason for selection according to the selected option (step S34). Specifically, the presentation control unit 210 generates an image of the reason for selection according to the selected option using UI data including the food and drink data, the option, and the reason for selection, and presents the image on the presentation unit 220.

Here, we will use Figure 4 to explain the images of food and drink data and options presented on the presentation unit 220, as well as the image of the reason for selection that is presented when an input is made for a presented option.

Figure 4 is a diagram showing an example of an image presented by the presentation device 200 according to the embodiment. The left side of Figure 4 shows an image of food and drink data and an image of options presented by the presentation device 200. The right side of Figure 4 shows an image of food and drink data presented by the presentation device 200, an image of the options, and an image of the reason for selection.

As shown on the left side of FIG. 4, in the process of step S32, an image of, for example, a chicken ball is presented as an image of food and drink data, and images of, for example, a "Yes" option and a "No" option are presented on the presentation device 200 as option images. In addition, an arrow image is presented around the image of the "Yes" option, encouraging the user to flick to the side of the screen where "Yes" is presented, and an arrow image is presented around the image of the "No" option, encouraging the user to flick to the side of the screen where "No" is presented. When a user sees these images, if the user wants to include chicken balls in the menu, the user flicks to the side where the image of the "Yes" option is presented, and if the user does not want to include chicken balls in the menu, the user flicks to the side where the image of the "No" option is presented. In this way, the user can input to the options.

For example, suppose that the user performs a flick input on the side where the image of the "No" option is presented as an input for a choice. In this case, as shown on the right side of FIG. 4, an image of the reason for selecting "No" is presented on the side of the screen where "No" is presented as an image of the reason for selecting "No" according to the choice, that is, an image of the reason for not including chicken balls in the menu. Here, an image of "I hate it" and an image of "I ate it yesterday" are presented as images of the reason for selecting "No" according to the choice. If the reason for not including chicken balls in the menu is that the user hates chicken balls, the user performs a flick input on the image of "I hate it," and if the reason for not including chicken balls in the menu is that the user ate chicken balls yesterday, the user performs a flick input on the image of "I ate it yesterday." In this way, the user can input a specific reason for selection.

Returning to the explanation in FIG. 3, the presentation control unit 210 determines whether or not there has been an input for the reason for selection presented on the presentation unit 220 (step S35). Specifically, the presentation control unit 210 determines whether or not there has been a user operation via the input unit 230 for the reason for selection presented on the presentation unit 220.

If the reason for selection presented to the presentation unit 220 is not input (No in step S35), the presentation control unit 210 repeats the process in step S35 until the reason for selection presented to the presentation unit 220 is input.

When the reason for selection presented to the presentation unit 220 is input (Yes in step S35), the presentation control unit 210 transmits the input result of the option and the reason for selection to the server 100 via the communication unit 240 (step S36). For example, if the user does not include chicken balls in the menu because he or she hates them, the input result of the option indicates that chicken balls are not included, and the input result of the reason for selection indicates that the reason for not including chicken balls is that he or she hates them.

Note that if the reason for selection is not received in step S31, that is, if the food and drink data and options are sent from the server 100 but the reason for selection is not sent as in step S27 of FIG. 2B, the processing in steps S34 and S35 does not need to be performed, and the input result of the reason for selection does not need to be sent in step S36.

Figure 5 is a flowchart showing an example of the operation of the server 100 when receiving an input result according to an embodiment.

The training unit 120 determines whether the communication unit 140 has received the input results of the options and the reasons for selection from the presentation device 200 (step S41).

If the training unit 120 determines that the communication unit 140 has not received the input results of the options and the reason for selection from the presentation device 200 (No in step S41), it repeats the processing in step S41 until the communication unit 140 receives the input results of the options and the reason for selection.

When the communication unit 140 determines that the input results of the options and the reasons for selection have been received from the presentation device 200 (Yes in step S41), the training unit 120 controls the training of the machine learning model using the food and drink data according to the input results of the reasons for selection (step S42).

For example, the training unit 120 may determine training parameters according to the input result of the reason for selection, and train the machine learning model with the determined parameters. The training parameters are, for example, weighting parameters for food and drink data. For example, the training unit 120 determines the weighting of the input vector corresponding to the reason for selection according to the input result of the reason for selection. The training parameters may also be rewards for reinforcement learning. For example, when the selected reason for selection is "hate", the training unit 120 trains the model with parameters that are more effective in training so that the food and drink data "chicken ball" is less likely to be suggested as a menu candidate. Also, for example, when the selected reason for selection is "eaten yesterday", the training unit 120 trains the model with parameters that are more effective in training so that the food and drink data "chicken ball" is less likely to be suggested as a menu candidate, but are less effective in training.

For example, the training unit 120 may also determine the training parameters according to the input results of the options. The input of the options is related to the input of the reason for selection, and the training is controlled using the input of the options related to the input of the reason for selection, thereby improving the effectiveness of the training.

The input result of the reason for selection or the input result of the option may include the direction, strength, or speed of the input for the reason for selection or the option. For example, if the option and the reason for selection are input by flick input, the input result of the reason for selection or the input result of the option may include the direction, strength, or speed of the flick input when flick inputting the reason for selection or the option.

For example, if the reason for selection is input slowly, the training unit 120 determines training parameters that will reduce the effectiveness of the training. For example, if the reason for selection is input slowly, the training unit 120 reduces the weighting parameters for the food and drink data. This is because if the reason for selection is input slowly, it is considered that the user is unsure of the reason for selection before inputting it, and the reason for selection input at this time may not be applicable to the user.

For example, if the reason for selection is input quickly, the training unit 120 determines training parameters that will increase the effectiveness of the training. For example, if the reason for selection is input quickly, the training unit 120 increases the weighting parameters for the food and drink data. This is because, if the reason for selection is input quickly, it is considered that the user has input the reason for selection without hesitation, and the reason for selection input at this time may be a reason for selection that applies to the user.

Also, for example, the training unit 120 may determine whether or not to perform training depending on the input result of the selection reason. In other words, the training unit 120 does not need to train the machine learning model depending on the input result of the selection reason. For example, training can be performed when the input result of the selection reason is data with good learning efficiency, and training can be not performed when the input result of the selection reason is data with poor learning efficiency, thereby training the machine learning model efficiently. In addition, since data with poor learning efficiency is not collected, the storage capacity for storing data for training the machine learning model can be reduced. For example, when the selected reason for selection is "hate", the training unit 120 trains the model so that the food and drink data "chicken ball" is less likely to be suggested as a menu candidate, but when the selected reason for selection is "eat it yesterday", training on the food and drink data "chicken ball" is not performed.

Next, the training unit 120 determines whether the input result of the option is positive (step S43). For example, if the user performs a flick input to the side where the image of the "Yes" option is presented, the input result of the option is positive, and if the user performs a flick input to the side where the image of the "No" option is presented, the input result of the option is negative.

If the input result of the option is positive (Yes in step S43), the content suggestion unit 110 determines other food and drink data associated with the food and drink data (step S44). An example of a case in which the input result of the option is positive is when the presented food and drink data is adopted in a menu. The other food and drink data associated with the food and drink data adopted in a menu is food and drink data that goes well with the food and drink data adopted in the menu. Specifically, if the adopted food and drink data is a Japanese main dish, the other food and drink data is a Japanese side dish or Japanese soup that goes well with the Japanese main dish.

On the other hand, if the input result of the option is negative (No in step S43), the content suggestion unit 110 determines the next food and drink data based on the input result of the reason for selection (step S45). An example of a case where the input result of the option is negative is when the presented food and drink data is not adopted in the menu. The next food and drink data is food and drink data that does not fit the selection reason for not adopting the presented food and drink data. For example, if the presented food and drink data is chicken balls, a main dish, and the selection reason for not adopting them is that it was also eaten yesterday, the next food and drink data is food and drink data that is different from the food and drink data included in yesterday's menu and is food and drink data that could be the main dish.

Then, the content suggestion unit 110 generates options for the determined food and drink data (step S46) and generates reasons for selection according to the options (step S47), and the communication unit 140 transmits the determined food and drink data, the options, and the reasons for selection to the presentation device 200 (step S48). As a result, the content suggestion unit 110 presents the determined food and drink data via the presentation device 200, and the processes shown in Figures 2A, 2B, 3, and 5 are performed.

Note that if several food and drink data are adopted and all components of the menu are determined, the process may end without performing steps S43 and subsequent steps.

In addition, if the input result of the reason for selection is not received in step S41, that is, if the food and drink data and the options are sent from the server 100 but the reason for selection is not sent as in step S27 of FIG. 2B, and the input result of the reason for selection is not sent in step S36 of FIG. 3, the processing in step S42 may not be performed.

Note that it is not necessary to repeat the process of presenting the food and drink data, options, and reasons for selection once and then controlling the training of the machine learning model according to the input results for each piece of food and drink data. That is, multiple pieces of food and drink data and the options and reasons for selection for each of the multiple food and drink data may be transmitted collectively to the presentation device 200. In this case, the presentation device 200 may repeat the processing from step S32 to step S35 of FIG. 3 for each of the multiple received food and drink data, and transmit the input results of the options and reasons for selection for each of the multiple food and drink data to the server 100 collectively. Then, the training of the machine learning model may be controlled according to the input results of the options and reasons for selection for each of the multiple food and drink data.

As described above, for example, a user's impressions of food and drink data may vary and be ambiguous depending on the weather, physical condition, mood, schedule, etc. at the time, and such ambiguous impressions may result in data with poor learning efficiency for a machine learning model using food and drink data. Therefore, selection reasons (specifically, multiple candidate selection reasons) are generated according to options for food and drink data, and the generated selection reasons are presented to the presentation device to have the user input a selection reason from among the selection reasons presented. Since the input result of the selection reason is not an impression that the user thinks of each time, but is one that the user adopts from the specifically generated and presented selection reasons, the training of the machine learning model can be controlled according to the input result of such a specific selection reason, thereby efficiently training the machine learning model using food and drink data.

(Modification)
Next, a modification of the embodiment will be described with reference to FIGS. 6A to 10. FIG.

The configurations of the server 100 and presentation device 200 in the modified embodiment are the same as those in the embodiment, but the functions are different from those in the embodiment.

Figure 6A is a flowchart showing an example of the operation of server 100 in a modified embodiment when transmitting food and drink data, etc.

The content suggestion unit 110 acquires food and drink data from the storage unit 130 (step S51) and generates options for the acquired food and drink data (step S52).

Next, the content suggestion unit 110 generates attributes of the food and drink data (step S53). The attributes of the food and drink data are, for example, the ingredients, cooking method, or seasoning of the food and drink data, and are specifically presented on the presentation device 200 as "ingredients," "cooking method," and "flavor."

Next, the content suggestion unit 110 generates a selection reason corresponding to the attribute (step S54). For example, the selection reason corresponding to the attribute is a candidate reason for including the acquired food and drink data in today's menu when the attribute is focused on, and a candidate reason for not including the acquired food and drink data in today's menu when the attribute is focused on, and is specifically presented on the presentation device 200 as "like," "dislike," and "also had it yesterday."

Then, the communication unit 140 transmits the food and drink data, the options, the attributes, and the reason for selection to the presentation device 200 (step S55). Specifically, the communication unit 140 transmits UI data including the food and drink data, the options, the attributes, and the reason for selection to the presentation device 200. As a result, the options, attributes, and the reason for selection are presented on the presentation device 200, and the content suggestion unit 110 presents the options, attributes, and the reason for selection via the presentation device 200.

In addition, once the training has progressed to a certain extent, the machine learning model, whose training is controlled by the training unit 120, may output a predicted value of whether or not to adopt a selection reason for each attribute of the food and drink data obtained by inputting the food and drink data into the machine learning model. In other words, the machine learning model may output a predicted value for each attribute as to whether or not the user will adopt each selection reason corresponding to the attributes of the input food and drink data. For example, the generation of selection reasons corresponding to attributes may be controlled using such a machine learning model. This will be described with reference to FIG. 6B.

Figure 6B is a flowchart showing another example of the operation of server 100 in a modified embodiment when transmitting food and drink data, etc.

The content suggestion unit 110 acquires food and drink data from the storage unit 130 (step S61), generates options for the acquired food and drink data (step S62), and generates attributes of the food and drink data (step S63).

Next, the content suggestion unit 110 obtains a predicted value of whether or not to adopt the selection reason for each attribute of the food and drink data output from the machine learning model (step S64). For example, if the training result predicts that the user is highly likely to adopt a specific selection reason for a certain attribute of the food and drink data input to the machine learning model, the predicted value of the certain attribute will be high. On the other hand, if it is unclear which selection reason the user will adopt for a certain attribute of the food and drink data input to the machine learning model, the predicted value will be low.

Next, the content suggestion unit 110 determines whether the obtained predicted value is less than the threshold value (step S65).

When the acquired predicted value is less than the threshold (Yes in step S65), the content suggestion unit 110 generates a selection reason for the attribute whose predicted value is less than the threshold (step S66). For attributes whose predicted value is less than the threshold, it is not clear which selection reason the user will adopt, so it is necessary to present the selection reason for the attribute to the presentation device 200 and have the user input the selection reason, and to further train the machine learning model for attributes whose measured value is less than the threshold using the input result.

Then, the communication unit 140 transmits the food and drink data, the options, the attributes, and the selection reasons corresponding to the attributes to the presentation device 200 (step S67).

When the acquired predicted value is equal to or greater than the threshold, the content suggestion unit 110 does not generate a selection reason for the attribute whose predicted value is equal to or greater than the threshold, and the communication unit 140 transmits the food and drink data, the options, and the attribute to the presentation device 200 (step S68). That is, the content suggestion unit 110 does not present a selection reason for the attribute whose predicted value is equal to or greater than the threshold. For attributes whose predicted value is equal to or greater than the threshold, it is clear which selection reason the user will adopt, and there is no need to present a selection reason for the attribute to the presentation device 200 to have the user input the selection reason, and there is no need to further train the machine learning model for attributes whose predicted value is equal to or greater than the threshold using the input result. In this way, for attributes whose predicted value for the adoption or rejection of the selection reason is equal to or greater than the threshold, training on which selection reason will be selected for that attribute has progressed, and it is possible not to present a selection reason for such attributes. In other words, it is possible to avoid giving the user the trouble of having to input a selection reason.

Figure 7 is a flowchart showing an example of the operation of the presentation device 200 according to a modified embodiment.

The presentation control unit 210 determines whether the communication unit 240 has received food and drink data, options, attributes, and reasons for selection from the server 100 (step S71). Specifically, the presentation control unit 210 determines whether the communication unit 240 has received UI data including food and drink data, options, attributes, and reasons for selection from the server 100.

If the presentation control unit 210 determines that the communication unit 240 has not received the food and drink data, options, attributes, and reasons for selection from the server 100 (No in step S71), it repeats the processing in step S71 until the communication unit 240 receives the food and drink data, options, and reasons for selection.

When the presentation control unit 210 determines that the communication unit 240 has received the food and drink data, options, attributes, and reasons for selection from the server 100 (Yes in step S71), the presentation control unit 210 presents the food and drink data and options to the presentation unit 220 (step S72). Specifically, the presentation control unit 210 generates images of the food and drink data and images of the options using UI data including the food and drink data, options, attributes, and reasons for selection, and presents these images to the presentation unit 220.

Next, the presentation control unit 210 determines whether or not there has been an input for the option presented on the presentation unit 220 (step S73). Specifically, the presentation control unit 210 determines whether or not there has been a user operation via the input unit 230 for the option presented on the presentation unit 220.

If there is no input for the options presented on the presentation unit 220 (No in step S73), the presentation control unit 210 repeats the process in step S73 until there is an input for the options presented on the presentation unit 220.

When an input is received for the option presented on the presentation unit 220 (Yes in step S73), the presentation control unit 210 presents the attributes of the food and drink data (step S74). Specifically, the presentation control unit 210 generates an image of the attributes of the food and drink data using UI data including the food and drink data, the option, the attribute, and the reason for selection, and presents the image on the presentation unit 220.

Next, the presentation control unit 210 determines whether or not there has been an input for the attribute presented on the presentation unit 220 (step S75). Specifically, the presentation control unit 210 determines whether or not there has been a user operation via the input unit 230 for the attribute presented on the presentation unit 220.

If there is no input for the attribute presented on the presentation unit 220 (No in step S75), the presentation control unit 210 repeats the process in step S75 until there is input for the attribute presented on the presentation unit 220.

When an input is received for an attribute presented to the presentation unit 220 (Yes in step S75), the presentation control unit 210 presents a selection reason corresponding to the attribute (step S76). Specifically, the presentation control unit 210 generates an image of the selection reason corresponding to the attribute using UI data including food and drink data, options, attributes, and selection reasons, and presents the image to the presentation unit 220.

Here, the images of food and drink data and options presented on the presentation unit 220, the images of attributes presented when an input is made for the presented options, and the images of reasons for selection presented when an input is made for the presented attributes will be described with reference to FIG. 8.

Figure 8 is a diagram showing an example of an image presented by the presentation device 200 according to a modified example of the embodiment. The left side of Figure 8 shows an image of food and drink data and an image of options presented by the presentation device 200. The center of Figure 8 shows an image of food and drink data, an image of options, and an image of attributes presented by the presentation device 200. The right side of Figure 8 shows an image of food and drink data presented by the presentation device 200, an image of options, an image of selected attributes, and an image of the reason for selection.

As shown on the left side of FIG. 8, in the process of step S72, an image of, for example, a chicken ball is presented as an image of food and drink data, and images of, for example, a "Yes" option and a "No" option are presented on the presentation device 200 as option images. In addition, an arrow image is presented around the image of the "Yes" option, encouraging the user to flick to the side of the screen where "Yes" is presented, and an arrow image is presented around the image of the "No" option, encouraging the user to flick to the side of the screen where "No" is presented. When a user sees these images, if the user wants to include chicken balls in the menu, the user flicks to the side where the image of the "Yes" option is presented, and if the user does not want to include chicken balls in the menu, the user flicks to the side where the image of the "No" option is presented. In this way, the user can input to the options.

For example, suppose the user performs a flick input on the side where the image of the "No" option is presented as an input for a choice. In this case, as shown in the center of FIG. 8, an image of a food or drink attribute is presented as an attribute image, which is the type of reason for selecting "No." Here, images of "ingredients," "cooking method," and "flavor" are presented as attribute images. If the reason for not using chicken balls in the menu is the ingredients (e.g., chicken), the user performs a flick input toward the image of "ingredients." If the reason for not using chicken balls in the menu is the cooking method (e.g., deep-frying), the user performs a flick input toward the image of "cooking method." If the reason for not using chicken balls in the menu is the seasoning (e.g., salty), the user performs a flick input toward the image of "flavor." In this way, the user can perform input for attributes.

For example, suppose that the user performs a flick input on the side where the image of the attribute "ingredients" is presented as an input for an attribute. In this case, as shown on the right side of FIG. 8, on the side of the screen where "No" is presented, an image of the selection reason for the ingredients among the selection reasons for selecting "No" as an image of the selection reason corresponding to the attribute, that is, an image of the reason for not using chicken balls in the menu when focusing on the ingredients, is presented. Here, an image of "I hate it" and an image of "I ate it yesterday" are presented as images of the selection reason corresponding to the attribute of "ingredients". If the reason for not using chicken balls in the menu is that the user hates chicken, the user performs a flick input toward the image of "I hate it", and if the reason for not using chicken balls in the menu is that the user ate chicken yesterday, the user performs a flick input toward the image of "I ate it yesterday". In this way, the user can input a selection reason corresponding to the attribute.

In the above description, the reason for selection is input by selecting one of specific reasons for selection, such as "I hate it" and "I ate it yesterday," but this is not limiting. For example, the degree of the reason for selection may be input. This will be described with reference to FIG. 9.

Figure 9 is a diagram showing another example of an image presented by the presentation device 200 according to a modified embodiment of the present invention. Figure 9 shows an image of the selection reason when an input is made for an attribute.

In FIG. 9, images of attributes are shown as "I like it, but," "For some reason," and "I don't like it," but there are no boundaries between these images, and they form a single image. "I like it, but" is presented at the top of the area within the integrated image, "For some reason" is presented in the middle, and "I don't like it" is presented at the bottom. When the top of the area is selected, the degree of preference for the ingredient tends to be toward "like," and when the bottom of the area is selected, the degree of preference for the ingredient tends to be toward "dislike." In this way, the degree of the reason for selection may be input.

In this case, parameters may be determined such that the more ambiguous the reason for selection, the less effective the training. For example, when the selected reason for selection is "just because," the training unit 120 trains the model with parameters that make it less likely to suggest the food and drink data "chicken balls" as a menu candidate, but have a low training effect. Alternatively, when the reason for selection is ambiguous, training may not be performed.

Returning to the explanation in FIG. 7, the presentation control unit 210 determines whether or not there has been an input for the reason for selection presented on the presentation unit 220 (step S77). Specifically, the presentation control unit 210 determines whether or not there has been a user operation via the input unit 230 for the reason for selection presented on the presentation unit 220.

If the reason for selection presented to the presentation unit 220 is not input (No in step S77), the presentation control unit 210 repeats the process in step S77 until the reason for selection presented to the presentation unit 220 is input.

When the reason for selection presented to the presentation unit 220 is input (Yes in step S77), the presentation control unit 210 transmits the input result of the option and the reason for selection to the server 100 via the communication unit 240 (step S78). For example, if the user does not include chicken balls in the menu because he or she dislikes the ingredient (chicken), the input result of the option indicates that chicken balls are not included, and the input result of the reason for selection indicates that the reason for not including chicken balls is that he or she dislikes chicken.

In addition, if the reason for selection is not received in step S71, that is, if the food and drink data, options, and attributes are sent from the server 100 but the reason for selection is not sent as in step S68 of FIG. 6B, the processing in steps S76 and S77 does not need to be performed, and the input result of the reason for selection does not need to be sent in step S78.

Figure 10 is a flowchart showing an example of the operation of the server 100 when receiving an input result according to a modified embodiment.

The training unit 120 determines whether the communication unit 140 has received the input results of the options and the reasons for selection from the presentation device 200 (step S81).

If the training unit 120 determines that the communication unit 140 has not received the input results of the options and the reason for selection from the presentation device 200 (No in step S81), it repeats the processing in step S81 until the communication unit 140 receives the input results of the options and the reason for selection.

When the communication unit 140 determines that the input results of the options and the reasons for selection have been received from the presentation device 200 (Yes in step S81), the training unit 120 controls the training of the machine learning model using the food and drink data according to the input results for each attribute of the reasons for selection (step S82).

For example, the training unit 120 may determine training parameters according to the input results for each attribute of the selection reason, and train the machine learning model with the determined parameters. Also, for example, the training unit 120 may determine whether or not to perform training according to the input results for the selection reason for each attribute. For example, when the selected attribute is an ingredient and the selected reason for selection is "hate," the training unit 120 trains the model with parameters that are more effective at training so that the food and drink data "chicken balls" and the ingredient "chicken" are less likely to be suggested as menu candidates. That is, the machine learning model is trained using the food and drink data and attributes as training data.

Next, the training unit 120 determines whether the input result of the option is positive (step S83).

If the input result of the option is positive (Yes in step S83), the content suggestion unit 110 determines other food and drink data associated with the food and drink data (step S84).

On the other hand, if the input result of the option is negative (No in step S83), the content suggestion unit 110 determines food and drink data with different attributes based on the input result of the reason for selection (step S85). Food and drink data with different attributes is food and drink data that does not fit the selection reason for the attribute that did not adopt the presented food and drink data. For example, if the presented food and drink data is chicken balls as a main dish and the selection reason for not adopting it is because the user ate chicken yesterday, the next food and drink data will be food and drink data that is different from food and drink data that uses chicken (i.e., food and drink data that does not use chicken), and will be food and drink data that can become a main dish. In this way, if the input result of the option for the presented food and drink data is negative, food and drink data with different attributes (for example, different ingredients, cooking methods, or seasonings) can be presented.

Then, the content suggestion unit 110 generates options for the determined food and drink data (step S86), generates attributes (step S87), and generates selection reasons corresponding to the attributes (step S88), and the communication unit 140 transmits the determined food and drink data, options, attributes, and selection reasons to the presentation device 200 (step S89). As a result, the content suggestion unit 110 presents the determined food and drink data via the presentation device 200, and the processes shown in Figures 6A, 6B, 7, and 10 are performed.

Note that if several food and drink data are adopted and all of the menu components have been determined, the process may end without performing steps S83 and subsequent steps.

In addition, if the input result of the reason for selection is not received in step S81, that is, if the food and drink data, options, and attributes are sent from the server 100 as in step S68 of FIG. 6B but the reason for selection is not sent, and if the input result of the reason for selection is not sent in step S78 of FIG. 7, the processing in step S82 may not be performed.

Note that it is not necessary to repeat the process of presenting the food and drink data, options, attributes, and selection reasons once and then controlling the training of the machine learning model according to the input results for each piece of food and drink data. That is, multiple pieces of food and drink data, as well as the options, attributes, and selection reasons for each of the multiple food and drink data, may be transmitted collectively to the presentation device 200. In this case, the presentation device 200 may repeat the processing in steps S72 to S77 of FIG. 7 for each of the multiple received food and drink data, and transmit the input results of the options and selection reasons for each of the multiple food and drink data to the server 100 collectively. Then, the training of the machine learning model may be controlled according to the input results of the options and selection reasons for each of the multiple food and drink data.

As described above, there are various reasons for selecting options for food and drink data, depending on the attributes of the food and drink data, such as ingredients, cooking methods, or seasonings. Therefore, by generating and presenting selection reasons corresponding to the attributes and controlling the training of the machine learning model according to the input results of the selection reasons for each attribute, the machine learning model can be trained efficiently.

(Other embodiments)
The information processing method and the server 100 (information processing system) according to one or more aspects of the present disclosure have been described above based on the embodiments, but the present disclosure is not limited to these embodiments. As long as it does not deviate from the spirit of the present disclosure, various modifications conceived by a person skilled in the art to each embodiment and forms constructed by combining components in different embodiments may also be included within the scope of one or more aspects of the present disclosure.

For example, the voice acquired during the flick input may be acquired as the input result of the reason for selection. For example, the voice input function of the presentation device 200 may be turned on from the timing when the flick input is started to the timing when it is completed, and the voice uttered by the user during that time may be picked up and the voice recognition result may be acquired as the input result of the reason for selection.

Also, for example, as described in FIG. 9, when the degree of the reason for selection is input, for example, if "Yes" is selected as an option but the reason for selection is "somehow" or "dislike", etc., "Yes" was not actively selected, so specific options may be generated to elicit the user's dissatisfaction, and the user may be prompted to input the dissatisfaction, and learning may be performed using the input result. Also, when the degree of the reason for selection is input, for example, if "No" is selected as an option and the reason for selection is "dislike", etc., strong learning may be performed that "No" is the answer.

In addition, for example, input of options and reasons for selection is not limited to flick input, but may be tap input, swipe input, etc.

In addition, the presentation of the reason for selection may be controlled, for example, according to the speed or strength of input of the options. For example, if the options are input quickly or with strong input, it is highly likely that the user is selecting an option based on the appearance of the image of the presented food and drink data. For this reason, something related to appearance may be presented as the reason for selection. In addition, for example, if the options are input slowly or with weak input, it is possible that the user is having some difficulty in selecting the food and drink data presented. For this reason, another specific reason for selection may be presented.

Also, for example, a selection reason may be generated based on the characteristics of the food and drink data, including an image showing the food and drink or the name of the food and drink dish. For example, color characteristics of the food and drink data may be extracted from the image of the food and drink data, and a selection reason may be generated based on the color characteristics. Specifically, if the image does not contain much green, a selection reason such as "not many vegetables" may be generated. Also, for example, attribute characteristics of the food and drink data may be extracted from the name of the dish of the food and drink data, and a selection reason may be generated based on the attribute characteristics. Specifically, if the name of the dish is "stir-fried pork and bean sprouts," selection reasons such as "pork," "bean sprouts," and "stir-fried" may be generated. Also, if a tag such as "time saving" is associated with the food and drink data, a selection reason such as "time saving" may be generated.

In addition, for example, an example in which positive or negative options for a menu are presented has been described, but positive or negative options may also be presented for recipes, eating out, ready-made meals, etc.

In addition, for example, an example has been described in which training is controlled according to the input result of the selection reason or the option, but the information processing system 100 may select data for training according to the input result of the selection reason or the option.

For example, in the example of FIG. 4, when the selected reason for selection is "I hate it," the training unit 120 classifies the food and drink data "chicken balls" as data to be trained with parameters that make it more difficult to suggest it as a menu candidate, but with a higher training effect. In addition, when the selected reason for selection is "I ate it yesterday," the training unit 120 classifies the food and drink data "chicken balls" as data to be trained with parameters that make it more difficult to suggest it as a menu candidate, but with a lower training effect. Alternatively, in the latter case, the food and drink data may be determined not to be data to be trained.

Furthermore, for example, in the example of FIG. 8, when the selected reason for selection is "hate," the training unit 120 classifies the combination of food and drink data "chicken balls" and ingredient "chicken" as data to be trained with parameters that are more effective in making it more difficult to suggest the combination as a menu candidate.

For example, in the example of FIG. 9, when the selected reason for selection is "just because," the training unit 120 trains the model with parameters that make it less likely to suggest the food and drink data "chicken balls" and the ingredient "chicken" as menu candidates, but have a low training effect. Alternatively, it may be determined that the combination of food and drink data and ingredients is not data that should be trained.

The present disclosure can be realized as a program for causing a processor to execute the steps included in the information processing method. Furthermore, the present disclosure can be realized as a non-transitory computer-readable recording medium, such as a CD-ROM, on which the program is recorded.

For example, when the present disclosure is realized by a program (software), each step is performed by executing the program using hardware resources such as a computer's CPU, memory, and input/output circuits. In other words, each step is performed by the CPU acquiring data from memory or input/output circuits, etc., performing calculations, and outputting the results of the calculations to memory or input/output circuits, etc.

In the above embodiment, each component included in the server 100 may be configured with dedicated hardware, or may be realized by executing a software program suitable for each component. Each component may be realized by a program execution unit such as a CPU or processor reading and executing a software program recorded on a recording medium such as a hard disk or semiconductor memory.

Some or all of the functions of the server 100 according to the above embodiment are typically realized as an LSI, which is an integrated circuit. These may be individually integrated into a single chip, or may be integrated into a single chip that includes some or all of the functions. Furthermore, the integrated circuit is not limited to an LSI, and may be realized by a dedicated circuit or a general-purpose processor. An FPGA (Field Programmable Gate Array) that can be programmed after LSI manufacture, or a reconfigurable processor that can reconfigure the connections and settings of circuit cells inside the LSI, may also be used.

Furthermore, as long as the modifications do not deviate from the spirit of this disclosure, they also include various modifications that are within the scope of what a person skilled in the art would think of in each embodiment of this disclosure.

The information processing method disclosed herein can be applied, for example, to a device that learns a user's food and drink preferences.

REFERENCE SIGNS LIST 100 Server 110 Content suggestion unit 120 Training unit 130 Storage unit 140 Communication unit 200 Presentation device 210 Presentation control unit 220 Presentation unit 230 Input unit 240 Communication unit

Claims (26)

1. A computer-implemented information processing method, comprising:
Acquire food and drink data and present it via a presentation device;
generating options for the food and drink data and presenting the options via the presentation device;
generating a reason for selection in accordance with the selected option and presenting the reason for selection via the presentation device;
Controlling training of a machine learning model using the food and drink data according to the input result of the reason for selection;
the machine learning model is a model that outputs a predicted value of a selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model,
generating the selection reason based on the characteristics of the food and drink data ;
In the control of the training,
determining parameters of the training in response to the input result of the reason for selection;
Controlling the training using the parameters
Or,
The decision as to whether or not to conduct the training is made depending on the input result of the reason for selection.
Information processing methods. The information processing method according to claim 1 , wherein the parameter is a weighting parameter for the food and drink data. The information processing method according to claim 1 , wherein the parameter is a reward in reinforcement learning. In determining the parameters,
The information processing method according to claim 1 , further comprising the step of determining a parameter of the training in accordance with an input result of the option. The information processing method according to claim 1 , wherein the input result of the reason for selection or the input result of the option includes a direction, a strength or a speed of an input for the reason for selection or the option. In determining the parameters,
The information processing method according to claim 5 , further comprising: determining the parameters such that the effect of the training is reduced when the reason for selection is input slowly. In determining the parameters,
The information processing method according to claim 5 or 6, further comprising: determining the parameters that will increase the effect of the training if the reason for selection is input quickly. The information processing method according to claim 5 , further comprising controlling the presentation of the reason for selection in accordance with a speed or strength of input of the option. obtaining the predicted value from the machine learning model;
The information processing method according to claim 1 , wherein the selection reason is not presented for the food and drink data whose predicted value is equal to or greater than a threshold value . The information processing method according to claim 1 , further comprising: presenting, via the presentation device, other food and drink data associated with the food and drink data when the input result of the option is positive. The information processing method according to claim 1 , further comprising: determining next food and drink data based on the input result of the reason for selection and presenting the next food and drink data via the presentation device when the input result of the option is negative. Presenting attributes of the food and drink data via the presentation device;
The information processing method according to claim 1 , wherein the presentation of the reason for selection comprises generating the reason for selection corresponding to the attribute and presenting the reason for selection via the presentation device. The machine learning model outputs a predicted value of the adoption or rejection of the selection reason for each attribute of the food and drink data obtained by inputting the food and drink data into the machine learning model;
obtaining the predicted value from the machine learning model;
The information processing method according to claim 12 , wherein the selection reason is not presented for an attribute whose predicted value is equal to or greater than a threshold value. The information processing method according to claim 12 or 13 , further comprising: determining food and drink data having different attributes based on the input result of the reason for selection when the input result of the option is negative, and presenting the food and drink data via the presentation device. The information processing method according to claim 1 , wherein the options and the reasons for selection are input by flick input. The information processing method according to claim 1 , wherein the options and the reasons for selection are input by swiping . The information processing method according to claim 1 , further comprising obtaining a voice acquired during the input as an input result of the reason for selection. An acquisition unit that acquires food and drink data;
an option generation unit that generates options for the food and drink data;
a selection reason generating unit that generates a selection reason corresponding to the selected option;
a presentation control unit that presents the food and drink data, the options, and the selection reasons via a presentation device;
A control unit that controls training of a machine learning model using the food and drink data in response to the input result of the reason for selection,
the machine learning model is a model that outputs a predicted value of a selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model,
the selection reason generating unit generates the selection reason based on a feature of the food and drink data ;
The control unit is
determining parameters of the training in response to the input result of the reason for selection;
Controlling the training using the parameters
Or,
The decision as to whether or not to conduct the training is made depending on the input result of the reason for selection.
Information processing system. 1. A computer-implemented information processing method, comprising:
Acquire food and drink data and present it via a presentation device;
generating options for the food and drink data and presenting the options via the presentation device;
generating a reason for selection in accordance with the selected option and presenting the reason for selection via the presentation device;
Controlling training of a machine learning model using the food and drink data according to the input result of the reason for selection;
the machine learning model is a model that outputs a predicted value of a selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model,
the input result of the reason for selection or the input result of the option includes a direction, a strength, or a speed of the input of the reason for selection or the option;
In the control of the training,
determining parameters of the training in response to the input result of the reason for selection;
Controlling the training using the parameters
Or,
The decision as to whether or not to conduct the training is made depending on the input result of the reason for selection.
Information processing methods. 1. A computer-implemented information processing method, comprising:
Acquire food and drink data and present it via a presentation device;
generating options for the food and drink data and presenting the options via the presentation device;
generating a reason for selection in accordance with the selected option and presenting the reason for selection via the presentation device;
Controlling training of a machine learning model using the food and drink data according to the input result of the reason for selection;
the machine learning model is a model that outputs a predicted value of a selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model,
obtaining the predicted value from the machine learning model;
The reason for selecting the food and drink data with the predicted value equal to or greater than a threshold is not presented,
In the control of the training,
determining parameters of the training in response to the input result of the reason for selection;
Controlling the training using the parameters
Or,
The decision as to whether or not to conduct the training is made depending on the input result of the reason for selection.
Information processing methods. 1. A computer-implemented information processing method, comprising:
Acquire food and drink data and present it via a presentation device;
generating options for the food and drink data and presenting the options via the presentation device;
generating a reason for selection in accordance with the selected option and presenting the reason for selection via the presentation device;
Controlling training of a machine learning model using the food and drink data according to the input result of the reason for selection;
the machine learning model is a model that outputs a predicted value of a selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model,
If the input result of the option is positive, presenting other food and drink data associated with the food and drink data via the presentation device ;
In the control of the training,
determining parameters of the training in response to the input result of the reason for selection;
Controlling the training using the parameters
Or,
The decision as to whether or not to conduct the training is made depending on the input result of the reason for selection.
Information processing methods. 1. A computer-implemented information processing method, comprising:
Acquire food and drink data and present it via a presentation device;
generating options for the food and drink data and presenting the options via the presentation device;
generating a reason for selection in accordance with the selected option and presenting the reason for selection via the presentation device;
Controlling training of a machine learning model using the food and drink data according to the input result of the reason for selection;
the machine learning model is a model that outputs a predicted value of a selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model,
If the input result of the option is negative, next food and drink data is determined based on the input result of the reason for selection, and presented via the presentation device ;
In the control of the training,
determining parameters of the training in response to the input result of the reason for selection;
Controlling the training using the parameters
Or,
The decision as to whether or not to conduct the training is made depending on the input result of the reason for selection.
Information processing methods. 1. A computer-implemented information processing method, comprising:
Acquire food and drink data and present it via a presentation device;
generating options for the food and drink data and presenting the options via the presentation device;
generating a reason for selection in accordance with the selected option and presenting the reason for selection via the presentation device;
Controlling training of a machine learning model using the food and drink data according to the input result of the reason for selection;
the machine learning model is a model that outputs a predicted value of a selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model,
Presenting attributes of the food and drink data via the presentation device;
In the presentation of the reason for selection, the reason for selection corresponding to the attribute is generated and presented via the presentation device ;
In the control of the training,
determining parameters of the training in response to the input result of the reason for selection;
Controlling the training using the parameters
Or,
The decision as to whether or not to conduct the training is made depending on the input result of the reason for selection.
Information processing methods. 1. A computer-implemented information processing method, comprising:
Acquire food and drink data and present it via a presentation device;
generating options for the food and drink data and presenting the options via the presentation device;
generating a reason for selection in accordance with the selected option and presenting the reason for selection via the presentation device;
Controlling training of a machine learning model using the food and drink data according to the input result of the reason for selection;
the machine learning model is a model that outputs a predicted value of a selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model,
The options and the reasons for selection are input by flick input;
In the control of the training,
determining parameters of the training in response to the input result of the reason for selection;
Controlling the training using the parameters
Or,
The decision as to whether or not to conduct the training is made depending on the input result of the reason for selection.
Information processing methods. 1. A computer-implemented information processing method, comprising:
Acquire food and drink data and present it via a presentation device;
generating options for the food and drink data and presenting the options via the presentation device;
generating a reason for selection in accordance with the selected option and presenting the reason for selection via the presentation device;
Controlling training of a machine learning model using the food and drink data according to the input result of the reason for selection;
the machine learning model is a model that outputs a predicted value of a selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model,
The input of the option and the reason for selection is a swipe input,
In the control of the training,
determining parameters of the training in response to the input result of the reason for selection;
Controlling the training using the parameters
Or,
The decision as to whether or not to conduct the training is made depending on the input result of the reason for selection.
Information processing methods. 1. A computer-implemented information processing method, comprising:
Acquire food and drink data and present it via a presentation device;
generating options for the food and drink data and presenting the options via the presentation device;
generating a reason for selection in accordance with the selected option and presenting the reason for selection via the presentation device;
Controlling training of a machine learning model using the food and drink data according to the input result of the reason for selection;
the machine learning model is a model that outputs a predicted value of a selection of an option for the food and drink data obtained by inputting the food and drink data into the machine learning model,
acquiring a voice acquired during the input as an input result of the reason for selection ;
In the control of the training,
determining parameters of the training in response to the input result of the reason for selection;
Controlling the training using the parameters
Or,
The decision as to whether or not to conduct the training is made depending on the input result of the reason for selection.
Information processing methods.

Images