JP7839615B2 - Information processing method, program, information processing device, and method for generating a learning model - Google Patents

Description

This invention relates to an information processing method, a program, an information processing device, and a method for generating a learning model.

Patent Document 1 discloses a recommendation device characterized by analyzing linguistic data used by a customer to specify a product (for example, sake), and outputting recommended products to the customer from among the products stored in a memory unit that stores information about the products.

Japanese Patent Publication No. 2019-139663

However, the invention described in Patent Document 1 may lead to the company proposing the same product to the customer again without referring to past proposal history.

One aspect of this is providing information processing methods, etc., that suggest appropriate beer beverages to customers.

The information processing method relating to one aspect involves acquiring customer attributes, acquiring answers to multiple questions from the customer, and outputting beer information related to beer beverages, including beer, low-malt beer, new genre beer, or beer-flavored beverages, which should be suggested when the acquired customer attributes and multiple answers are input to a learning model. The learning model then outputs beer information, which consists of characteristic values including at least two of the following: alcohol content, bitterness value, chromaticity, and ethyl acetate, for the beer beverages to be suggested when the customer attributes and multiple answers are input to a learning model. The history of beer beverages suggested based on the outputted characteristic values including at least two of the following: alcohol content, bitterness value, chromaticity, and ethyl acetate is stored in association with the customer, and the previous suggestion In a screen that includes a first display field for displaying beer information relating to a proposed beer beverage, a second display field for displaying beer information relating to a modified beer beverage different from the previously proposed beer beverage, and a third display field for displaying beer information relating to a candidate beer beverage different from the modified beer beverage, the screen is displayed when the beer beverage proposed to the customer based on the history matches the previously proposed beer beverage, and the computer performs a process to display the previously proposed beer beverage in the first display field and the modified beer beverage in the second display field.

One aspect of this is that it becomes possible to suggest appropriate beer beverages to customers.

This is an explanatory diagram showing an overview of the beer beverage proposal system. This is a block diagram showing an example server configuration. This is an explanatory diagram showing an example of a record layout for the customer database. This is an explanatory diagram showing an example of a record layout in the store database. This is an explanatory diagram showing an example of a record layout in the product database. This is an explanatory diagram showing an example of a record layout in the history database. This is an explanatory diagram showing an example of a record layout for an information collection database. This is a block diagram showing an example of a terminal configuration. This is a functional block diagram illustrating the processing operation of a beer beverage suggestion system. This is an explanatory diagram illustrating the beer information output model. This flowchart shows the procedure for changing to a different beer beverage than the one originally proposed. This flowchart shows the procedure for changing to a different beer beverage than the one originally proposed. This flowchart shows the processing procedure of the subroutine that suggests beer beverages. This is an explanatory diagram showing an example of a screen on a terminal that accepts customer attributes. This is an explanatory diagram showing an example of a screen on a device that accepts answers to multiple questions. This is an explanatory diagram showing an example of a screen displaying beer beverages before the change. This is an explanatory diagram showing an example of a screen displaying the modified beer beverage. This flowchart shows the processing procedure for receiving feedback results for beer beverages. This flowchart shows the processing steps for modifying beer beverages based on feedback results. This is an explanatory diagram showing an example of a screen for receiving feedback on a device. This is a block diagram showing an example of the server configuration of Embodiment 3. This is an explanatory diagram showing an example of a record layout for a category classification database. This flowchart shows the procedure for changing to a beer beverage belonging to the same category. This flowchart shows the procedure for updating the beer information output model.

The present invention will be described in detail below with reference to the drawings illustrating its embodiments.

(Embodiment 1)
Embodiment 1 relates to a method of changing a beer beverage to a different beer beverage based on a history of proposed beer beverages. Beer beverages include beer, low-malt beer, new genre beverages, and beer-flavored beverages. New genre beverages (third-category beer) are low-malt beer with a malt ratio of less than 50% to which spirits derived from barley are added, or beverages fermented using ingredients other than sugars, hops, water, and malt (grains, etc. as specified by government ordinance). In this embodiment, an example of a beer beverage is described, but the same applies to other types of beverages such as wine, sake, shochu, and whiskey.

Figure 1 is an explanatory diagram illustrating the outline of a proposed system for beer beverages. The system of this embodiment includes an information processing device 1 and an information processing terminal 2, and each device transmits and receives information via a network N such as the Internet.

Information processing device 1 is an information processing device that performs processing, storage, and transmission/reception of various types of information. Information processing device 1 is, for example, a server device, a personal computer, or a general-purpose tablet PC (personal computer). In this embodiment, information processing device 1 is assumed to be a server device, and for simplicity, it will be referred to as server 1 below.

Information processing terminal 2 is a terminal device that receives customer attributes and answers to multiple questions, and receives and displays suggested beer beverages. Information processing terminal 2 is an information processing device such as a smartphone, mobile phone, wearable device such as Apple Watch (registered trademark), tablet, or personal computer terminal. For simplicity, information processing terminal 2 will be read as terminal 2 below.

In this embodiment, Server 1 acquires customer attributes and answers to multiple questions from Terminal 2. Server 1 then inputs the acquired customer attributes and multiple answers into a learning model that outputs beer information, including beer, low-malt beer, or beer-flavored beverages, to suggest beer beverages. The learning model then outputs beer information to suggest beer beverages.

Server 1 suggests beer beverages based on the beer information output by the learning model and stores the history of suggested beer beverages, associating them with the customer. Server 1 determines whether the beer beverage suggested to the customer matches the previously suggested beer beverage. If Server 1 determines that the suggested beer beverage matches the previously suggested beer beverage, it changes it to a different beer beverage.

Figure 2 is a block diagram showing an example configuration of Server 1. Server 1 includes a control unit 11, a storage unit 12, a communication unit 13, an input unit 14, a display unit 15, a reading unit 16, and a large-capacity storage unit 17. Each component is connected via bus B.

The control unit 11 includes arithmetic processing units such as a CPU (Central Processing Unit), MPU (Micro-Processing Unit), and GPU (Graphics Processing Unit), and performs various information processing and control processing related to the server 1 by reading and executing the control program 1P stored in the storage unit 12. Although Figure 2 describes the control unit 11 as a single processor, it may also be a multi-processor system.

The memory unit 12 includes memory elements such as RAM (Random Access Memory) and ROM (Read Only Memory), and stores the control program 1P or data necessary for the control unit 11 to execute processing. The memory unit 12 also temporarily stores data necessary for the control unit 11 to execute arithmetic processing. The communication unit 13 is a communication module for performing communication-related processing, and transmits and receives information with terminals 2, etc., via the network N.

The input unit 14 is an input device such as a mouse, keyboard, touch panel, or buttons, and outputs the received operation information to the control unit 11. The display unit 15 is a liquid crystal display or an organic electroluminescent (EL) display, and displays various information according to the instructions of the control unit 11.

The reading unit 16 reads a portable storage medium 1a, including a CD (Compact Disc)-ROM or DVD (Digital Versatile Disc)-ROM. The control unit 11 may read the control program 1P from the portable storage medium 1a via the reading unit 16 and store it in the large-capacity storage unit 17. Alternatively, the control unit 11 may download the control program 1P from another computer via a network N or the like and store it in the large-capacity storage unit 17. Furthermore, the control unit 11 may also read the control program 1P from the semiconductor memory 1b.

The large-capacity storage unit 17 includes recording media such as an HDD (Hard Disk Drive) or SSD (Solid State Drive). The large-capacity storage unit 17 includes a customer database (DB) 171, a store database (DB) 172, a product database (DB) 173, a history database (DB) 174, an information collection database (DB) 175, and a beer information output model (Model) 176. The customer database (DB) 171 stores information about customers. The store database (DB) 172 stores information about stores. The product database (DB) 173 stores information about products such as beer beverages. The history database (DB) 174 stores a history of beer beverage recommendations made to customers. The information collection database (DB) 175 stores information about questions and feedback.

In this embodiment, the storage unit 12 and the large-capacity storage unit 17 may be configured as a single integrated storage device. Furthermore, the large-capacity storage unit 17 may be composed of multiple storage devices. Moreover, the large-capacity storage unit 17 may be an external storage device connected to the server 1.

In this embodiment, Server 1 is described as a single information processing device, but it may be configured as a distributed system using multiple devices, or as a virtual machine.

Figure 3 is an explanatory diagram showing an example of the record layout of Customer DB 171. Customer DB 171 includes columns for Customer ID, Nickname, Password, Email Address, Gender, Age, Frequently Consumed Alcohol, Drinking Frequency, Drinking Occasion, and Answers to Questions. The Customer ID column stores a uniquely identifiable customer ID to identify each customer. The Nickname column stores the customer's nickname. The Password column stores the login password. The Email Address column stores the customer's email address. The Gender column stores the customer's gender. The Age column stores the customer's age.

The "Frequently Drinked Alcohol" column stores the types of alcohol the customer frequently consumes. The "Drinking Frequency" column stores the frequency of beer consumption. Drinking frequency can be expressed as, for example, "Once a week" or "Almost daily." The "Drinking Scene" column stores information about the situations in which the customer feels like drinking alcohol. This column may include information such as "Prefers drinking at home" or "Prefers drinking out." The "Answers to Questions" column stores the customer's answers to multiple questions.

Figure 4 is an explanatory diagram showing an example of the record layout of Store DB 172. Store DB 172 includes columns for Store ID, Store Name, Product ID, and Start Date. The Store ID column stores the unique ID of each store to identify it. The Store Name column stores the name of the store. The Product ID column stores the ID of the product (the product being offered) handled by the store. The Start Date column stores the date the store began handling the product.

Figure 5 is an explanatory diagram showing an example of the record layout of the product database 173. The product database 173 includes columns for product ID, product type, brand name, and characteristics. In this embodiment, the product is described as a beer beverage. The product ID column stores the ID of each beer beverage, which is used to uniquely identify each beverage. The product type column stores the type of beer beverage. Types of beer beverages include beer, low-malt beer, new-genre beer, and beer-flavored beverages.

Beer is a beverage fermented using malt, hops, and water as raw materials. Alternatively, beer is a beverage fermented using malt, hops, water, and barley, or other items specified by government ordinance, provided that the total weight of the specified items does not exceed 50% of the weight of the malt. Happoshu (low-malt beer) is a sparkling alcoholic beverage (with an alcohol content of less than 20%) that uses malt as part of its raw materials, and has a lower proportion of malt than beer. The "New Genre" category is classified into two types: other brewed alcoholic beverages and liqueurs. Other brewed alcoholic beverages (sparkling) are beverages fermented using grains, sugars, and other items, with an alcohol content of less than 20% and an extract content of 2% or more. Liqueurs (sparkling) are sparkling beer with a malt ratio of less than 50% to which spirits are added, and have an extract content of 2% or more. Beer-flavored beverages are beverages that contain no alcohol or less than 1% alcohol.

The Brand column stores the brand name of the beer beverage. The Characteristics column includes Original Extract (OE), True Extract (ER), True Fermentation Level (RDF), Appearance Extract (AE), Appearance Fermentation Level (ADF), and Alcohol Content columns.

The Original Extract column stores the concentration of the original extract in the beer beverage. The True Extract column stores the concentration of the true extract (soluble evaporation residue) in the beer beverage. The True Fermentation Degree column stores the value indicating the true fermentation degree. The Appearance Extract column stores the concentration of the appearance extract in the beer beverage. The Appearance Fermentation Degree column stores the value indicating the appearance fermentation degree. The Alcohol Content column stores the alcohol content of the beer beverage.

Furthermore, regarding the characteristics of beer beverages, in addition to the original extract, true extract, true fermentation degree, appearance extract, appearance fermentation degree, and alcohol content exemplified above, other characteristic parameters may also be included. For example, the characteristics may further include characteristic parameters such as appearance final fermentation degree, bitterness value, extract content, residual fermentation degree, pH, color, ethyl acetate, n-propanol, isobutanol, isoamyl alcohol, isoamyl acetate, amyl alcohol, active amyl alcohol, acetaldehyde, diacetyl, amino acid sulfate ions, phosphate ions, citric acid, succinic acid, malic acid, acetic acid, lactic acid, isobutyric acid, caprylic acid, capric acid, total polyphenols, β-glucan, turbidity, foam retention time, carbon dioxide, linalool, α-acids, iso-α-acids, S-fraction analysis values, yeast used (top-fermenting yeast or bottom-fermenting yeast), information on whether or not a Japanese regional name is included in the product name, or information on whether or not malt, hops, wheat, sugars, spices, fruits, vegetables, rice, or miso are used as raw materials.

Figure 6 is an explanatory diagram showing an example of the record layout of History DB 174. History DB 174 includes columns for Customer ID, Proposed Beer Beverage ID, Feedback Result, and Proposal Date/Time. The Customer ID column stores the customer ID that identifies the customer. The Proposed Beer Beverage ID column stores the beer beverage ID of the beer beverage proposed to the customer. The Feedback Result column stores the feedback results for the proposed beer beverage. Feedback results include, for example, responses to multiple feedback items. The Proposal Date/Time column stores the date and time information when the beer beverage was proposed to the customer.

Figure 7 is an explanatory diagram showing an example of the record layout of the information collection DB 175. The information collection DB 175 includes columns for Management ID, Type, Problem ID, Problem, and Answer Candidates. The Management ID column stores the ID of the data to be collected, which is uniquely identified to identify each piece of data for information collection. The Type column stores the type of data to be collected. For example, the Type column might store "Question" or "Feedback." The Problem ID column stores the question ID or feedback item ID to be submitted to the customer. The Problem column stores the question or feedback item to be submitted to the customer. The Answer Candidates column stores the answer choices for the problem.

Figure 8 is a block diagram showing an example configuration of terminal 2. Terminal 2 includes a control unit 21, a storage unit 22, a communication unit 23, an input unit 24, and a display unit 25. Each component is connected by bus B.

The control unit 21 includes a processing unit such as a CPU and MPU, and performs various information processing and control processing related to the terminal 2 by reading and executing the control program 2P stored in the storage unit 22. Although Figure 8 describes the control unit 21 as a single processor, it may be a multiprocessor. The storage unit 22 includes memory elements such as RAM and ROM, and stores the control program 2P or data necessary for the control unit 21 to execute processing. The storage unit 22 also temporarily stores data necessary for the control unit 21 to execute arithmetic processing.

The communication unit 23 is a communication module for performing communication-related processing, and it sends and receives information with the server 1, etc., via the network N. The input unit 24 may be a keyboard, mouse, or a touch panel integrated with the display unit 25. The display unit 25 is a liquid crystal display or an organic EL display, etc., and displays various information according to the instructions of the control unit 21.

Figure 9 is a functional block diagram showing the processing operation of the beer beverage suggestion system. Terminal 2 accepts input of customer attributes. Customer attributes include gender, age, most frequently consumed alcoholic beverage, frequency of beer consumption, and drinking occasion (e.g., home drinking, dining out, etc.). Terminal 2 retrieves multiple questions pre-stored in the information collection DB 175 of Server 1 from Server 1 and displays the retrieved questions on the screen.

The questions are designed to gather information for suggesting beer beverages that suit the customer's attributes. For example, the questions could be "Do you like ethnic food?", "Do you prefer sweet or dry sake?", or "Are you trend-conscious?". In this embodiment, questions pre-stored in the information gathering database 175 of server 1 were sent to terminal 2, but this is not limited to this. For example, questions could be pre-stored in the storage unit 22 of terminal 2. Terminal 2 receives input from the customer with answers to multiple questions.

Terminal 2 transmits the acquired customer attributes and multiple responses to Server 1. Server 1 receives the customer attributes and multiple responses transmitted from Terminal 2 and stores them in Customer DB 171. Specifically, Server 1 assigns a customer ID and stores the nickname, gender, age, frequently consumed alcoholic beverages, frequency of beer consumption, drinking occasions, and multiple responses as a single record in Customer DB 171.

Server 1 outputs beer information to a beer information output model 176, which outputs beer information regarding beer, low-malt beer, or beer-flavored beverages that should be suggested based on the customer's attributes and multiple responses. The beer information includes multiple characteristic values that indicate the characteristics of the beer beverage. The process of outputting beer information using the beer information output model 176 will be described later.

Server 1 proposes beer beverages based on the beer information output from the beer information output model 176. Specifically, Server 1 calculates the similarity between multiple characteristic values of the beer beverages output from the beer information output model 176 and multiple characteristic values associated with various beer beverages stored in the product database 173. For the similarity calculation process, conventionally known methods can be used, such as vectorizing the characteristic values of the beer beverages and calculating the similarity using the distance between vectors, clustering methods such as k-means, or cosine similarity.

Server 1 extracts the beer beverage with the highest similarity from the product database 173 as the beer beverage to be proposed. However, this process is not limited to simply extracting beer beverages with high similarity. For example, it can prioritize proposing beer beverages with high similarity within a specified period from the registration date of the beer beverage. Specifically, if a predetermined similarity threshold (e.g., 70%) is set in advance, Server 1 retrieves all beer beverage IDs from the product database 173 whose calculated similarity is equal to or greater than the predetermined threshold. Based on the retrieved beer beverage IDs, Server 1 extracts beer beverages from the store database 172 within the specified period. The specified period is, for example, within three months from the start date (registration date) of handling the beer beverage. From the extracted beer beverages within the specified period, Server 1 extracts the beer beverage with the highest similarity.

Server 1 stores the extracted beer beverage in the history database 174, associating it with the customer ID. Specifically, Server 1 stores the extracted beer beverage ID and extraction date and time as a single record in the history database 174, associating them with the customer ID. Based on the customer ID, Server 1 retrieves the beer beverage previously offered to that customer from the history database 174. Server 1 then determines whether the extracted beer beverage matches the previously offered beer beverage.

Server 1, if it determines that the extracted beer beverage matches the previously proposed beer beverage, will change it to a different beer beverage. For example, as an example of beer beverage proposals based on similarity as described above, it will change the beer beverage to the next most similar beer beverage. Note that if the previous proposal date was more than a predetermined period (e.g., six months), it is not necessary to change the beer beverage proposed to the customer through the beer information output model 176.

Furthermore, the process of changing beer beverages is not limited to the similarity of the characteristic values of the beer beverages. For example, from among multiple beer beverages classified in the same category as the proposed beer beverage, the beer beverage with the highest popularity ranking may be extracted, and the proposed beer beverage may be changed to the extracted beer beverage. Also, the various processes described above may be executed on terminal 2.

Server 1 sends the modified beer beverage to Terminal 2 as the proposed beer beverage. Terminal 2 receives the beer beverage sent from Server 1 and displays it on its screen. If Server 1 determines that the extracted beer beverage does not match the previously proposed beer beverage, it sends the beverage to Terminal 2 without modification.

Next, we will explain the process of outputting beer information using the beer information output model 176. Figure 10 is an explanatory diagram illustrating the beer information output model 176. The beer information output model 176 is used as a program module, which is part of the artificial intelligence software. The beer information output model 176 is a machine learning model with pre-trained parameters. For example, Server 1 uses customer attributes and answers to multiple questions as training data to generate (construct) the beer information output model 176. The training data is combination data created by associating beer information with customer attributes and multiple answers.

As shown in Figure 10, Server 1 generates a decision tree model as the beer information output model 176. For example, Server 1 generates the beer information output model 176 using gradient boosting, a type of ensemble learning. That is, Server 1 generates one weak classifier (decision tree) using the training data, and then sequentially generates the next weak classifier based on the error (more precisely, the residual) between the predicted value and the actual value. Server 1 sequentially generates weak classifiers by referencing the gradient of the loss function defined by the error between the predicted value and the actual value, taking into account the learning results of the previous weak classifier, and generates the final classifier, i.e., the beer information output model 176.

In addition to the learning methods described above, other ensemble learning methods may be used, such as bagging, which generates multiple weak classifiers in parallel. Furthermore, the beer information output model 176 may be generated using learning methods other than ensemble learning. For example, Server 1 generates the beer information output model 176 using the random forest method. Specifically, based on the learning data sampled from the training data, Server 1 randomly selects feature items to be used for identification (classification) at non-terminal nodes, creating multiple decision trees with low correlation, and then generates the beer information output model 176 using these multiple decision trees.

Furthermore, dimensionality reduction may be applied to the input data entered into the beer information output model 176.

As described above, Server 1 generates a beer information output model 176 using customer attributes and answers to multiple questions as training data. While this embodiment uses customer attributes and answers to multiple questions as training data, it is not limited to these. For example, information such as temperature, humidity, season, date and time, and the type of food served at the store may also be included in the training data.

Server 1 retrieves customer attributes and answers to multiple questions from the customer database 171 and uses them as training input parameters to generate a beer information output model 176 so that the output parameters approximate the correct values. Beer information consists of multiple characteristic values that represent the characteristics of beer beverages. Specifically, Server 1 generates the beer information output model 176 by training a machine learning model (such as a decision tree model) with parameters using the multiple characteristic values of beer beverages associated with the input customer attributes and multiple answers as correct values.

Server 1 uses the generated beer information output model 176 to output estimation results for multiple characteristic values of the beer beverage. As shown in the figure, the estimated values for "Original Extract," "True Extract," "True Fermentation Degree," "Appearance Extract," "Appearance Fermentation Degree," "Alcohol Content," "Extract Content," "Residual Fermentation Degree," "pH," and "Color" are output to the customer as "13.8," "5.5," "60.5," "3.8," "70.3," "5%," "22.6," "6.7," "5.2," and "260," respectively. Note that the estimation results output from the output layer are not limited to the format described above. For example, the estimation results may be discrete values indicating the presence or absence of breweries or adjuncts (e.g., values of "0" or "1"), or continuous probability values (e.g., values in the range from "0" to "1"). While the above-described beer beverage characteristic values were used for explanation, the illustration and explanation of other characteristic values (e.g., values of ethyl acetate, isobutanol, isoamyl alcohol, etc.) as output results are omitted.

Furthermore, training data can be created based on customer feedback, and the beer information output model 176 can be retrained using this training data. Specifically, Server 1 retrieves customer feedback results from the history database 174 based on the customer ID. Server 1 analyzes the retrieved feedback results (responses to feedback items) and extracts beer beverage IDs that match the customer's preferences. For example, if the customer's response to the feedback item "How did you like the taste of the beer you drank?" is "It was delicious" or "It was very delicious," Server 1 extracts the beer beverage ID for that beer. Alternatively, if the customer's response to the feedback item "How many beers have you had now?" is "Three or more," Server 1 extracts the beer beverage ID for that beer.

Server 1 retrieves multiple characteristic values of beer beverages from the product database 173 based on the extracted beer beverage ID. Server 1 then creates training data by associating the retrieved characteristic values of beer beverages with customer attributes and multiple responses. Using this training data, Server 1 retrains the beer information output model 176 in the same manner as the learning process described above. Note that the training data is not limited to collecting beer beverages corresponding to high ratings such as "delicious" or "very delicious," but may also be created based on beer beverages corresponding to low ratings such as "not delicious" or "mediocre."

In this embodiment, the beer information output model 176 is described as a decision tree. However, the beer information output model 176 is not limited to a decision tree; it may be a pre-trained model constructed using any learning algorithm other than a decision tree, such as a CNN (Convolutional Neural Network), R-CNN (Regions with Convolutional Neural Networks), SVM (Support Vector Machine), Bayesian network, or regression tree.

In this embodiment, an example was described in which the beer information output from the beer information output model 176 consists of multiple characteristic values of a beer beverage, but this is not the only example. For example, the beer information related to a beer beverage may be a beer beverage ID. That is, customer attributes and multiple responses are input to the beer information output model 176, and the beer beverage ID of the beer beverage to be suggested is output.

Figures 11 and 12 are flowcharts illustrating the processing procedure when changing to a different beer beverage than the one initially proposed. Note that the customer's basic information (e.g., nickname, email address, etc.) is pre-registered in the customer database 171. Specifically, terminal 2 transmits the received basic information to server 1. Server 1 receives the basic information transmitted from terminal 2. Server 1 assigns a customer ID and stores the received basic information in customer database 171, associating it with the customer ID. Then, it performs a login process for registered customers.

The control unit 21 of terminal 2 determines whether the customer is already logged in (step S291). If the control unit 21 determines that the customer is already logged in (YES in step S291), it proceeds to step S201, which will be described later. If the control unit 21 determines that the customer is not logged in (NO in step S291), the control unit 21 sends authentication information, including the customer ID and password, to the server 1 via the communication unit 23 (step S292). Note that the authentication information is not limited to the customer ID and password; for example, it may include information for fingerprint or facial recognition.

The control unit 11 of server 1 receives authentication information transmitted from terminal 2 via the communication unit 13 (step S191). Based on the received authentication information, the control unit 11 performs authentication processing (step S192) and transmits the authentication result (login successful, login failed, etc.) to terminal 2 via the communication unit 13 (step S193). The control unit 21 of terminal 2 receives the authentication result transmitted from server 1 via the communication unit 23 (step S293). Based on the received authentication result, the control unit 21 determines whether the login was successful or not (step S294).

If the control unit 21 determines that the login was unsuccessful (NO in step S294), it proceeds to step S291. If the control unit 21 determines that the login was successful (YES in step S294), it accepts the customer's attributes via the input unit 24 (step S201). Note that this process is not limited to accepting customer attributes via login. For example, if customer attributes can be obtained using local storage or cookies, the input of customer attributes may be omitted. The control unit 11 of the server 1 obtains multiple questions from the information collection DB 175 of the large-capacity storage unit 17 (step S101). The control unit 11 transmits the obtained multiple questions to the terminal 2 via the communication unit 13 (step S102).

The control unit 21 of terminal 2 receives multiple questions transmitted from server 1 via the communication unit 23 (step S202), and displays the received multiple questions via the display unit 25 (step S203). The control unit 21 receives input from the customer for answers to the multiple questions via the input unit 24 (step S204). The control unit 21 transmits the customer ID, the received customer attributes, and the multiple answers to server 1 via the communication unit 23 (step S205).

The control unit 11 of server 1 receives the customer ID, customer attributes, and multiple responses transmitted from terminal 2 via the communication unit 13 (step S103). The control unit 11 stores the received customer attributes and multiple responses in the customer database 171, associating them with the customer ID (step S104). The control unit 11 outputs beer information using a beer information output model 176, which outputs beer information related to beer beverages, including beer, low-malt beer, or beer-flavored beverages, that should be suggested based on the customer attributes and multiple responses (step S105). The control unit 11 suggests beer beverages based on the beer information output from the beer information output model 176 (step S106). The beer beverage suggestion subroutine will be described later.

The control unit 11 stores the proposed beer beverage in the history database 174, associating it with the customer ID (step S107). Based on the customer ID, the control unit 11 retrieves the previously proposed beer beverage from the history database 174 and determines, via the beer information output model 176, whether the proposed beer beverage matches the previously proposed one (step S108). If the control unit 11 determines, via the beer information output model 176, that the proposed beer beverage does not match the previously proposed one (NO in step S108), it proceeds to step S111, which will be described later.

If the control unit 11 determines that the beer beverage suggested to the customer via the beer information output model 176 matches the previously suggested beer beverage (YES in step S108), it changes it to a different beer beverage (step S109). For example, if the control unit 11 suggests beer beverages based on similarity, it changes the beer beverage suggested to the customer via the beer information output model 176 to the next most similar beer beverage. The control unit 11 updates the history database 174 by associating the changed beer beverage with the customer ID (step S110).

The control unit 11 transmits the proposed beer beverage to the terminal 2 via the communication unit 13 (step S111). The control unit 21 of the terminal 2 receives the beer beverage transmitted from the server 1 via the communication unit 23 (step S206), and displays the received beer beverage on the screen via the display unit 25 (step S207).

In this embodiment, we have described an example where, if the beer beverage suggested to the customer via the beer information output model 176 matches the previously suggested beer beverage, the system changes to a different beer beverage. However, this is not the only example. For instance, based on the history of beer beverages suggested to the customer, if the same beer beverage has been suggested a predetermined number of times consecutively (e.g., three times), the system may change it to a different beer beverage.

Furthermore, the method of modification is not limited to the one described above. For example, the beer beverage could be changed to a different one for each glass. For instance, for the first glass of beer, the beer beverage suggested to the customer through the beer information output model 176 could be changed to the next most similar beer beverage. For the second glass of beer, the beer beverage suggested to the customer through the beer information output model 176 could be changed to the third most similar beer beverage.

Figure 13 is a flowchart showing the processing procedure of a subroutine that suggests beer beverages. The control unit 11 calculates the similarity between multiple characteristic values of beer beverages output from the beer information output model 176 and multiple characteristic values associated with various beer beverages stored in the product database 173, for example, using cosine similarity (step S10). The control unit 11 obtains a predetermined threshold for similarity (for example, 70%) from the storage unit 12 (step S11). The control unit 11 obtains all beer beverage IDs from the product database 173 whose calculated similarity is equal to or greater than the predetermined threshold (step S12).

The control unit 11 extracts beer beverages from the store database 172 within a specified period (for example, within three months from the start of sales) based on the acquired beer beverage ID (step S13). The control unit 11 then extracts the beer beverage with the highest similarity from the extracted beer beverages within the specified period (step S14) and returns to the original process that called the subroutine.

Figure 14 is an explanatory diagram showing an example of a screen for receiving customer attributes on terminal 2. This screen includes a customer attribute input field 11a and a confirmation button 11b. The customer attribute input field 11a is an input field for receiving customer attribute input. The confirmation button 11b is a button for sending the customer attributes entered by the customer to server 1. As shown in the figure, customer attributes (customer information) include nickname, gender, age, most frequently consumed alcoholic beverage, frequency of beer consumption, drinking occasion (drinks at home or drinks out), etc.

When terminal 2 receives a touch (click) operation on the confirmation button 11b, it sends the customer attributes entered in the customer attribute input field 11a to server 1. Server 1 receives the customer attributes sent from terminal 2 and stores the received customer attributes in customer DB 171, associating them with the customer ID. Note that if the customer attributes have already been stored in customer DB 171, or if the customer attributes are retrieved from local storage, terminal 2 may omit the process in step S201 (Figure 11) from the second time onward. Login authentication information, including the customer ID, may be entered on this screen, or on a separate login screen.

Figure 15 is an explanatory diagram showing an example of a screen on terminal 2 that accepts answers to multiple questions. This screen includes a progress display area 12a, a question display area 12b, and answer candidate buttons 12c. The progress display area 12a is a display area for showing the progress of answering questions. For example, the numbers of questions that have been answered and are being answered may be shown with solid lines, and the numbers of unanswered questions may be shown with dotted lines. The question display area 12b is a display area for showing the content of the questions. The answer candidate buttons 12c are buttons for selecting answer candidates (options) for a question.

Server 1 retrieves multiple questions stored in the information collection DB 175 and sends them to Terminal 2. Terminal 2 receives the multiple questions sent from Server 1 and displays them on its screen. For example, one question might be, "Do you prefer Japanese or Western-style breakfast? Please answer with the most common answer." When Terminal 2 receives a touch operation on any of the answer candidate buttons 12c, it retrieves the corresponding answer and displays the next question. For example, if Terminal 2 receives a touch operation on the answer candidate button 12c for "Japanese-style breakfast," it retrieves the answer "Japanese-style breakfast." Terminal 2 then displays the next question in the question display area 12b.

Terminal 2 sends the customer ID and multiple answers selected via the answer candidate button 12c to Server 1. Server 1 receives the customer ID and multiple answers sent from Terminal 2, associates the received multiple answers with the customer ID, and stores them in the customer database 171.

Figure 16 is an explanatory diagram showing an example of a screen displaying the beer beverage before the change. Figure 16 illustrates a screen example where, if the beer beverage suggested to the customer via the beer information output model 176 does not match the previously suggested beer beverage, the original beer beverage (the original beer beverage) is suggested to the customer without change. This screen includes a character 61, a beer display area 71, an order button 72, and a cancel button 73. Character 61 is a character that provides guidance to the customer. As shown in the figure, terminal 2 displays a character 61 with a troubled expression regarding the original beer beverage.

Terminal 2 displays character 61 in the display area at the top of the screen, and also displays a comment (text) as character 61's dialogue, recommending a beer beverage to be displayed in the beer display area 71. The dialogue could be, for example, "Here's our recommended beer for you, Mr./Ms. XX," or "I'll do my best with the next suggestion!" The dialogue is not limited to text; it can also be voice.

The beer display area 71 is a display area for displaying beer information about the beer beverage proposed to the customer through the beer information output model 176. For example, terminal 2 displays the product name, description, and product image of the proposed beer beverage in the beer display area 71. Terminal 2 also displays a detailed profile of the proposed beer beverage, including its beer style, alcohol content (concentration), calories, manufacturer (brewery), and production region. Beer style refers to the classification of beer based on ingredients, brewing method, alcohol content, etc., such as IPA (India Pale Ale) and Weizen. In addition, terminal 2 displays a radar chart representing the flavor (body, acidity, sweetness, aroma, bitterness, etc.) of the beer beverage using multiple scales. The radar chart may be automatically created by, for example, server 1 processing multiple characteristic values of the beer beverage stored in the product database 173, or it may be created manually by the system administrator.

The order button 72 and the cancel button 73 are operation objects for selecting whether or not to order (drink) the suggested beer beverage. If the order button 72 is pressed, the suggested beer beverage is ordered. In this case, for example, server 1 processes the order for the beer beverage and notifies the store staff of the order. Note that if the order button 72 is pressed, a confirmation screen for entering the order quantity or confirming the order details may be displayed first. If integration with the store's ordering system is possible, the order processing for the suggested beer beverage may be performed by the store's server. If the cancel button 73 is pressed, the suggested beer beverage is not ordered and the process ends. In this case, for example, server 1 may select another beer beverage (such as the next most similar beer beverage) and display it on the suggestion screen, or it may end the series of suggestion processes.

Figure 17 is an explanatory diagram showing an example of a screen displaying the changed beer beverage. Note that content that overlaps with Figure 16 is denoted by the same reference numerals and its explanation is omitted. Figure 17 illustrates an example screen when, if the beer beverage proposed to the customer via the beer information output model 176 (the beer beverage before the change) matches the previously proposed beer beverage, a different beer beverage (the changed beer beverage) is proposed to the customer.

If Server 1 determines that the beer beverage suggested to the customer via the beer information output model 176 matches the previously suggested beer beverage, it transmits beer information about the changed beer beverage and a character (image or icon, etc.) to guide the customer to Terminal 2. Terminal 2 displays the received beer information about the changed beer beverage in the beer display area 71. Similar to Figure 16, the beer information includes the product name, description, product image, beer style, alcohol content (concentration), calories, manufacturer (brewery), production region, or a radar chart representing the flavor of the changed beer beverage on multiple scales.

Terminal 2 displays the received character on character 61. Server 1 changes the display of character 61 if it changes the beer beverage suggested to the customer via the beer information output model 176. Specifically, Server 1 changes character 61's facial expression and comment from the previous version. For example, as shown in Figure 17, if Server 1 suggests a changed beer beverage, it changes character 61's facial expression from the screen in Figure 16 to a confident expression. Alternatively, Terminal 2 may display a phrase such as "A recommendation for thoughtfulness!" as a comment for character 61.

Furthermore, the screen includes a pre-change beer display field 74. The pre-change beer display field 74 is a display field for showing beer information related to the beer beverage before the change. Server 1 acquires beer information related to the beer beverage before the change and transmits it to terminal 2. Terminal 2 displays the beer information transmitted from server 1 in the pre-change beer display field 74. The beer information displayed in the pre-change beer display field 74 includes, for example, the product name and description of the beer beverage before the change.

Furthermore, the screen includes a candidate field 75. The candidate field 75 is a display field for displaying other beer beverages as candidates, in addition to the modified beer beverage displayed in the beer display field 71. Terminal 2 displays one or more other beer beverages as candidates in the candidate field 75, including product images, product names, etc. Specifically, for example, Server 1 extracts the beer beverage with the next highest similarity to the modified beer beverage displayed in the beer display field 71 and transmits beer information related to the extracted beer beverage to Terminal 2. Terminal 2 receives the beer information about the next highest similarity beer beverage transmitted from Server 1 and displays the received beer information in the candidate field 75. Note that the selection method for other beer beverages besides the modified beer beverage is not limited to similarity. For example, if the release date is used as the selection criterion for the beer beverage to be proposed, the next newest beer beverage with a release date may also be selected as a candidate, and the selection method for candidate beer beverages depends on the method used.

Furthermore, the beer information regarding the beer beverage before the change, the beer information regarding the beer beverage after the change, and the beer information regarding the candidate beer beverage after the change may be simultaneously transmitted from server 1 to terminal 2.

If the order button 72 is pressed, the suggested modified beer beverage will be ordered. In this case, for example, server 1 will process the order for the beer beverage and notify the store staff of the order. Note that when the order button 72 is pressed, a confirmation screen may be displayed first, allowing the user to enter the order quantity or confirm the order details. If integration with the store's ordering system is possible, the order processing for the suggested beer beverage may be performed by the store's server. If the cancel button 73 is pressed, the suggested modified beer beverage will not be ordered, and the process will end. In this case, for example, server 1 may display the suggested beer beverage shown in the candidate field 75 in the beer display field 71, or it may end the series of suggestion processes.

Furthermore, the screen includes a "Pre-Change Order" button 74a and a "Candidate Order" button 75a. The "Pre-Change Order" button 74a is an operation object for ordering the beer beverage displayed in the "Pre-Change Beer Display" section 74. When the "Pre-Change Order" button 74a is pressed, the pre-change beer beverage is ordered. The "Candidate Order" button 75a is an operation object for ordering the beer beverage displayed in the "Candidate" section 75. When the "Candidate Order" button 75a is pressed, the candidate "Changed Beer Beverage" is ordered.

When either the original beer beverage, the modified beer beverage, or a candidate modified beer beverage is ordered, Server 1 stores the order details of the ordered beer beverage in the history DB 174. The order details include, for example, the beer beverage ID, order quantity, and order date and time. Furthermore, the beer information output model 176 can be retrained depending on whether or not a beer beverage was ordered. For example, if the customer did not order (did not drink) the proposed modified beer beverage, the characteristic values of the proposed modified beer beverage are modified (for example, changed to the characteristic values of a different beer beverage ordered by the customer), and the beer information output model 176 is updated with the modified characteristic values as the correct values for retraining. Also, if a candidate modified beer beverage is ordered (wanted to drink), the characteristic values of the proposed modified beer beverage are modified (for example, changed to the characteristic values of the candidate modified beer beverage), and the beer information output model 176 is updated with the modified characteristic values as the correct values for retraining.

According to this embodiment, based on customer attributes and answers to multiple questions, it becomes possible to output beer information regarding beer beverages to be recommended using the beer information output model 176.

According to this embodiment, based on the history of proposed beer beverages, it becomes possible to change to a different beer beverage if it matches the previously proposed beer beverage.

According to this embodiment, by changing the beer beverage to one different from the one previously proposed, duplication of proposals can be avoided, making it possible to propose a suitable beer beverage to the customer.

(Embodiment 2)
Embodiment 2 relates to a configuration in which, if the beer beverage to be proposed matches the previously proposed beer beverage, the proposed beer beverage is changed to a different one, taking into account the feedback results of the previous proposal. Note that explanations of content that overlaps with Embodiment 1 will be omitted.

Server 1, via the beer information output model 176, determines that the beer beverage it suggested to the customer matches the previously suggested beer beverage, and that the feedback result for that beer beverage is negative, then changes it to a different beer beverage. For example, let's consider a scenario where the feedback item for the suggested beer beverage, "How did you like the taste of the beer you drank?", uses a five-point rating scale: "Not tasty," "Not great," "Average," "Delicious," and "Very delicious." For instance, if Server 1 receives negative feedback from the customer regarding the taste of the previously suggested beer beverage A ("Not tasty," "Not great," or "Average"), it changes it to a different beer beverage B.

Figure 18 is a flowchart illustrating the processing procedure for receiving feedback results for beer beverages. The control unit 11 of server 1 retrieves multiple feedback items from the information collection DB 175 of the large-capacity storage unit 17 (step S121). The control unit 11 then transmits the beer beverage ID proposed to the customer and the retrieved feedback items to terminal 2 via the communication unit 13 (step S122).

The control unit 21 of terminal 2 receives the beer beverage ID and multiple feedback items transmitted from server 1 via the communication unit 23 (step S221), and displays the received items via the display unit 25 (step S222). The control unit 21 receives the customer's responses to the multiple items via the input unit 24 (step S223). The control unit 21 transmits the customer ID, beer beverage ID, and feedback results including the received responses to server 1 via the communication unit 23 (step S224).

The control unit 11 of server 1 receives the feedback results transmitted from terminal 2 via the communication unit 13 (step S123). The control unit 11 stores the received feedback results in the history DB 174 of the large-capacity storage unit 17 (step S124). Specifically, the control unit 11 stores the responses to multiple feedback items in the feedback result column, associated with the customer ID and beer beverage ID.

Figure 19 is a flowchart showing the processing procedure when changing the beer beverage based on feedback results. Note that the same reference numerals are used for elements that overlap with Figure 12, and their explanations are omitted. If the control unit 11 of the server 1 determines that the beer beverage proposed to the customer via the beer information output model 176 matches the previously proposed beer beverage (YES in step S108), it obtains the feedback result for that beer beverage from the history DB 174 of the large-capacity storage unit 17 based on the customer ID and beer beverage ID (step S112). The feedback result may be, for example, "No. 1 (How did the beer you drank taste?): Not tasty, No. 2 (What food did you drink the beer with?): Snacks, No. 3 (What number beer have you drunk now?): 1st glass," etc.

The control unit 11 determines whether the feedback result for the beer beverage indicates a negative result (step S113). For example, if the response to feedback item No. 1 for the beer beverage is "not tasty," "not great," or "average," the control unit 11 determines that the feedback result for the beer beverage indicates a negative result. Conversely, if the response to feedback item No. 1 for the beer beverage is "tasty" or "very tasty," the control unit 11 determines that the feedback result for the beer beverage indicates a positive result.

If the control unit 11 determines that the feedback result for the beer beverage indicates a negative (YES in step S113), it executes the process in step S109 to change the beer beverage. For example, if the control unit 11 proposes a beer beverage based on similarity (e.g., cosine similarity), it identifies multiple beer beverages as output candidates in descending order of similarity value. The control unit 11 changes the beer beverage from the identified multiple beer beverages to the next most similar beer beverage. Alternatively, the control unit 11 extracts the beer beverage ranked number one in popularity from multiple beer beverages belonging to the same category as the beer beverage in question, and changes the beer beverage to the extracted beer beverage. In other words, the process of changing the beer beverage is not limited to the processing method described above, and any algorithm may be adopted. If the control unit 11 determines that the feedback result for the beer beverage indicates a positive (NO in step S113), it proceeds to the process in step S111.

Figure 20 is an explanatory diagram showing an example of a screen for receiving feedback on terminal 2. This screen includes a logo display area 14a, a beer beverage display area 14b, an item display area 14c, a suggested answer button 14d, and an item progress display area 14e. The logo display area 14a is for displaying the logo of the beer beverage. The beer beverage display area 14b is for displaying the name and alcohol content of the beer beverage.

The item display area 14c is a display area for displaying the feedback items. The answer candidate button 14d is a button for selecting an answer candidate for the feedback item. Note that the answer candidate button 14d may be represented by numbers or icons (e.g., a star mark) instead of text. The item progress display area 14e is a display area for displaying the progress of responses for multiple feedback items. For example, the numbers of items that have been answered and are being answered may be shown with a solid line frame, and the numbers of unanswered items may be shown with a dotted line frame.

Server 1 retrieves multiple feedback items stored in the information collection DB 175 and retrieves the name (brand) and alcohol content of the beer beverage stored in the product DB 173. Server 1 transmits the retrieved feedback items, the beer beverage name, and the alcohol content to Terminal 2. Terminal 2 receives the feedback items, the beer beverage name, and the alcohol content transmitted from Server 1. Terminal 2 displays the received feedback items in the item display field 14c and the beer beverage name and alcohol content in the beer beverage display field 14b. The feedback items are, for example, "How did you like the taste of the beer you drank?". When Terminal 2 receives a touch operation on any of the answer candidate buttons 14d, it retrieves the corresponding answer for that item and displays the next item. For example, if Terminal 2 receives a touch operation on the answer candidate button 14d for "Normal," it retrieves the answer "Normal." Terminal 2 displays the next item in the item display field 14c.

Terminal 2 transmits the customer ID, beer beverage ID, and multiple responses selected via the response candidate button 14d to Server 1. Server 1 receives the customer ID, beer beverage ID, and multiple responses transmitted from Terminal 2, and stores the received multiple responses (feedback results) in the history database 174, associating them with the customer ID and beer beverage ID.

According to this embodiment, by considering feedback results on beer beverages, it becomes possible to accurately grasp customer needs and propose beer beverages that precisely meet those needs.

(Embodiment 3)
Embodiment 3 relates to a method of changing the proposed beer beverage to one belonging to the same category as the previously proposed beer beverage, if the proposed beer beverage is the same as the previously proposed beer beverage. Note that explanations of content that overlaps with Embodiments 1 and 2 will be omitted.

Figure 21 is a block diagram showing an example configuration of the server 1 of Embodiment 3. Note that components overlapping with those in Figure 2 are denoted by the same reference numerals and their descriptions are omitted. The large-capacity storage unit 17 stores the category classification model 177 and the category classification database 178. The category classification model 177 is used as a program module, which is part of the artificial intelligence software. It is a trained model that classifies each beer beverage into multiple categories based on learning from multiple characteristic values of each beer beverage. The category classification database 178 stores the classification information of the categories classified through the category classification model 177.

Figure 22 is an explanatory diagram showing an example of the record layout of Category Classification DB 178. Category Classification DB 178 includes a Category ID column and a Product ID column. The Category ID column stores the uniquely identified category ID to identify each category. The Product ID column stores the product ID to identify a product.

Next, we will explain the process of changing the beer beverage to one belonging to the same category as the original beer beverage. First, if the category of the beer beverage is not yet classified, Server 1 uses the category classification model 177 to classify each beer beverage into one of several categories. In this embodiment, we will explain an example of category classification using an unsupervised learning algorithm. Unsupervised learning is a method in which a device learns the distribution of input data by providing only a large amount of input data, and then performs compression, classification, and formatting of the input data without providing corresponding supervised output data.

Specifically, Server 1 retrieves characteristic data (multiple characteristic values) for each beer beverage from the product database 173 based on each beer beverage ID. Server 1 inputs the retrieved characteristic data for each beer beverage into the category classification model 177. The category classification model 177 accepts the characteristic data and determines the category associated with each beer beverage by clustering based on the vector representation of the received characteristic data. The category classification model 177 may include models such as k-means or random forest. Server 1 assigns a category ID to the categories determined through the category classification model 177. Server 1 stores the assigned category IDs in association with each beer beverage ID in the category classification database 178.

Furthermore, the category classification process may be performed using not only unsupervised learning algorithms, but also supervised learning machine learning techniques such as SVM. Alternatively, the output of the hidden layer of a neural network trained using supervised learning may be extracted and used as a clustering algorithm.

Furthermore, in the event of new beer beverage arrivals or discontinuation of beer beverage sales, Server 1 reclassifies each beer beverage into multiple categories using the category classification model 177. The reclassification process is the same as the classification process described above, and therefore, its explanation is omitted.

Next, if Server 1 determines that the beer beverage suggested to the customer via the beer information output model 176 matches the previously suggested beer beverage, it extracts beer beverages belonging to the same category as the suggested beer beverage. For example, Server 1 may extract beer beverages with high similarity based on multiple characteristic values from multiple beer beverages belonging to the same category as the suggested beer beverage, similar to Embodiment 1. Alternatively, Server 1 may extract beer beverages to be suggested from multiple beer beverages belonging to the same category in order of popularity ranking, price (lowest first), or date of availability (newest first). Server 1 transmits the extracted beer beverages to Terminal 2 as the beer beverages to be suggested. Terminal 2 receives and displays the beer beverages transmitted from Server 1.

Figure 23 is a flowchart showing the processing procedure when changing to a beer beverage belonging to the same category. Note that the same reference numerals are used for parts that overlap with Figure 12, and their explanations are omitted. If the control unit 11 of the server 1 determines that the beer beverage proposed to the customer through the beer information output model 176 matches the previously proposed beer beverage (YES in step S108), it extracts all beer beverage IDs belonging to the same category as the proposed beer beverage (step S114). Specifically, the control unit 11 obtains the category ID of the beer beverage from the category classification DB 178 based on the beer beverage ID. Based on the obtained category ID, the control unit 11 extracts all beer beverage IDs belonging to that category from the category classification DB 178. Based on all the extracted beer beverage IDs, the control unit 11 obtains the beer beverages to be changed from the product DB 173, for example, in order of the newest start date (step S115). Note that the control unit 11 may randomly select beer beverages other than the previously proposed beer beverage from among multiple beer beverages belonging to the same category as the proposed beer beverage. Subsequently, the control unit 11 executes the process in step S109.

According to this embodiment, if the beer beverage suggested to the customer via the beer information output model 176 matches the previously suggested beer beverage, it becomes possible to change the suggested beer beverage to one belonging to the same category as the previously suggested beer beverage.

According to this embodiment, it becomes possible to propose the optimal beer beverage tailored to the customer's preferences.

(Embodiment 4)
Embodiment 4 relates to a configuration in which the beer information output model 176 is updated using customer feedback results. Customer feedback results include responses to multiple feedback items. Note that explanations of content that overlaps with Embodiments 1 to 3 will be omitted.

If the beer beverage suggested to the customer via the beer information output model 176 does not match the previously suggested beer beverage, the original beer beverage (the one before the change) is suggested to the customer without modification. Conversely, if the beer beverage suggested to the customer via the beer information output model 176 matches the previously suggested beer beverage, a different beer beverage (the changed beer beverage) is suggested to the customer. As a result, the feedback results include feedback for both the original and changed beer beverages. Either feedback result includes information such as the customer's preference for the beer beverage and the number of glasses consumed.

The feedback results stored in the history database 174 are used to retrain the beer information output model 176. Specifically, Server 1 creates first training data based on the feedback results for the beer beverage before the change. The first training data includes customer attributes and answers to multiple questions, beer information about the beer beverage before the change, and customer feedback on that beer information. Server 1 then creates second training data based on the feedback results for the beer beverage after the change. The second training data includes customer attributes and answers to multiple questions, beer information about the beer beverage after the change, and customer feedback on that beer information.

Server 1 uses the customer attributes included in the created first and second training data, along with the answers to multiple questions, as input data for retraining (training). It then uses the proposed beer beverages as the ground truth values for retraining and updates the beer information output model 176.

In this case, it is preferable for Server 1 to retrain based on whether the suggested beer beverage was ordered, the responses to multiple feedback items entered after the order, etc. For example, if the customer did not order the suggested beer beverage, Server 1 modifies the characteristic values of the suggested beer beverage (for example, by changing them to the characteristic values of a different beer beverage ordered by the customer) and updates the beer information output model 176 using the modified characteristic values as the correct values for retraining. Also, if a beer beverage displayed in the candidate field 75 instead of the beer display field 71 is ordered, Server 1 modifies the characteristic values of the suggested beer beverage (for example, by changing them to the characteristic values of a candidate beer beverage) and updates the beer information output model 176 using the modified characteristic values as the correct values for retraining. Furthermore, for example, if the customer's feedback result regarding the beer beverage consumed was negative, Server 1 modifies the characteristic values of the suggested beer beverage (for example, by changing them to the characteristic values of a beer beverage corresponding to a positive response to the feedback) and updates the beer information output model 176 using the modified characteristic values as the correct values for retraining.

The above retraining method is merely an example, and this embodiment is not limited to it. For example, if the feedback result includes a question asking about the beer beverage requested by the user, Server 1 may change the characteristic values of the suggested beer beverage (e.g., change the alcohol content) according to the answer to that question (e.g., "I'd prefer something with a higher alcohol content"), and update the beer information output model 176 using the changed characteristic values as the correct values for retraining. In other words, the correct values of the beer information (characteristic values) used for retraining may be changed according to the content of the answers to multiple items in the feedback.

In this way, the accuracy of proposals can be improved through the operation of this system, based on customer feedback and other factors.

Figure 24 is a flowchart showing the procedure for updating the beer information output model 176. Based on Figure 24, the process for updating the beer information output model 176 through retraining will be explained.

The control unit 11 of server 1 obtains feedback results for retraining from the history DB 174, etc. (step S131). The control unit 11 creates first training data for retraining for the beer beverage before the change (step S132). The first training data includes customer attributes, answers to multiple questions, characteristic values of the beer beverage before the change, whether the beer beverage before the change was ordered, and answers to multiple items of feedback entered after the order. The control unit 11 creates second training data for retraining for the beer beverage after the change (step S133). The second training data includes customer attributes, answers to multiple questions, characteristic values of the beer beverage after the change, whether the beer beverage after the change was ordered, and answers to multiple items of feedback entered after the order.

The control unit 11 updates the beer information output model 176 based on the first and second training data for retraining (step S134). For example, if the customer does not order the suggested beer beverage, the control unit 11 updates the parameters of the beer information output model 176 so that the loss function increases. Also, for example, if the customer's feedback on the suggested beer beverage is positive or negative, the control unit 11 updates the parameters of the beer information output model 176 so that the loss function decreases or increases. The control unit 11 then completes the series of processes.

According to this embodiment, by retraining the beer information output model 176 using customer feedback results, it is possible to improve the estimation accuracy of the beer information output model 176.

The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of the invention is indicated by the claims, not in the sense described above, and all modifications within the sense and scope equivalent to the claims are intended.

1. Information processing device (server)
11 Control Unit 12 Storage Unit 13 Communication Unit 14 Input Unit 15 Display Unit 16 Reading Unit 17 Large Capacity Storage Unit 171 Customer Database
172 Store Database
173 Product DB
174 History DB
175 Information Gathering Database
176 Beer Information Output Model 177 Category Classification Model 178 Category Classification Database
1a Portable storage medium 1b Semiconductor memory 1P Control program 2 Information processing terminal (terminal)
21 Control Unit 22 Storage Unit 23 Communication Unit 24 Input Unit 25 Display Unit 2P Control Program

Claims (9)

Obtain customer attributes,
We obtained answers to several questions from the aforementioned customer,
A learning model that outputs beer information regarding beer, low-malt beer, new genre beer, or beer-flavored beverages to be suggested when customer attributes and multiple responses are input, outputs beer information that includes characteristic values of at least two of the following for the beer beverage to be suggested when customer attributes and multiple responses are input: alcohol content, bitterness value, color, and ethyl acetate.
Based on the beer information, which consists of characteristic values including alcohol content, bitterness value, color, and at least two of ethyl acetate, the history of the proposed beer beverage is stored in association with the customer.
When displaying a screen that includes a first display field for displaying beer information relating to the previously proposed beer beverage, a second display field for displaying beer information relating to a modified beer beverage different from the previously proposed beer beverage, and a third display field for displaying beer information relating to a candidate beer beverage that is different from the modified beer beverage,
The aforementioned screen is displayed when the beer beverage to be suggested to the customer based on the history matches the beer beverage suggested previously, and the computer performs a process to display the previously suggested beer beverage in the first display field and the changed beer beverage in the second display field. The aforementioned learning model outputs multiple characteristic values that represent the characteristics of beer beverages,
The information processing method according to claim 1, wherein a memory unit that stores multiple characteristic values associated with beer beverages is referenced to extract beer beverages that have a high similarity to the multiple characteristic values output from the learning model. The information processing method according to claim 2, wherein if the beer beverage suggested to the customer through the learning model matches the beer beverage suggested previously, the method changes to the next beer beverage with the highest similarity. The information processing method according to any one of claims 1 to 3, wherein if the beer beverage suggested to the customer through the learning model matches the beer beverage suggested previously, and the feedback result for the beer beverage indicates a negative, the beer beverage is changed to a different beer beverage. Each beer beverage is classified into multiple categories based on unsupervised learning using multiple characteristic values of each beer beverage.
The information processing method according to any one of claims 1 to 4, wherein if the beer beverage suggested to the customer through the learning model matches the beer beverage suggested previously, the method extracts beer beverages belonging to the same category as the previously suggested beer beverage. Obtain customer attributes,
We obtained answers to several questions from the aforementioned customer,
A learning model that outputs beer information regarding beer, low-malt beer , new genre beer , or beer-flavored beverages to be suggested when customer attributes and multiple responses are input, outputs beer information that includes characteristic values of at least two of the following for the beer beverage to be suggested when customer attributes and multiple responses are input: alcohol content, bitterness value, color, and ethyl acetate.
Based on the beer information, which consists of characteristic values including alcohol content, bitterness value, color, and at least two of ethyl acetate, the history of the proposed beer beverage is stored in association with the customer.
When displaying a screen that includes a first display field for displaying beer information relating to the previously proposed beer beverage, a second display field for displaying beer information relating to a modified beer beverage different from the previously proposed beer beverage, and a third display field for displaying beer information relating to a candidate beer beverage that is different from the modified beer beverage,
The aforementioned screen is displayed when the beer beverage to be suggested to the customer, based on the history, matches the beer beverage suggested previously. The program causes the computer to perform a process that displays the previously suggested beer beverage in the first display field and the changed beer beverage in the second display field. A first acquisition unit that acquires customer attributes,
A second acquisition unit that obtains answers to multiple questions from the aforementioned customer,
A learning model that outputs beer information regarding beer, low-malt beer , new genre beer , or beer-flavored beverages to be suggested when customer attributes and multiple responses are input, includes an output unit that outputs beer information which is a characteristic value including at least two of the following for the beer beverage to be suggested when customer attributes and multiple responses are input:
A storage unit that stores a history of beer beverages proposed based on beer information, which is a characteristic value including at least two of the outputted alcohol content, bitterness value, color, and ethyl acetate, in association with the customer.
When displaying a screen that includes a first display field for displaying beer information relating to the previously proposed beer beverage, a second display field for displaying beer information relating to a modified beer beverage different from the previously proposed beer beverage, and a third display field for displaying beer information relating to a candidate beer beverage that is different from the modified beer beverage,
The information processing device is characterized in that the screen is displayed when the beer beverage to be suggested to the customer based on the history matches the beer beverage suggested previously, and the first display field displays the previously suggested beer beverage , and the second display field displays the changed beer beverage. By inputting customer attributes and answers to multiple questions into a learning model that outputs beer information, which is a characteristic value including at least two of the following: alcohol content, bitterness value, color, and ethyl acetate, if the beer beverage, including beer, sparkling wine, or beer-flavored beverage, suggested based on the outputted beer information does not match the previously suggested beer beverage stored in the history of the beer beverage, first training data is obtained, which includes beer information about the beer beverage suggested based on the beer information and the customer's feedback on the suggested beer beverage.
If the beer beverage suggested based on the customer's attributes and answers to multiple questions, and the beer information output from the learning model, matches the previously suggested beer beverage stored in the beer beverage history, then second training data is acquired, which includes beer information about a modified beer beverage different from the previously suggested beer beverage, and the customer's feedback on the modified beer beverage.
A method for generating a learning model, wherein a computer performs a process to generate the learning model using the first training data and the second training data. Obtain customer attributes,
We obtained answers to several questions from the aforementioned customer,
A learning model that outputs beer information regarding beer, low-malt beer, new genre beer, or beer-flavored beverages should be suggested when the acquired customer attributes and multiple responses are input, outputs beer information regarding beer beverages that should be suggested when the customer attributes and multiple responses are input,
Based on the outputted beer information, the history of the beer beverages suggested is stored in association with the customer.
A character that provides guidance to customers is displayed on the screen.
If the beer beverage suggested to the customer through the learning model matches the beer beverage suggested previously based on the history, the beer beverage suggested to the customer will be changed to a different beer beverage from the one previously suggested .
An information processing method in which a computer performs a process to change the character's comments and expressions to a second comment and a second expression that are different from the first comment and first expression used when informing the customer about the beer beverage suggested to the customer through the learning model, when the beer beverage is changed to a different one.