JP7736962B1 - Restaurant information system - Google Patents

Abstract

[Problem] To enable users to easily find restaurants that suit their preferences, their dining purposes, and the circumstances they envision.
[Solution] A counselor selection screen that allows the user to select one of a plurality of different virtual counselors is displayed on the user terminal 3 by the information providing server 1, and the user selects a counselor and then inputs consultation information on the consultation information input screen. An answer request program 12 on the information providing server 1 sends a question to the artificial intelligence generating means 5 in accordance with the selected counselor and the input consultation information, and displays the returned answer on the answer screen. The answer request program 12 adds counselor additional information for causing the artificial intelligence generating means to function as the selected counselor, and restaurant additional information for causing the answer from the artificial intelligence generating means 5 to include any of the restaurants whose information is recorded in the restaurant information file 21 on the storage unit 2, to the consultation information, and then transmits the answer request.
[Selected figure] Figure 5

Description

Application of Article 30, Paragraph 2 of the Patent Act 1. Facts of Disclosure (1) Facts of Disclosure 1 ▲1▼Date of Disclosure: December 16, 2024 ▲2▼Publication Location: https://gourmet-calendar.com/ ▲3▼Discloser: Tokyo Calendar Co., Ltd. ▲4▼Contents of the Disclosed Invention: The applicant has implemented an "AI Restaurant Concierge" function on the website "Gourmet Calendar" (https://gourmet-calendar.com/) operated by the applicant, which allows users to consult AI in chat format about restaurant searches, and has launched the service. The website provides three AIs with different personalities, and explains that they can be used differently depending on the purpose and application. (2) Facts of Disclosure 2 ▲1▼Date of Disclosure: December 18, 2024 ▲2▼Publication Location: https://x. com/tokyo_calendar/status/1869312638750208417 ▲3▼Discloser Tokyo Calendar Co., Ltd. ▲4▼Disclosed Contents of the Invention A post was made on social media X (formerly Twitter) introducing the "AI Restaurant Concierge" service, which allows users to consult AI in chat format to find restaurants. The post explains that three AIs with different personalities are available, and that they can be used differently depending on the purpose and application. (3) Disclosure Facts 3 ▲1▼Disclosure Date December 19, 2024 ▲2▼Disclosure Location https://prtimes.jp/main/html/rd/p/000000072.000032791. html ▲3▼Discloser Tokyo Calendar Co., Ltd. ▲4▼Contents of the disclosed invention An article about "AI Restaurant Concierge," a service that allows users to consult AI about restaurant searches in chat format, was released on the press release site (https://prtimes.jp/) operated by PR TIMES Co., Ltd. (1-11-44 Akasaka, Minato-ku, Tokyo), and the article was posted on the site. The article explains that three AIs with different personalities are available, and that they can be used depending on the purpose and use.

The present invention relates to a restaurant information system that provides information on various restaurants.

Currently, there are several information sites on the Internet that provide information on restaurants and other food establishments, and many of these also provide intermediary services for reserving tables at restaurants. A user planning to dine at a restaurant uses their own device (such as a smartphone) to access one of several gourmet sites and use the search function to display restaurant information on the device. They then browse the list of restaurants, select one, and make a reservation by phone or using the reservation function on the site.

JP 2025-38236 A

However, it is quite difficult for users to find a restaurant that perfectly matches their tastes based solely on the information posted on information websites. User preferences vary widely, including the restaurant's atmosphere, concept, menu composition, and staff service style. Many users have often found a restaurant they thought would be good on an information website, only to find that it did not quite match their tastes. Furthermore, when a specific purpose or situation is in mind for dining, finding a restaurant that perfectly matches those purposes or situations can be just as difficult. Even if users select a candidate restaurant, they often find that when they call to make a reservation, it is actually not suitable for the purpose of the dinner, resulting in repeated calls to different restaurants.
The present invention has been made to solve the above-mentioned problems in providing restaurant information, and aims to enable users to easily find restaurants that suit their preferences, their dining purposes, and the circumstances they envision.

To solve the above problems, this specification discloses a restaurant information system. The restaurant information system according to the disclosed invention is a restaurant information system used to receive inquiries from users who are planning to dine at restaurants and provide information on suitable restaurants. This system uses artificial intelligence (AI) generation means to respond to inquiries from users.
Specifically, this system includes an information server that can be accessed via a network from a user terminal, which is a terminal operated by a user, and a storage unit.
The information providing server is connected to a generating artificial intelligence means that generates and outputs an answer from the output layer via an intermediate layer in response to information input to the input layer of the neural network, and the information providing server is a server that can display a counselor selection screen, a consultation information input screen, and an answer screen on a user terminal.
The counselor selection screen allows the user to select one counselor from among a number of different virtual counselors, and allows the user to select one counselor based on the user's preferences, the purpose of the dining experience the user intends to have at the restaurant, or the dining situation the user envisions.
The consultation information input screen is a screen that allows the user to input consultation information, which is information about the eating and drinking that the user is going to do at the restaurant.
The counselor selection screen is provided with a selection button for selecting one counselor from a plurality of different virtual counselors.
The answer screen is a screen on which an answer from the artificial intelligence generating means is displayed when the user selects one counselor using the selection button and inputs consultation information on the consultation information input screen.
The information providing server is equipped with an answer request program that sends an answer request to the generating artificial intelligence means when the user selects one counselor using a selection button and inputs consultation information on the consultation information input screen, and an answer receiving program that displays the answer on the answer screen when the answer is sent from the generating artificial intelligence means after the answer request program has sent the answer request.
The storage unit stores restaurant information files that store information on a plurality of restaurants.
The answer request program is a program that adds counselor additional information for making the generating artificial intelligence means function as a virtual counselor selected by the user using a selection button, and restaurant additional information for making the generating artificial intelligence means include any restaurant whose information is recorded in the restaurant information file in its answer, to the consultation information entered by the user, and then sends the answer request to the generating artificial intelligence means.
In order to solve the above problem, in the restaurant information providing system according to the disclosed invention, the answer request program may be a program that, when adding additional restaurant information, adds different additional restaurant information depending on the selected counselor.
In order to solve the above problem, in the restaurant information provision system of the disclosed invention, the response receiving program may be a program that displays a response screen incorporating an operation unit for displaying detailed information about the restaurant related to the name on the user terminal when the name of a restaurant whose information is recorded in the restaurant information file is included in the response from the generating artificial intelligence means.
In order to solve the above problem, in the restaurant information provision system of the disclosed invention, the information provision server can be equipped with a reservation acceptance program that allows a user to make a reservation for dining at a restaurant related to the name recorded in the restaurant information file at a specific date and time if the name of the restaurant is included in the response from the generation artificial intelligence means.

As explained below, according to the restaurant information provision system of the disclosed invention, when providing information on restaurants suitable for a user who is planning to eat at a restaurant, the information on suitable restaurants is provided by a generating artificial intelligence means, so that the user can easily find a restaurant that suits their preferences, purpose, or anticipated situation.
At this time, the answer request program adds additional restaurant information and transmits the answer request to the generating artificial intelligence means, so that the restaurant whose information is recorded in the restaurant information file is more likely to be included in the answer, and it is possible to induce the restaurant whose information is recorded in the restaurant information file to be included in the answer.
Furthermore, one of a plurality of virtual counselors is selected according to the user's preferences, purpose of dining, or anticipated situation, and the answer request program adds counselor additional information for causing the AI generating means to function as the selected counselor and sends the answer request to the AI generating means, making it easier to find a restaurant that suits the user's preferences, purpose, or anticipated situation. This is advantageous in that it does not require the huge costs that would be incurred if separate AI generating means were used for each counselor.
Furthermore, if the answer request program is configured to add different restaurant information depending on the selected counselor when adding additional restaurant information, this is even more preferable because it increases the likelihood of obtaining a restaurant that is more suitable for the user's preferences, purpose for eating and drinking, or anticipated situation.
Furthermore, if the response receiving program is configured to be a program that displays a response screen incorporating an operation unit for displaying detailed information about the restaurant related to the name on the user terminal when the name of the restaurant whose information is recorded in the restaurant information file is included in the response from the generating artificial intelligence means, it becomes easier for the user to obtain more detailed information about the response from the generating artificial intelligence means.
Furthermore, if the response from the generating artificial intelligence means includes the name of a restaurant whose information is recorded in the restaurant information file, a reservation reception program that allows users to make reservations for meals at the restaurant associated with that name on a specific date and time can be implemented on the information providing server, which can encourage more reservations at the restaurant and contribute to increased sales for the restaurant.

1 is a schematic diagram of a restaurant information providing system according to an embodiment. FIG. 1 is a schematic diagram showing an example of a top page of a gourmet site. FIG. 2 is a schematic diagram showing an example of the structure of a restaurant information file. FIG. 10 is a schematic diagram showing an example of the structure of a user information file. FIG. 10 is a schematic diagram showing an example of a counselor selection page. FIG. 10 is a schematic diagram showing an example of a consultation page displayed by a consultation page display program. FIG. 10 is a schematic diagram showing the configuration of the use of artificial intelligence generating means on the consultation page. 10 is a flowchart showing an outline of a response request program. 10 is a schematic diagram showing an example of a display state of a user terminal when a reply request program is executed. FIG. 10 is a schematic diagram showing an example of a display state of a user terminal when a reply request program is executed. FIG. FIG. 10 is a schematic diagram showing an example of a details page.

Next, a mode for carrying out the present invention (hereinafter referred to as an embodiment) will be described. Fig. 1 is a schematic diagram of a restaurant information providing system according to an embodiment.
The restaurant information providing system of the embodiment is a restaurant information providing system used to provide information on suitable restaurants to a user who is planning to eat at a restaurant. This system accepts inquiries from users and provides information on suitable restaurants, and in this case, a configuration is adopted in which a generating artificial intelligence means 5 is used.
Specifically, as shown in FIG. 1, the restaurant information system includes an information server 1 that can be accessed via a network from a user terminal 3, which is a terminal operated by a user, and a memory unit 2.

In this embodiment, the network is the Internet 9, and therefore the information providing server 1 is a server that functions as a web server. The memory unit 2 is a storage device such as a hard disk provided in the information providing server 1, but it may also be provided in a server separate from the information providing server 1, such as a storage server. Note that the information providing server 1 may not consist of only one server, and multiple servers may work together to function as the information providing server 1. The same is true for the memory unit 2, and multiple scattered memory units may function as the memory unit 2 in this embodiment.

In this embodiment, the information providing server 1 is managed by a business that provides restaurant information and mediates restaurant reservations (hereinafter referred to as the intermediary business). The intermediary business uses the information providing server 1 to provide users with a website (hereinafter referred to as the gourmet site) that provides information on various restaurants and allows them to make restaurant reservations. Therefore, the memory unit 2 stores HTML files (not shown) for each page of the gourmet site. The intermediary business operates as a business that advertises and promotes each restaurant on behalf of the restaurant, and can also be considered an advertising agency.

Figure 2 is a schematic diagram showing an example of the top page of a gourmet site. The gourmet site is designed to display information on various restaurants by narrowing down the search results by region (area) or genre. For this reason, the top page is provided with a region search page link button 61 and a genre search page link button 62. Figure 2 shows an example of the top page of the smartphone version, but the PC version is basically the same.
The area search page is a page for viewing restaurant information by narrowing down the area that restaurants are located in. Although not shown in the figure, the area search page has an area selection field and a search button, and when you select an area and press the search button, a list of information about restaurants located in the selected area is displayed.

The genre selection page is similar, with a genre selection field and a search button. By selecting a genre such as "izakaya" or "restaurant" and pressing the search button, a list of restaurants in the selected genre will be displayed. The region selection page also has a genre selection button, and the region selection page also has a region selection button, allowing you to view restaurant information by narrowing down your search by both region and genre.

A restaurant information file 21 that records information about each restaurant introduced on these pages is stored in the storage unit 2. Fig. 3 is a schematic diagram showing an example of the structure of the restaurant information file.
As shown in FIG. 3, the restaurant information file 21 is a database file consisting of fields such as "Store ID,""StoreName,""Genre,""Address,""TelephoneNumber," and "Introduction Text.""IntroductionText" is a field in which text introducing the restaurant is recorded. In addition to the fields shown in FIG. 3, there are also fields for recording information such as the official website URL, image file path, budget amount (for lunch and dinner), availability of private rooms, maximum number of people for banquets, and whether or not a reservation intermediation contract exists. The availability of a reservation intermediation contract is information on whether or not a contract for providing reservation intermediation services has been concluded for the restaurant in question.

The script embedded in the search button on the area search page is coded to search the restaurant information file in the area selected in the area selection field, retrieve each piece of information from the corresponding record, and display it on the area search page. The same is true for the genre search page, which searches the restaurant information file 21 for the selected genre, retrieves each piece of information from the corresponding record, and displays it.
The business premises of the intermediary business are provided with a management terminal 4. The management terminal 4 has special access authority and is able to access the information providing server 1 and the storage unit 2. Each piece of information in the restaurant information file 21 is information that has been input and recorded by the management terminal 4.

Users can register to use the gourmet site, and some functions, such as the reservation intermediation service, are only available to registered users. For this reason, a user information file 20 is stored in the storage unit 2. Figure 4 is a schematic diagram showing an example of the structure of a user information file. The user information file 20 is a database file that stores each user's name, age, gender, nickname, email address, as well as their user ID and password. As shown in Figure 2, the top page has a login button 60, which is linked to the login page. The login page has input fields for the user ID and password and a send button. When the user ID and password are correctly entered and the send button is pressed, the user is logged in with that user ID. Once logged in, the user ID is stored in a session variable.

As mentioned above, such a gourmet site is configured to allow restaurant information to be viewed using the artificial intelligence generating means 5. This will be described in detail below.
A major feature of the system of the embodiment is that the artificial intelligence generating means 5 is used as a virtual counselor, and in particular, multiple different virtual counselors are simulated as such counselors. "Consultation" here refers to consulting about restaurants that suit one's preferences, or restaurants that are suitable for the purpose of a planned dining trip or the anticipated dining situation. The consultation service using the artificial intelligence generating means 5 is available only to registered users, and requires that the user is logged in with a user ID.
More specifically, as shown in Fig. 2, the top page has a counselor selection page button 63 titled "Consult with AI." The counselor selection page button 63 is linked to a counselor selection page. Fig. 5 is a schematic diagram showing an example of the counselor selection page.

The counselor selection page corresponds to the counselor selection screen in this embodiment. As shown in FIG. 5, the counselor selection page has multiple counselor selection buttons 64A, 64B, and 64C. Each counselor selection button 64A, 64B, and 64C is a button for selecting a virtual counselor (concierge). In this example, three counselor selection buttons 64A, 64B, and 64C are provided, each of which is a command button incorporating an illustration of the counselor and explanatory text. For convenience of explanation, the counselors will be referred to as counselors A, B, and C below, and the counselor selection buttons will be referred to as counselor selection button 64A, counselor selection button 64B, and counselor selection button 64C. In this example, counselor A is a hypothetical counselor who is a gourmet who is knowledgeable about high-end gourmet food, counselor B is a hypothetical counselor who is knowledgeable about restaurants where couples dine on dates, and counselor C is a hypothetical counselor who is knowledgeable about restaurants suitable for business dinners and other occasions.

A consultation page display program 11 that displays a consultation page is embedded in each of the counselor selection buttons 64A, 64B, and 64C. The counselor ID of the counselor being displayed is embedded in each of the counselor selection buttons 64A, 64B, and 64C, and the consultation page display program is executed using the counselor ID as an argument.
The consultation page display program first obtains the user ID from the session variable, searches the user information file 20 using the user ID, and records the counselor ID argument in the "Counselor ID" field of the corresponding record. The consultation page display program 11 then displays the consultation page. Figure 6 is a schematic diagram showing an example of a consultation page displayed by the consultation page display program.

Figure 7 is a schematic diagram showing the configuration of the generative artificial intelligence means used on the consultation page. As shown in Figure 7, the generative artificial intelligence means 5 includes an AI server 51 and a large-scale database 52 that supports the deep learning performed by the AI server 51. The neural network 53 in the AI server 51 is composed of an input layer 531 to which information is input, an output layer 532 to which information generated in response to the input information is output, and an intermediate layer 533 located between the input layer 531 and the output layer 532. The intermediate layer 533 is multi-layered, and the neural network 53 is a multi-layer perceptron. The neural network 53 here refers to a group of programs that perform large-scale processing (large-scale language processing in this embodiment) according to a mathematical model that mimics the neural network in the brain. In the following description, the information 501 input to the input layer 531 of the neural network 53 is referred to as a question prompt, and the answer 502 output from the output layer 532 is referred to as an answer prompt.

As shown in Figure 6, the consultation page has a consultation information input field 65 and a response display field 68. Therefore, the consultation page in this embodiment is a page that serves as both a consultation information input screen and a response screen. The consultation information input field 65 is located at the bottom of the consultation page and is a field for entering text.

In this embodiment, restaurant information is provided in chat format using the artificial intelligence generating means 5. For this reason, the consultation page includes a chat window in which a counselor comment display field and a user comment display field are alternately displayed. In the chat window shown in FIG. 6, the counselor comment display field 66 displayed at the top is called the counselor initial comment display field, and the user comment display field 67 displayed below it is called the user initial comment display field. The counselor comment display field immediately below the user initial comment display field 67 (the second counselor comment display field) is a field in which responses from the artificial intelligence generating means 5 are displayed, and corresponds to the response display field 68 in the embodiment.

Figure 6 shows the consultation page in its initial state (hereinafter referred to as the consultation initial page). In the example of Figure 6, counselor A has been selected. The consultation page display program 11 is coded to initially display the counselor's comments in the counselor initial comment display field 66 and leave the user initial comment display field 67 blank. In this case, the consultation page display program 11 is coded to display the consultation initial page according to the counselor ID embedded in the counselor selection buttons 64A, 64B, and 64C.

That is, the storage unit 2 stores an image (avatar) representing each counselor and text to be displayed in the counselor initial comment display field 66. The consultation page display program 11 is coded to obtain the counselor avatar and text specified by arguments from the storage unit 2 and incorporate them into the consultation initial page.
6, a send button 69 is provided next to the consultation information input field 65. The information providing server 1 is equipped with an answer request program 12, and the send button 69 is an execution button for the answer request program 12. The information providing server 1 is also equipped with an answer receiving program 13, and the answer from the artificial intelligence generation means 5 is displayed in an answer display field 68 on the consultation page by the answer receiving program 13.

Specifically, in this embodiment, ChatGPT is used as the generating artificial intelligence means 5. Therefore, the AI server 51 is a ChatGPT server, and the information providing server 1 is equipped with the ChatGPT API. The information providing server 1 is configured to use ChatGPT in a web browser (consultation page), and the answer request program 12 embedded in the send button 69 is coded to create a consultation prompt based on the text entered in the consultation information input field 65 and input it as an HTTP request to the input layer 531 of the AI server 51 using the ChatGPT API function. In addition, the answer receiving program 13 is coded to receive the answer prompt output from the output layer 532 of the AI server 51 and display it in the active counselor comment display field on the consultation page.

The answer request program 12 is coded to send the text entered in the consultation information input field 65 to the AI server 51 not simply as a consultation prompt, but also together with additional information, which is a major feature of the restaurant information provision system of this embodiment.
More specifically, the answer request program 12 is coded to send an HTTP request by adding additional information to the text entered in the consultation information input field 65. The answer request program 12 employs the RAG (Retrieval-Augmented Generation) method, and in this embodiment, the OpenAI Embedding model is used to vectorize the question (the text entered in the consultation information input field 65) and each restaurant information (information such as each introduction text in the restaurant information file 21), and a cosine similarity calculation (calculation of the distance between vectors) is used to extract information about restaurants that are similar to the question.

In the restaurant information provision system of this embodiment, two types of information are added as additional information. One is additional information for guiding the answer from the artificial intelligence generating means 5. For convenience of explanation, this additional information is referred to as restaurant additional information. The other is information that enables the artificial intelligence generating means 5 to simulate a selected counselor. Hereinafter, this additional information will be referred to as counselor additional information. The answer request program 12 is coded to add the restaurant additional information and counselor additional information, generate a consultation prompt, and send it to the AI server 51.

The above points will be explained in more detail with reference to Figure 8. Figure 8 is a flowchart showing an overview of the answer request program 12. The answer request program 12 is executed using the text entered in the consultation information input field 65 as an argument. For the sake of explanation, the text entered in the consultation information input field 65 will be referred to as user-input information. As shown in Figure 8, the answer request program 12 first displays the user-input information in the user initial comment display field 67. It then performs processing to generate additional restaurant information. In this embodiment, the additional restaurant information is information on multiple restaurants that are highly correlated with the user-input information, from the restaurant information recorded in the restaurant information file 21.

More specifically, the information providing server 1 is equipped with a candidate store acquisition program 13 as a subroutine of the response request program 12. The candidate store acquisition program 13 is a program that acquires the names of multiple highly correlated restaurants from the restaurant information file 21. There are several possible codes for acquiring the names of highly correlated restaurants, but one example is a configuration that performs language processing for vectorization. That is, the value (text) of the "introduction text" field of each record in the restaurant information file 21 is embedded (vectorized) in advance and recorded in the database file 22 together with the store ID. Hereinafter, this file 22 will be referred to as the comparison file. Note that the text may also be divided into text chunks before being vectorized.

The candidate store acquisition program 13 vectorizes the text entered in the consultation information input field 65 and stores it as a vector data file 23. In the above example, this process can be performed using the ChatGPT API functions. The program then compares the vector data for each restaurant in the comparison file 22 with that in the vector data file 23 for the text acquired from the consultation information input field 65, and acquires the store ID of any record where the difference is within a certain threshold. The program then acquires each store name and each introduction text according to the acquired store ID, records them in a simple database file 24 such as a CSV, and uses them as additional restaurant information. Hereinafter, this simple database file 24 will be referred to as the restaurant additional information file.

As shown schematically in Figure 8, the answer request program 12 is coded to display an error and terminate the program if a return value is not returned within the time set by the timer. On the other hand, as shown in Figure 8, if a return value (additional restaurant information) is received from the candidate store acquisition program 13, the answer request program 12 performs processing to add additional counselor information. There are various possible configurations for adding additional counselor information, but one possible configuration is to set text information as a different constant for each counselor and add this as additional counselor information.

To explain the above in more detail as an example, one possible configuration is to prepare constant text for each counselor specifying their personality, speaking style, and preference for favorite restaurants. For example, for counselor A, the following text could be set as a constant: "Please answer assuming that you have a gentle personality, speak in a soft tone, are knowledgeable about high-end gourmet food, and tend to recommend Japanese restaurants." For counselor B, the following text could be set as a constant: "Please answer assuming that you have a passionate personality, speak in a casual tone, are knowledgeable about restaurants where couples go for dates, and tend to recommend Western restaurants." For counselor C, the following text could be set as a constant: "Please answer assuming that you have a meticulous personality, speak in a polite tone, and tend to recommend restaurants suitable for entertaining."

It is preferable to use Markdown notation when adding additional counselor information to the question prompt. For example, in the above example,
# Role: A gentle-natured expert on high-end gourmet food who tends to recommend Japanese restaurants.
# How to answer Please answer calmly and politely.
It will look something like this.

As shown in Figure 8, the answer request program 12 acquires the restaurant additional information, then generates a question prompt 501 by adding a pseudo-constant to the user-input information according to the counselor ID. The counselor ID is acquired by acquiring the user ID from the session variable, searching the user information file 20 using the user ID, and obtaining the corresponding record. The answer request program 12 is then coded to send the generated question prompt 501, along with the restaurant additional information file 24, to the input layer 511 of the AI server 51.

The answer receiving program 13 is coded to incorporate and display answers from the generating artificial intelligence means 5 obtained via an API (ChatGPT API in the above example) in the counselor comment display field (answer display field 68) on the consultation page. Figures 9 and 10 are schematic diagrams showing examples of display states on the user terminal following execution of the answer request program. Of these, Figure 9 shows the state immediately after execution of the answer request program 12, and Figure 10 shows the state after the answer is displayed by the answer receiving program 13.
As shown in Figure 8, the answer receiving program 13 is a program that is called and executed by the answer request program 12 when an answer is received from the artificial intelligence generating means 5. However, it is also possible to provide a separate program that detects the receipt of an answer from the artificial intelligence generating means 5 as an event, and execute the program from there. As shown schematically in Figure 8, if an answer is not received within the time set by the timer, the answer request program 12 is coded to display an error message and then terminate the program.

When an answer is received, the answer receiving program 13 performs language processing on the text of the answer prompt 502 and identifies the location where the restaurant name is displayed. If the restaurant name is recorded in the restaurant information file 21, the program is coded to obtain the minimum amount of information from the restaurant information file 21 and display it simply on the consultation page. Specifically, as shown in Figure 10, the answer is displayed in the answer display field 68, and a frame 70 is inserted below it, and the restaurant name, image, etc. are incorporated and displayed in this frame (hereinafter referred to as the simple introduction frame) 70.

The reply receiving program 13 is coded to embed a link to a details page that displays detailed information about the restaurant when inserting the simple introduction frame 70. Fig. 11 is a schematic diagram showing an example of a details page.
As shown in FIG. 11, the details page displays detailed information about the restaurant. This information is obtained from the record in question in the restaurant information file 21. At this time, if the presence or absence of a reservation intermediary contract is a true value, elements related to the reservation are also incorporated and displayed on the details page. The example in FIG. 11 is an example where the presence or absence of a reservation intermediary contract is a true value. That is, as shown in FIG. 11, the details page has a reservation button 71. The reservation button 71 is linked to the reservation page.

Details of the reservation page are omitted, but it includes a calendar control, with available reservation dates displayed with a circle. A reservation information file 25 is stored in the memory unit 2, and the reservation acceptance program 14 is implemented in the information providing server 1. Pressing the circle in the calendar control displays the reservation details page, allowing you to make a reservation. When you select the time and number of people on the reservation details page and press the send button, the reservation acceptance program 14 records the reservation information in the reservation information file 14. These configurations can be configured in the same way as regular restaurant reservation sites, so further explanation will be omitted. When making a reservation, information about the person making the reservation is obtained from the user information file 26 according to the user ID obtained when logging in, and recorded in the reservation information file 25.

On the consultation page, the top blank user comment display column is the active column. Therefore, when the answer to the first question prompt 501 is displayed in the answer display column 68 and the user further enters a question in the consultation information input column 65 and presses the send button 69, the question is displayed in the active user comment display column immediately below. At the same time, the answer request program 12 creates a further question prompt 501 including the Q&A up to that point and sends it to the generating artificial intelligence means 5. When the answer prompt 502 is returned, the answer is displayed in the active counselor comment display column (answer display column 68) immediately below the user comment display column that displayed the question. The answer request program 12 and answer display program 13 are coded to allow chat to take place in this manner.

The operation of the restaurant information providing system of this embodiment will be described below.
The intermediary company operating the restaurant information system conducts marketing activities with each restaurant and invites them to register on its gourmet website. In response, the restaurant owners provide their establishment information to the intermediary company. The intermediary company's staff then operates the management terminal 4 to record the restaurant information in the restaurant information file 21. The information on the numerous restaurants collected in this manner is recorded in the restaurant information file 21, and vector data generated from each introduction text is recorded in the comparison file 22. While a registration fee (advertising fee) such as a monthly fee is often charged for registration on a gourmet website, the website's value increases with the amount of information posted, so registrations may be offered free of charge. Reservation intermediary services may be charged a flat monthly fee, pay as you go, or even free of charge. If a reservation intermediary service contract has been signed, this information is also recorded in the restaurant information file 21.

On the other hand, if a user wants to eat out at a restaurant but can't decide which one to go to, they operate the user terminal 3 to access the gourmet site, enter their ID and password to log in, and the top page is displayed on the user terminal 3. They then find it interesting that the AI is recommending restaurants, and press the counselor selection page button 63. From the multiple counselor selection buttons 64A, 64B, 64C displayed on the user terminal 3, they select and press one counselor selection button 64A, 64B, 64C according to their preferences, the purpose of their planned dining out, or the expected situation. This causes a consultation page like the one shown in Figure 6 to be displayed on the user terminal 3.

The user enters appropriate text in the consultation information input field 65 displayed at the bottom of the consultation page and presses the send button 69. This activates the answer request program 12, and the user-entered information is displayed in the user's initial comment display field 67 on the user terminal 3, as shown in Figure 9. The answer request program 12 compares the vector data (data in the vector data file 23) obtained by vectorizing the text of the user-entered information with the vector data of each restaurant in the comparison file 22. It then obtains multiple store names and introduction texts of multiple restaurants that are determined to have a high correlation, and generates a CSV-like restaurant additional information file 24.

The answer request program 12 then adds the text of the counselor additional information to the text entered in the consultation information input field 65 (user-input information) to generate a question prompt 501, which it then sends to the artificial intelligence generation means 5 along with the restaurant additional information file 24 and inputs into the input layer 511 of the AI server 51. The AI server 51 performs large-scale language processing (LLM) while referencing the large-scale database 52 along with the restaurant additional information file 24, generates an answer prompt 502 through deep learning, and outputs it from the output layer 532. The text of the output answer prompt 502 is incorporated into the answer display field 68 by the answer receiving program 13 and displayed. If the answer prompt contains a store name that is the same as a store name recorded in one of the records in the restaurant information file 21, the answer receiving program 13 obtains an image file or the like of the store, inserts a simple introduction frame 70, and displays the image file therein.

The user references the answers displayed in the answer display field 68 and the information in the simple introduction frame 70. If the user is interested in the restaurant introduced in the answer, they click the simple introduction frame 70 to display a details page on the user terminal 3. If the user decides to eat at the restaurant, they can make a reservation if necessary. That is, if a reservation intermediary service contract has been made, the reservation button 71 will be displayed, and the user can click the reservation button 71 to display the reservation page on the user terminal 3. Then, after selecting an available date using the calendar control, the user selects an available time in the time selection field. The user then clicks the send button to confirm the reservation. The reservation information is recorded in the reservation information file 25 by the reservation acceptance program 14.

In this embodiment of the restaurant information provision system, when providing information on restaurants suitable for a user planning to dine at a restaurant, the artificial intelligence generating means 5 provides the information on suitable restaurants, allowing the user to quickly obtain information on suitable restaurants. Compared to conventional methods where users had to narrow down their search by genre or region and then consider each restaurant by viewing its individual introduction page, this system allows users to obtain information on suitable restaurants more quickly and easily.

In this case, according to the system of the embodiment, one artificial intelligence means 5 is used to simulate multiple different counselors. Therefore, by selecting an appropriate counselor based on the user's preferences, intended dining purpose, or anticipated situation, the user can easily obtain information about restaurants that suit their preferences. A possible configuration for using artificial intelligence means to simulate counselors is to use a separate artificial intelligence means for each counselor. Specifically, each AI server is provided, and large amounts of data are input to each server for deep learning. In this case, a configuration is conceivable in which different data is input depending on the simulated counselor, resulting in different answers to the same question. However, this configuration has the disadvantage of requiring significant server and deep learning costs. In the system of the embodiment, one artificial intelligence means 5 is simulated as multiple different counselors by devising information added to user-input information, and the user is then allowed to select one counselor from among them. This makes it possible to provide users with a service that allows them to easily find restaurants that suit their preferences, dining purpose, or anticipated situation without incurring significant costs.

Although not explained here, the text of the user-input information is not directly converted into question prompt 501; in fact, in many cases, some pre-processing is performed before sending it to the artificial intelligence generating means 5 as question prompt 501. For example, the text "Please check the following sentence for typos and omissions" is added to the text of the user-input information, and the text is sent to the artificial intelligence generating means 5 to check for typos and omissions, and the returned answer prompt is then used as question prompt 501. In addition, the text of the user-input information may be summarized for later processing, and may be passed through the artificial intelligence generating means 5 for summarization.

In addition, when generating additional restaurant information, pre-processing can be performed that takes into account regional characteristics and budget. Specifically, the text of the user-entered information is analyzed to extract information on place names and budgets, and the restaurant information file 21 is searched and narrowed down based on the extracted information. The vector values of the introduction text of each of the narrowed-down number of restaurants are then compared with the vector values of the text of the user-entered information to obtain the restaurant names and introduction texts of restaurants with a high correlation above a certain level, and these are used as additional restaurant information. The reason for this is that the artificial intelligence generating means 5 is not good at narrowing down the search by region or budget, and pre-processing (pre-narrowing) makes it easier to obtain more accurate answers. The artificial intelligence generating means 5 is often used to extract information. Alternatively, separate fields for entering the place names of restaurants and the budgets may be provided, and the restaurant information file 21 may be searched based on the information entered there.

As mentioned above, the user information file 20 records and saves the counselor ID of the counselor selected by each user for their most recent consultation. This is particularly significant from the perspective of user analysis as part of marketing. That is, the user information file 20 records the age, gender, etc. of each user, and the counselor ID information can be used to analyze counselor selection. For example, it is possible to analyze what type of counselor is likely to be selected depending on age, gender, etc.
In the above example, the number of simulated counselors was set to three, but the number may be four, five, or even more. In this case, by using one generating AI means 5 in different ways, a system that can respond more precisely to preferences, dining purposes, and situations can be provided at low cost.
The above input by the user and the response from the generating artificial intelligence means 5 may be repeated several times, and a configuration may be adopted in which an ID is assigned to this exchange (chat) and recorded. That is, a chat database file is stored in the memory unit 2, and a chat ID is assigned to each chat, and the user ID of the user who conducted the chat, the counselor ID of the selected counselor, etc. are recorded.

In addition, in the above embodiment, the answer receiving program 13 displays the answer screen with an operation unit incorporated for displaying detailed information about the restaurant related to the name on the user terminal 3 when the answer from the artificial intelligence generating means 5 includes the name of a restaurant whose information is recorded in the restaurant information file 21. This makes it easier for the user to obtain more detailed information about the answer from the artificial intelligence generating means 5. In the above configuration, the operation unit is a simple introduction frame 70 with an embedded link, but various other configurations can be adopted. For example, a hyperlink can be embedded in the store name, or a configuration in which a details page is displayed in a pop-up window when the pointer is moved over the store name can also be used. "Operation unit" means the part where a details page is displayed when the user performs some operation.

Furthermore, if the response from the artificial intelligence generation means 5 includes the name of a restaurant whose information is recorded in the restaurant information file 21, the information providing server 1 is equipped with a reservation acceptance program 14 that allows users to make reservations for dining at the restaurant associated with that name for a specific date and time, which can encourage more reservations at the restaurant and contribute to increased sales for the restaurant. This also contributes to increased sales for the intermediary business in the embodiment, if the intermediary business has concluded a contract with the restaurant under which it receives a success fee for securing a reservation.

In the above embodiment, when generating additional restaurant information, it is also possible to take into account the perspective of additional counselor information. Specifically, when vectorizing the text of the user-input information to generate vector data, different text is added as a constant depending on the selected counselor. In the example above, if counselor A is selected, text such as "luxury gourmet" and "gourmet food" is added to the text of the user-input information; if counselor B is selected, text such as "couple" and "date" is added; and if counselor C is selected, text such as "business" and "entertainment" is added. In this way, highly correlated restaurants are extracted from the restaurant information file 21 according to the counselor's attributes. This further increases the likelihood of obtaining restaurant information that is more suitable for the purpose or situation of dining, making this a preferred configuration.

Even if the additional text is changed so that different restaurant additional information is added depending on the selected counselor, the resulting additional restaurant information may be the same. Depending on the text of the consultation information entered by the user and the information in the introduction text recorded in the restaurant information file 21, the restaurant additional information file 24 may end up with the same content even if different text is added depending on the counselor. Even in such cases, the answer request program 12 contains code intended to result in different content, and is therefore a program that adds different restaurant additional information depending on the counselor.

Furthermore, the additional information may not be a CSV file, but may simply be a context variable listing restaurant names. However, since there may be cases where the artificial intelligence generating means 5 responds with information about a restaurant with the same restaurant name that is not recorded in the restaurant information file 21, a CSV file that also includes the introduction text is more preferable as additional restaurant information. Alternatively, a request to reference the URL of each restaurant's individual introduction page on the gourmet site may be included in the question prompt, or a PDF file containing the introduction text for each restaurant in the restaurant information file 21 may be created and attached as additional information for reference.

In the above embodiment, the artificial intelligence generating means 6 was used via a website (gourmet site), but this is not required for implementing the present invention. Alternatively, the artificial intelligence generating means 5 can be accessed from an application program (app) installed on the user terminal 3. In this configuration, the consultation page in the above embodiment becomes a consultation screen provided by the app. Since the restaurant information file 21 is often stored in the memory 2 on the information providing server 1 or in another memory on the Internet, the answer request program 12 and answer receiving program 13 are often implemented on the information providing server 1 and requested to be executed by the app. However, the restaurant information file 21 and each program 12, 13 may also be installed on the user terminal 3, and each program 12, 13 may be executed on the user terminal 3. Since a web page is also a screen on the user terminal 3, an "XX page" is one form of an "XX screen."

In the above embodiment, the AI server 51 constituting the artificial intelligence generating means 5 was a separate server from the information providing server 1, but they may be a single server. In other words, a server that can function as an AI server may also be the information providing server. In such a case, it can also be said that the information providing server is connected to the artificial intelligence generating means.

1 Information providing server 11 Consultation page display program 12 Answer request program 13 Answer receiving program 2 Storage unit 21 Restaurant information file 22 Comparison file 23 Vector data file 24 Restaurant additional information file 3 User terminal 4 Management terminal 5 Generating artificial intelligence means 51 AI server 511 Input layer 512 Intermediate layer 513 Output layer 52 Large-scale database 9 Internet

Claims (4)

A restaurant information providing system used to receive inquiries from users who are planning to eat at restaurants and provide information on suitable restaurants,
The information providing server is capable of receiving access via a network from a user terminal, which is a terminal operated by a user, and includes a storage unit.
The information providing server is connected to a generating artificial intelligence means that generates and outputs an answer from an output layer via an intermediate layer in response to information input to an input layer of the neural network,
The information providing server is a server capable of displaying a counselor selection screen, a consultation information input screen, and a response screen on a user terminal,
The counselor selection screen is a screen that allows the user to select one counselor from a plurality of different virtual counselors, and allows the user to select one counselor based on the user's preference for restaurants, the purpose of the dining experience the user intends to have at the restaurant, or the dining experience the user envisions,
The consultation information input screen is a screen that allows the user to input consultation information, which is information about the eating and drinking that the user is going to do at the restaurant.
The counselor selection screen is provided with a selection button for selecting one counselor from a plurality of different virtual counselors,
The answer screen is a screen on which an answer from the generation artificial intelligence means is displayed when the user selects one counselor using the selection button and inputs consultation information on the consultation information input screen,
The information providing server is equipped with an answer request program that transmits an answer request to the generating artificial intelligence means when the user selects one counselor by pressing a selection button and inputs consultation information on the consultation information input screen, and an answer receiving program that displays the answer on the answer screen when an answer is transmitted from the generating artificial intelligence means after the answer request program transmits the answer request,
The storage unit stores a restaurant information file that stores information about a plurality of restaurants.
The answer request program is a program that adds counselor additional information for making the generating artificial intelligence means function as a virtual counselor selected by the user using a selection button, and restaurant additional information for making any of the restaurants whose information is recorded in the restaurant information file included in the answer from the generating artificial intelligence means, to the consultation information entered by the user, and then sends the answer request to the generating artificial intelligence means. The restaurant information provision system described in claim 1, characterized in that the response request program is a program that, when adding the additional restaurant information, adds different additional restaurant information depending on the selected counselor. The restaurant information provision system described in claim 1, characterized in that the response receiving program is a program that, if the response from the generation artificial intelligence means includes a name of a restaurant whose information is recorded in the restaurant information file, displays the response screen with an operation unit incorporated for displaying detailed information about the restaurant related to the name on the user terminal. The restaurant information provision system described in claim 1, 2, or 3, characterized in that the information provision server is equipped with a reservation acceptance program that allows a user to make a reservation for dining at a restaurant associated with a name recorded in the restaurant information file for a specific date and time if the response from the generation artificial intelligence means includes the name of the restaurant.