relay device

Abstract

This allows for effective questioning of AI chatbots, efficient acquisition of usable answers, and control over access to AI chatbots. [Solution] The relay device 1 stores the attribute information of the PC3 user in the attribute table 106. Through the question reception unit 111, it receives question texts from PCs connected to LAN2 to the generation AI provision server device 5. The answer request formation unit 114 takes the table data in the attribute table 106 into consideration to form answer requests for each attribute information to obtain answers to the received question texts and provides them to the generation AI provision server device 5. The answer acquisition unit 115 acquires answers for each attribute information from the generation AI provision server device 5, stores them in the answer storage unit 107, and sends back answers to the user who sent the question text, corresponding to the user's attributes.

Description

Technical Field

[0001] This invention relates to a relay device that relays communication between a terminal device connected to a LAN (Local Area Network), such as a proxy server, and a server device provided on the Internet.

Background Art

[0002] The generative AI (Artificial Intelligence) technology that can generate various contents has advanced rapidly, and services using artificial intelligence chatbots utilizing this technology have become available. An artificial intelligence chatbot is a program provided in a server device on the Internet. Its characteristic function is that an AI self-learned based on various available information accepts questions from a user's terminal device in the form of text data or voice data, generates an answer to the question, and can return this to the user's terminal device. Regarding artificial intelligence chatbots, various improvements have been made. For example, Patent Document 1 described later discloses an invention related to a dialogue processing device and a dialogue processing program that, when an answer to a question cannot be found, further prompts the question so that dialogue can be carried out more efficiently.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Artificial intelligence (AI) chatbots can instantly answer both simple inquiries and specialized questions. Furthermore, by adding detailed information to the questions asked by the AI ​​chatbot, or by asking further questions in response to the chatbot's answers, users can obtain the information they need (information that is more relevant to their purpose). Therefore, the number of characters and the number of questions asked to the AI ​​chatbot tend to increase. If the AI ​​chatbot uses a usage-based billing system, where the fee changes depending on the amount of service used, the cost of using the AI ​​chatbot may increase. Specifically, the fee may change depending on the number of characters in the question asked by the user and the number of characters in the AI ​​chatbot's response.

[0005] Traditionally, so-called proxy servers have been designed to temporarily store content sent from a web server and provide that content to the client without accessing the web server when the same content is requested again. Therefore, it is conceivable to use this technology to accumulate and reuse responses from artificial intelligence chatbots. However, responses from artificial intelligence chatbots often vary depending on the attribute information of the questioner, such as gender and age. For this reason, even if responses from artificial intelligence chatbots are accumulated, if the questioner changes, their attribute information also changes, so it is not always possible to reuse the accumulated responses from the artificial intelligence chatbots.

[0006] In view of the above, the purpose of this invention is to enable effective questioning of an artificial intelligence chatbot, efficient acquisition of usable answers, and appropriate control over access to the artificial intelligence chatbot. [Means for solving the problem]

[0007] To solve the above problems, the relay device of the invention described in claim 1 is A relay device installed between a LAN (Local Area Network) and the Internet, which relays communication between multiple terminal devices connected to the LAN and a generation AI (Artificial Intelligence) providing server device connected to the Internet. An attribute storage means for storing and maintaining attribute information of users of multiple terminal devices connected to the LAN, A question receiving means that receives a question sentence from the terminal device to the AI ​​generation server device input by the user, A response request forming means for forming a response request to obtain a response to the question received by the question receiving means for each attribute information stored in the attribute storage means, and providing this response request to the generation AI providing server device, A response acquisition means that acquires responses for each attribute information from the AI ​​generating server device in response to the response request and records them in a response storage means, The acquired response providing means, when an answer is obtained from the generating AI providing server device through the answer acquisition means, identifies the user's attribute information based on the user information provided along with the question text received through the question receiving means, and sends back the answer associated with the identified user's attribute information from the answer acquired by the answer acquisition means to the terminal device used by the user. It is characterized by having the following features.

[0008] The relay device of the invention described in claim 1 is provided between a LAN and the internet and relays communication between a terminal device connected to the LAN and a generation AI providing server device on the internet. The relay device stores attribute information of the terminal device user in an attribute storage means. It receives question texts from terminal devices connected to the LAN to the generation AI providing server device through a question receiving means. The answer request formation means forms an answer request to obtain an answer for each attribute information for the received question text and provides it to the generation AI providing server. The answer acquisition means acquires the answer for each attribute information from the generation AI providing server, stores it in an answer storage means, and returns the answer according to the user's attributes to the user who sent the question text. [Effects of the Invention]

[0009] This invention allows for the effective creation of usable answers by prompting questions to an artificial intelligence chatbot. This helps to curb the excessive use of artificial intelligence chatbots and enables their effective utilization. [Brief explanation of the drawing]

[0010] [Figure 1] This is a diagram illustrating an example of the configuration of the network system according to the embodiment. [Figure 2] This is a block diagram illustrating an example configuration of the relay device according to the embodiment. [Figure 3] This is a diagram illustrating an example of data stored in the employee database of the relay device according to the embodiment. [Figure 4] This diagram illustrates an example of data stored in the terminal management database of the relay device according to the embodiment. [Figure 5] This figure illustrates an example of the data stored in the attribute table of the relay device according to the embodiment. [Figure 6] This is a diagram illustrating an example of data stored in the response storage unit of the relay device according to the embodiment. [Figure 7] This is a flowchart illustrating the processing performed by the relay device of the embodiment. [Modes for carrying out the invention]

[0011] Hereinafter, an embodiment of the relay device according to this invention will be described with reference to the figures. First, an example of the configuration of a network system using the relay device 1 according to this invention will be described.

[0012] [Example of a network system configuration] Figure 1 is a diagram illustrating an example of the network system configuration of the embodiment. As shown in Figure 1, the relay device 1 is installed between the Internet 4 and the LAN 2. The Internet 4 is a computer network that enables interconnection between various terminal devices using so-called Internet Protocol Suite technology. Various web servers are provided on the Internet 4, but in this embodiment, for the sake of simplicity, we show the case where the generation AI provision server device 5, which is an important access destination, is connected.

[0013] The AI ​​generation server device 5 provides an artificial intelligence chatbot service. As mentioned above, the artificial intelligence chatbot service uses an AI that has self-learned based on various available information to receive questions from the user's terminal device, generate answers to those questions, and send them back to the user's terminal device. Questions and answers are provided as text data or voice data.

[0014] LAN2 is a network formed within a limited and narrow range such as within a home or various facilities, enabling communication between terminals such as PCs (Personal Computers) connected to it. In this embodiment, LAN2 is a relatively large-scale one formed in a company's office. As shown in FIG. 1, PCs 3(1), 3(2), 3(3),... as terminal devices are connected to LAN2. Each of PCs 3(1), 3(2), 3(3),... is used by the company's employees in their work where LAN2 is formed. In FIG. 1, only three PCs, 3(1), 3(2), and 3(3), are shown connected to LAN2, but in reality, a large number of PCs are connected according to the number of employees.

[0015] Each of the PCs 3(1), 3(2), 3(3),... connected to LAN2 can be connected to the Internet 4 through the relay function of the relay device 1, enabling communication with various Web servers provided on the Internet 4. The relay device 1 of this embodiment effectively asks questions and obtains efficiently usable answers in consideration of the attribute information of the users of the PCs connected to LAN2 when using the generative AI-providing server device 5 through the PCs 3(1), 3(2), 3(3),... connected to LAN2.

[0016] To explain with a simple specific example, for instance, assume that a user of PC 3(1) who is male sends a question text to the generative AI-providing server device 5. The relay device 1 that receives the question text forms the question text so as to obtain both the answer when the questioner is male and the answer when the questioner is female simultaneously, and sends the question text to the generative AI-providing server device 5. Thereby, the relay device 1 holds the answer text returned from the generative AI-providing server device 5 and provides the answer when the questioner is male to PC 3(1).

[0017] After that, assume that a female user of PC3(2) sends a similar question text to the generative AI providing server device 5. The relay device 1 that has received the similar question text provides, to PC3(2), an answer to the question text held by itself without transmitting the question text to the generative AI providing server device 5, where the answer is for the case where the questioner is female. As a result, the question text from PC3(2) can be provided with the answer already acquired by the relay device 1 without being transmitted to the generative AI providing server device 5.

[0018] In this way, the relay device 1 of this embodiment can effectively send questions to the generative AI providing server device 5 (artificial intelligence chatbot) and obtain efficiently usable answers. By using the answers obtained in this way, overuse of the generative AI providing server device 5 can be suppressed, and the generative AI providing server device 5 can be used effectively and efficiently. Hereinafter, the configuration and operation of the relay device 1 of this embodiment will be described more specifically.

[0019] In FIG. 1, PC3(1), PC3(2), PC3(3),... have the same functions, and thus, unless there is a need to show them separately, they are collectively referred to as PC3. Also, since PC3(1), PC3(2), PC3(3),... can be used as they are as commercially available products, the description of their configuration examples is omitted. Similarly, since the generative AI providing server device 5 is already provided on the Internet 4 and provides services, the description of its configuration example is also omitted.

[0020] [Configuration Example of Relay Device 1] Figure 2 is a block diagram illustrating an example configuration of the relay device 1 of the embodiment. The connection terminal 101T constitutes the connection terminal to the Internet 4. The WAN (Wide Area Network) I / F 101 is the part that performs communication processing via the Internet 4. That is, the WAN I / F 101 receives signals addressed to itself that are transmitted via the Internet 4, converting them into a signal format that can be processed by the device. The WAN I / F 101 also converts signals to be transmitted from the device to the target destination into a signal format for transmission, sends it to the Internet 4, and transmits it to the destination.

[0021] The control unit 102, although not shown, is a microprocessor equipped with a CPU (Central Processing Unit), ROM (Read Only Memory), RAM (Random Access Memory), non-volatile memory, etc., and controls each part of the relay device 1. The storage device 103 is a device consisting of a recording medium and its driver, such as an HDD (hard disk drive) or SSD (solid state drive), and performs various operations such as recording, modifying, deleting, and reading data. In addition to storing necessary data and programs, the storage device 103 is also used as a work area to temporarily store intermediate data generated in various processes.

[0022] The employee database 104, terminal management database 105, attribute table 106, and response storage unit 107 are formed on the recording medium of the device unit, which consists of a recording medium and its driver, such as an HDD or SSD, and store and retain the data used for processing. Note that "DB" in employee database 104, etc., is used as an abbreviation for database (data base). Furthermore, the employee database 104, terminal management database 105, attribute table 106, and response storage unit 107 can be formed on different recording media, or they can be formed on the same recording medium with different storage areas. In addition, the employee database 104 and terminal management database 105 can be formed on recording medium A, and the attribute table 106 and response storage unit 107 can be formed on recording medium B, for example, by distributing them across different recording media.

[0023] The employee database 104 stores information about each employee who belongs to a company where LAN2 is formed and uses a PC 3 connected to LAN2. Figure 3 is a diagram illustrating an example of the data stored in the employee database 104 of the relay device 1. As shown in Figure 3, the employee database 104 stores information about each employee of the company, such as employee ID, name, job title, work location, gender, age, department, position, date of joining the company, current address, and other information. In this way, the employee database 104 stores information about each employee, and can be referenced as needed to understand each employee's attribute information. The employee ID is, for example, a unique identification piece of information assigned to each employee upon joining the company.

[0024] The terminal management DB 105 stores information that allows for the management of which employee is using which terminal device, such as PC3, connected to LAN2. Figure 4 is a diagram illustrating an example of the data stored in the terminal management DB 105 of the relay device 1. As shown in Figure 4, the terminal management DB 105 associates identification information that can uniquely identify each terminal connected to LAN2, such as MAC (Media Access Control) address and IP (Internet Protocol) address, with information to identify employees, such as employee ID and name. This makes it possible to identify which employee is using PC3, which is identified by its MAC address or IP address.

[0025] Note that it is not necessary to have both a MAC address and an IP address. A configuration with only one of them is also possible. However, since the IP address may change, it is necessary to ensure that PC3 can always be properly identified, such as by receiving notification from PC3 of the change and updating the IP address accordingly.

[0026] The attribute table 106 stores attribute information of employees who are users of PC3 connected to LAN2, and stores a list of attribute information selected to be used for questioning the generation AI provision server device 5, as well as a list of the specific contents of each attribute information. Figure 5 is a diagram illustrating an example of the data stored in the attribute table 106 of the relay device 1. The example shown in Figure 5(A) shows that the attribute information selected from the attribute information of individual employees to be used for questioning the generation AI provision server device 5 is job title, work location, and gender. Specifically, the attribute information shows that the job title is engineer and sales, the work location is Yokohama, Tokyo, and Yonezawa, and the gender is male, female, and no answer (no particular distinction).

[0027] When a question is asked to the AI ​​generation server device 5 using PC3 connected to LAN2, the system is configured to obtain and store answers for each combination of attribute information registered in attribute table 106 when a single employee asks a question. Specifically, it is configured to obtain answers for each employee classified by attribute information, such as answers for a user (employee) whose job title is engineer, whose workplace is Yokohama, and whose gender is male, and answers for a user (employee) whose job title is engineer, whose workplace is Yokohama, and whose gender is female. This allows the system to obtain the same answers all at once, even if each employee asks the same question.

[0028] The attribute table 106 can store tables with various contents. For example, as shown in Figure 5(B), it is possible to add age in addition to occupation, work location, and gender as attribute information. In the example shown in Figure 5(B), responses can be obtained even when categorized as under 30, under 40, under 50, etc. Of course, it is also possible to use only occupation and gender as attribute information. In other words, various combinations of various attribute information about employees are possible for the table data registered in the attribute table 106. In the following explanation, it will be assumed that the table data shown in Figure 5(A) is used.

[0029] As described above, the response storage unit 107 stores the responses obtained when the generating AI providing server device 5 is instructed to obtain responses for combinations of attribute information registered in the attribute table 106. Figure 6 is a diagram illustrating an example of the data stored in the response storage unit 107 of the relay device 1. As shown in Figure 6, the response data stored in the response storage unit 107 consists of a response ID, generation date and time, question content, and responses for each combination of attribute information.

[0030] The response ID is an identifier that uniquely identifies the response in question, and uniquely identifies the response data that is subject to recording, reading, and deletion in the response storage unit 107. The generation date and time is information indicating the date and time when the response data was generated and recorded in the response storage unit 107. The question content indicates the content of the main question to the generation AI providing server device 5. "Main" means that the question to the generation AI providing server device 5 excludes the part of the question that will be added by the relay device 1 in order to obtain answers for each attribute information, as will be described in more detail later. In other words, the question content means the question text itself entered by the questioner. In the example shown in Figure 6, it is a question asking about the color of a jacket to wear in autumn, such as "I'm looking for an autumn jacket. Please tell me what color jacket you recommend." Based on this question content, the generation AI providing server device 5 is checked to determine whether a similar question has already been asked.

[0031] In the example shown in Figure 6, the responses for each combination of attribute information refer to responses for each job type, work location, and gender. Specifically, an employee whose job type is engineer, work location is Yokohama, and gender is male answered that "navy" is a good color for a jacket to wear in the fall. Similarly, an employee whose job type is engineer, work location is Yokohama, and gender is female answered that "burgundy" is a good color for a jacket to wear in the fall. Furthermore, an employee whose job type is engineer, work location is Yokohama, and gender is not specified answered that "gray" is a good color for a jacket to wear in the fall. In this way, the response data is stored in the response storage unit 107 with so-called header information such as response ID, generation date and time, and question content added to the responses for each combination of attribute information corresponding to the table data registered in the attribute table 106.

[0032] This answer storage unit 107 stores and holds answer information that can be provided to users of PC3 connected to LAN2. Therefore, the answer storage unit 107 functions as a so-called cache storage unit, storing data that has been used once in a way that allows for high-speed access when it is used again later.

[0033] The connection terminal 108T constitutes the connection terminal to LAN2. LANI / F108 is the part that performs communication processing through LAN2. Specifically, LANI / F108 receives signals addressed to the local device that are transmitted via LAN2, converting them into a format that can be processed by the local device. LANI / F108 also converts signals to be transmitted from the local device to the target recipient into a transmission format, sends them to LAN2, and transmits them to the recipient.

[0034] The AI ​​generation unit 110 is a processing unit that functions when using the AI ​​generation server device 5 via a PC 3 connected to LAN 2. As shown in Figure 2, the AI ​​generation unit 110 includes a question reception unit 111, a stored answer search unit 112, a stored answer provision unit 113, an answer request formation unit 114, an answer acquisition unit 115, and an acquired answer provision unit 116.

[0035] The question receiving unit 111 receives information from a PC 3 connected to LAN2, received via the connection terminal 108T and LANI / F 108, if it is a question to the AI ​​generation server 5, including its URL and IP address. The question receiving unit 111 temporarily stores the received question and other information in a predetermined storage area of ​​the storage device 103, for example, and requests a search from the stored answer search unit 112. This information includes information such as the URL that identifies the AI ​​generation server 5 and the question, as well as information such as the MAC address and IP address of the sending PC 3, so the sending PC 3 can be identified.

[0036] If the information from PC3 is not a question for the AI ​​generation server device 5, it is sent to the Internet 4 via WANI / F101 and connection terminal 101T under the control of the control unit 102, and transmitted to the target web server, etc. This allows users to view the target web page or perform processes such as downloading and uploading data to and from the target web server.

[0037] The stored answer search unit 112, in response to a request from the question reception unit 111, compares the question text of a series of information, including the question text, temporarily stored in a predetermined storage area of ​​the storage device 103 with the question content of the answer data (Figure 6) stored in the answer storage unit 107 to determine the degree of similarity. If there is answer data with a similarity of a predetermined value or higher, the stored answer search unit 112 determines that the same question has already been asked and identifies that answer data to be used for the reply. Next, the stored answer search unit 112 identifies the combination of attribute information of the identified answer data. That is, in the example shown in Figure 6, the combination of attribute information of the answer data is identified as occupation, work location, and gender, and each of the combined attribute information is identified.

[0038] The stored response search unit 112 uses the MAC address and IP address of the PC3 that sent the series of information, including the question, temporarily stored in a predetermined storage area of ​​the storage device 103, to refer to the terminal management DB 105 and identify the employee ID of the employee using the PC3 that sent the question. Based on this identified employee ID, it refers to the employee DB 104 and obtains attribute information for that employee that corresponds to the combination of attribute information of the identified response data, i.e., in the example shown in Figure 6, the job title, work location, and gender. The stored response search unit 112 obtains the response information for the identified combination of attribute information of the identified response data, i.e., the recommended color in the example shown in Figure 6, and requests the stored response provision unit 113 to send this back to the PC3 that sent the question.

[0039] The stored answer provision unit 113, in response to a request from the stored answer search unit 112, forms an answer for PC3, the source of the question received, and sends it to LAN2 via LANI / F 108 and connection terminal 108T, and transmits it to the source PC3. For example, if the user (employee) of PC3, the source of the question, is an engineer, works in Yonezawa, and is female, then, based on the answer data shown in Figure 6, an answer will be returned informing that the recommended color for a jacket to wear in autumn is "olive". In this case, the answer can be returned using the stored information in the answer storage unit 107 without asking a question to the generation AI provision server device 5.

[0040] Furthermore, the stored response search unit 112 compares the question text received this time with the question content of the response data (Figure 6) stored in the response storage unit 107 to determine the similarity, and if it determines that there is no response data with a similarity of a predetermined value or higher, then it determines that a question similar to the question text received this time has not been asked in the past. In this case, the stored response search unit 112 requests the response request formation unit 114 to generate a question text for the generation AI providing server device 5 and to ask the question to the generation AI providing server device 5.

[0041] The response request formation unit 114 uses the question text received by the question reception unit 111, as well as the table data formed in the attribute table 106 (Figure 5(A)), to form a question for the generation AI provision server device 5, and then transmits it to the generation AI provision server device 5. Specifically, let's assume that the question text from the PC 3 connected to LAN2 is, for example, "I'm looking for an autumn jacket. Please tell me what color jacket you would recommend." Also, let's assume that the attribute table 106 has the table data shown in Figure 5(A) stored therein.

[0042] In this case, the response request formation unit 114 first forms the instruction "#Instruction, please create a table of answers for each of the following attributes. Please create a table of all combinations," and provides it to the generation AI providing server device 5. Next, the response request formation unit 114 organizes attribute information such as "#Attribute, Occupation: Engineer, Sales, Work location: Yokohama, Tokyo, Yonezawa, Gender: Male, Female, No answer," based on the table data of attribute table 106, and provides it to the generation AI providing server device 5. Finally, it provides the question received from PC3, such as "I'm looking for an autumn jacket. Please tell me your recommended jacket colors," to the generation AI providing server device 5.

[0043] This is just one example; a series of questions can also be formed and sent to the AI ​​generation server device 5. For example, a question such as, "I'm looking for an autumn jacket. What colors of jackets would you recommend? For this question, please create a table of answers for each of the following attributes. Please create a table of all possible combinations. #Attributes, Occupation: Engineer, Sales, Work Location: Yokohama, Tokyo, Yonezawa, Gender: Male, Female, No Answer." can be formed and sent. The response request formation unit 114 then sends the instruction, attribute information, question, or series of questions formed in this manner to the Internet 4 via WANI / F101 and connection terminal 101T, and transmits them to the AI ​​generation server device 5.

[0044] The response acquisition unit 115, upon receiving a response from the AI ​​generation server device 5 after the question sentence formed by the response request formation unit 114 has been sent to the AI ​​generation server device 5, adds header information to it to form response data in the manner shown in Figure 6, and records this in the response storage unit 107. The response ID, which is the header information, is, for example, obtained by storing the previously assigned response ID (number information) and adding the value 1 to it. The generation date and time, although not shown in Figure 2, can be obtained from the clock circuit provided by the relay device 1. The clock circuit only needs to be able to provide the current year, month, day and time. The question content can be the one temporarily stored in a predetermined storage area of ​​the storage device 103 by the question reception unit 111. The data other than this header information becomes the response (summarized in tabular format) provided by the AI ​​generation server device 5.

[0045] The acquired response provision unit 116 identifies the attribute information of the employee who is using the PC3 that sent the question, based on the response provided by the generation AI provision server device 5, acquires the corresponding response information for that attribute information, and sends it back to the PC3 that sent the question. Specifically, the acquired response provision unit 116 functions as follows: First, the acquired response provision unit 116 refers to the terminal management DB 105 using the MAC address and IP address of the PC3 that sent the series of information, including the question, temporarily stored in a predetermined storage area of ​​the storage device 103, to identify the employee ID of the employee using the PC3 that sent the question.

[0046] Based on this identified employee ID, the employee database 104 is referenced to obtain the attribute information of the employee corresponding to the combination of attribute information in the response data, i.e., in the example shown in Figure 6, the job title, work location, and gender. The acquired response provision unit 116 obtains the response information (recommended color in the example shown in Figure 6) for the identified combination of attribute information in the identified response data, sends it to LAN2 via LANI / F 108 and connection terminal 108T, and returns it to PC3, the source of the question.

[0047] In this way, when the relay device 1 uses the AI ​​generation server device 5 through the PC 3 connected to LAN 2, it also takes into account the attribute information of each user on the PC 3 connected to LAN 2 to effectively ask questions and obtain usable answers efficiently. As a result, when the same question is asked through the PC 3 connected to LAN 2, the relay device can answer using the answer data already obtained and held from the AI ​​generation server device 5 without having to access the AI ​​generation server device 5.

[0048] [Summary of processing performed by relay device 1] The processing performed by the relay device 1, which has the configuration described using Figures 2 to 6, will be summarized with reference to the flowchart. Figure 7 is a flowchart for explaining the processing performed by the relay device 1. The processing shown in the flowchart in Figure 7 is executed in the control unit 102, which controls each part of the generation AI utilization unit 110.

[0049] The control unit 102 controls the question receiving unit 111 to accept questions from the PC 3 connected to LAN2 to the AI ​​generation server device 5, among the information addressed to the device received through the connection terminal 108T and LANI / F 108 (step S101). In step S101, the question receiving unit 111 accepts the information from the PC 3 as a target for processing if it includes the URL or IP address of the AI ​​generation server device 5 and a question. Next, the control unit 102 determines whether or not it accepted a question for the AI ​​generation server device 5 in step S101 (step S102).

[0050] In the determination process of step S102, it is determined that no question has been received for the AI ​​generation server device 5. In this case, the control unit 102 determines that the information received from the PC is not a question for the AI ​​generation server device 5, and performs processing according to the received information (step S103), and repeats the process from step S101. That is, in step S103, various processes are performed according to the received information, such as browsing a web page, uploading data to the target server device, downloading data from the target server device, and sending and receiving emails.

[0051] In the determination process of step S102, it is determined that a question has been received by the AI ​​generation server device 5. In this case, the question receiving unit 111, under the control of the control unit 102, temporarily stores a series of pieces of information, including the received question, in a predetermined storage area of ​​the storage device 103, and identifies the terminal ID of the PC 3 that sent the question (step S104). Also in step S104, the question receiving unit 111 requests the stored answer search unit 112, through the control unit 102, to search for an answer corresponding to the question received.

[0052] In response to a request from the question receiving unit 111, the control unit 102 controls the stored answer search unit 112 to search whether answer data with a question similar to (considered identical to) the question received this time has already been stored in the answer storage unit 107 (step S105). Specifically, in step S105, the stored answer search unit 112 compares the question received this time with the question content of the answer data (Figure 6) stored in the answer storage unit 107 to determine the similarity between the two. Next, the stored answer search unit 112 determines whether or not there is answer data with a question content that is so similar that it can be considered identical to the question received this time, based on the determined similarity being above a predetermined value (step S106).

[0053] In the determination process of step S106, it is determined that there is (or exists) answer data with a question whose similarity to the question received is greater than or equal to a predetermined value. In this case, the stored answer search unit 112 identifies the employee ID by referring to the terminal management DB 105 based on the terminal ID of the questioner identified in step S103, and identifies the attribute information of the questioner by referring to the employee DB 104 based on the identified employee ID (step S107). In step S107, the stored answer search unit 112 requests the stored answer provision unit 113, through the control unit 102, to acquire and return answer information for the combination of attribute information identified for the identified answer data.

[0054] The stored response provision unit 113, under the control of the control unit 102, acquires response information that matches the attribute information of the identified questioner in response to a request from the stored response search unit 112 (step S108). That is, in step S108, the stored response provision unit 113 acquires response information that matches the attribute information of the questioner (job type, workplace, gender) identified in step S107 from the response data identified in step S105 from the stored data of the response storage unit 107. The response information acquired here, in the case of the response data shown in Figure 6, is, for example, a recommended color such as navy or burgundy.

[0055] Next, the stored answer provision unit 113 performs the process of providing the answer information acquired in step S107 to the questioner (step S109). That is, in step S109, the stored answer provision unit 113 first forms an answer to PC3, the source of the question received, based on the acquired answer information. The answer formed here corresponds to the received question and may be, for example, "The recommended color for a jacket to wear in autumn is navy." Next, in step S109, the stored answer provision unit 113 sends the formed answer to LAN2 via LANI / F108 and connection terminal 108T, and transmits it to the source PC3.

[0056] After the processing in step S109, the processing from step S101 is repeated. In this way, when a question is received by the generation AI providing server device 5, it is first checked whether the same question has already been asked by searching the answer data (Figure 6) in the answer storage unit 107. If it is confirmed that the same question has already been asked, an answer to the questioner can be formed using the answer data already stored for the same question, and a reply can be sent to the questioner. In this case, no access is made to the generation AI providing server device 5.

[0057] On the other hand, suppose that in the discrimination process of step S106, it is determined that there is no answer data with question content that has a similarity to the question received this time that is equal to or greater than a predetermined value. In this case, the stored answer search unit 112 requests the answer request formation unit 114, via the control unit 102, to generate a question for the generation AI providing server device 5 and send the question to the generation AI providing server device 5. The answer request formation unit 114 uses the question received this time, as well as the table data formed in the attribute table 106 (Figure 5(A)), to form a question for the generation AI providing server device 5 and sends it to the generation AI providing server device 5 (step S110). In the process of step S110, as described above, the answer request formation unit 114 requests the generation AI providing server device 5 to provide answers in a table, summarizing the answers for each specified attribute in response to the question from PC3.

[0058] The response acquisition unit 115 receives the response to the question sent in step S110 when it is returned from the generation AI providing server device 5, pairs the received response with the question, and records it in the response storage unit 107 (step S111). Specifically, in step S111, the response acquisition unit 115 adds header information to the tabular response from the generation AI providing server device 5 to form response data in the manner shown in Figure 6, and records this in the response storage unit 107.

[0059] Next, the acquired response provision unit 116 operates under the control of the control unit 102 and performs the process of providing the questioner with an answer to the question received (step S112). In step S112, the acquired response provision unit 116 performs the following processes. First, (1) it refers to the terminal management DB 105 using the terminal ID (MAC address or IP address) identified in step S104 to identify the employee ID of the employee using the sending PC 3. Next, (2) it refers to the employee DB 104 based on the identified employee ID and obtains the attribute information of the employee corresponding to the combination of attribute information in the response data. Next, (3) it obtains the response information corresponding to the combination of attribute information obtained from the response data acquired from the generation AI provision server device 5. Finally, (4) it forms a response including the acquired response information and sends it to LAN2 via LANI / F 108 and connection terminal 108T, returning it to the PC 3 that sent the question.

[0060] After the processing in step S112, the processing from step S101 is repeated. As described above, when the relay device 1 receives a question from the generation AI providing server device 5, it first checks whether the same question has already been asked by searching the answer data (Figure 6) in the answer storage unit 107. If it is not confirmed that the same question has already been asked, the relay device 1 considers not only the attribute information of the current questioner but also the table data of the attribute table 106 to form a question and send it to the generation AI providing server device 5.

[0061] As explained using Figure 6, this allows for obtaining answers that are expanded based on attribute information and summarized in a tabular format, which can then be stored in the answer storage unit 107. Subsequently, even if all employees ask the same question, they can receive an answer using the stored information in the answer storage unit 107 without having to repeatedly ask the AI ​​generation server device 5. Of course, the system can also form and provide an answer to the sender of the question received, based on the answers provided, tailored to the questioner's attribute information.

[0062] [Effects of the embodiment] As described above, the relay device of the above embodiment allows for effective questioning of the AI ​​generation server device (artificial intelligence chatbot) 5 and efficient acquisition of usable answers. This helps to suppress excessive use of the artificial intelligence chatbot and enables its effective utilization.

[0063] Specifically, suppose relay device 1 receives the question "I'm looking for an autumn jacket. What jacket colors would you recommend?" from PC 3 connected to LAN 2. In this case, relay device 1 first issues the instruction "#Instruction, please create a table of answers for each of the following attributes. Please create a table of all combinations." Next, the response request formation unit 114 provides attribute information such as "#Attribute, Occupation: Engineer, Sales, Work location: Yokohama, Tokyo, Yonezawa, Gender: Male, Female, No answer." Finally, it sends the question received from PC 3, "I'm looking for an autumn jacket. What jacket colors would you recommend?" In this way, questions can be effectively made to the generation AI providing server device (artificial intelligence chatbot) 5.

[0064] This allows for obtaining responses corresponding to all employees using PC3 connected to LAN2, and storing the response data in the response storage unit 107 in the manner shown in Figure 6. In other words, responses can be obtained efficiently. Therefore, if a similar question is sent from another PC3 connected to LAN2, the relay device 1 can use the response data in the response storage unit 107 to form a response that matches the user of the original PC3 and provide it to the original PC3. In this case, there is no access to the generation AI provision server device 5, so excessive use of the generation AI provision server device 5 can be suppressed, and the generation AI provision server device 5 can be used effectively. Therefore, if the generation AI provision server device 5 uses a pay-per-use system, the usage fee can be reduced.

[0065] Furthermore, the questioner does not even need to input their own attribute information; they only need to enter the main question text to obtain response data that takes into account their own attribute information as well as the attribute information of other employees.

[0066] [Differentiations, etc.] In the embodiment described above, as explained with reference to Figure 5, the attribute table 106 allows for various combinations of various attribute information about employees. For this reason, multiple types of table data are prepared in the attribute table 106. When the relay device 1 receives a question from the generation AI provision server device 5, it may present an overview of the multiple types of table data to the questioner via the PC 3 used by the questioner, allowing the questioner to select which table data to use.

[0067] In this case, the table data outline presents combinations of attribute information, such as (1) job title, work location, gender, (2) job title, work location, gender, age, (3) job title, work location, gender, department, etc., and allows the user to select one of these. This allows the user to choose the table data that best suits their intent from among the multiple types of table data stored in attribute table 106.

[0068] Furthermore, while the above-described embodiment was explained as combining multiple types of attribute information, it is not limited to this. For example, it is certainly possible to select one type of attribute information and obtain responses for each aspect of that attribute information, such as obtaining responses for each job type, each work location, or each gender. In other words, if the attribute information is only job type, it is certainly possible to obtain responses for both engineers and sales positions, or if the attribute information is work location, it is possible to obtain responses for Yokohama, Tokyo, and Yonezawa.

[0069] Furthermore, the employee database 104 and terminal management database 105 can also be configured in various ways by increasing or decreasing the amount of information they hold. Also, although only one generation AI providing server device 5 is shown in the embodiment described above, it is not limited to this. Various types of generation AI providing server devices 5 exist. For this reason, a PC 3 connected to LAN 2 can select a target generation AI providing server device and request the relay device 1 to send a question to this selected generation AI providing server device. In this case, the relay device 1 should be instructed with the URL or IP address of the generation AI providing server device to which the question will be sent.

[0070] Furthermore, in the embodiment described above, the question from PC3 was to ask for a recommended color for a jacket to wear in autumn, but it goes without saying that this is not the only example. Various questions can be asked to obtain answers for each attribute based on the table data of attribute table 106, and these answers can be used. [Explanation of Symbols]

[0071] 1...Relay device, 101T...Connection terminal, 101...WAN I / F, 102...Control unit, 103...Storage device, 104...Employee DB, 105...Terminal management DB, 106...Attribute table, 107...Answer storage unit, 108T...Connection terminal, 108...LAN I / F, 110...Generating AI utilization unit, 111...Question reception unit, 112...Stored answer search unit, 113...Stored answer provision unit, 114...Answer request formation unit, 115...Answer acquisition unit, 116...Acquired answer provision unit, 2...LAN, 3, 3(1), 3(2), 3(3)...PC, 4...Internet, 5...Generating AI provision server device

Claims

1. A relay device installed between a LAN (Local Area Network) and the Internet, which relays communication between multiple terminal devices connected to the LAN and an artificial intelligence (AI) providing server device connected to the Internet. An attribute storage means for storing and maintaining attribute information of users of multiple terminal devices connected to the LAN, A question receiving means that receives a question sentence from the terminal device to the AI ​​generation server device input by the user, A response request forming means for forming a response request to obtain a response to the question received by the question receiving means for each attribute information stored in the attribute storage means, and providing the response request to the generation AI providing server device, A response acquisition means that acquires responses for each attribute information from the AI ​​generating server device in response to the response request and records them in a response storage means, When an answer is obtained from the generating AI providing server device through the answer acquisition means, the acquired answer providing means identifies the user's attribute information based on the user information provided along with the question text received through the question receiving means, and sends back the answer associated with the identified user's attribute information from the answer acquired by the answer acquisition means to the terminal device used by the user. A relay device characterized by being equipped with the following features.

2. A relay device according to claim 1, Based on the information about the user provided along with the question received through the question receiving means, the stored answer search means identifies the attribute information of the user, and searches for the answer stored in the answer storage means that corresponds to the attribute information of the identified user. When the stored response search means can retrieve the response associated with the attribute information of the identified user, the stored response provision means sends a reply to the terminal device used by the user. Equipped with, The aforementioned response request formation means, response acquisition means, and acquired response provision means function when the stored response search means cannot retrieve the response associated with the attribute information of the identified user. A relay device characterized by the following features.

Images