System, computer device, and method

Abstract

Provided is a system that makes it possible to obtain a response by means of simple user input.　Provided is a system that comprises at least one computer device. The system comprises a selection means that receives selection of at least one option from among a plurality of options that respectively correspond to a plurality of keywords in response to input from a user terminal that is operated by a user and a first generation means that generates a prompt for input into a language model or a provisional prompt for generating the prompt so as to include the keywords that correspond to the options received as selections by the selection means.

Description

Systems, computer devices and methods

[0001] The present invention relates to a system, a computer device and a method.

[0002] In recent years, the accuracy of generative AI, which can autonomously generate text and images, has improved, and it is expected to be used in various fields. For example, a system has been provided that uses large-scale language models (LLMs) to perform natural language processing tasks such as question and answering, text summarization, and text generation (see, for example, Patent Literature 1). An LLM is a natural language processing model trained using a large amount of text data. When using an LLM to perform a task such as question and answering, text is input into the LLM, and an answer is output from the LLM. The text input into the LLM is called a prompt.

[0003] However, it was difficult for users who were not familiar with language models to input prompts themselves. Also, depending on the content of the prompts input into the language model, the output answer might differ from the answer the user wanted. For this reason, inputting prompts required individual skills, making it difficult for users who were not familiar with language models to use them.

[0004] Patent No. 7441366

[0005] The present invention can solve, for example, any of the following problems: A first problem of the present invention is to provide a system that allows a user to obtain an answer with a simple input. A second problem of the present invention is to provide a system that reduces the number of times a language model outputs an answer.

[0006] The present invention addresses the following problems: [1] A system comprising at least one computer device, the system comprising: a selection means for accepting a selection of one or more options from a plurality of options corresponding to each of a plurality of keywords in response to an input from a user terminal operated by a user; and a first generation means for generating a prompt to be input to a language model so as to include a keyword corresponding to the option selected by the selection means, or a provisional prompt for generating the prompt; [2] The system according to [1], wherein the user terminal comprises a prompt display means for controlling the first generation means to display the prompt or provisional prompt generated by the first generation means; [3] The system according to [1] or [2], wherein the user terminal comprises an option display means for controlling the display of options, the selection means accepts a selection of one or more options from the options controlled to be displayed by the option display means, and when one option is selected from the plurality of options by the selection means, the option display means controls the display of options corresponding to a keyword associated with the selected option; [4] The system according to [3], wherein a keyword associated with a first option and a keyword associated with a second option are different from the plurality of options displayed; [5] The system according to any one of [2] to [4], wherein the prompt display means controls to display a keyword corresponding to an option selected by the selection means in a manner different from other text included in the generated prompt or provisional prompt; [6] The system according to any one of [1] to [5], wherein the system includes an input means for receiving input of text information in response to an input from the user terminal, and the first generation means generates the prompt or the provisional prompt based on the keyword corresponding to the option selected by the selection means and the text information received by the input means; [7] The system according to any one of [1] to [6], wherein the first generation means generates the prompt or the provisional prompt by embedding a keyword corresponding to the option selected by the selection means into a preset template;[8] The system according to any one of [1] to [7], further comprising an answer acquisition means for acquiring an answer from a language model based on the prompt generated by the first generation means or a prompt generated from a provisional prompt; [9] The system according to any one of [1] to [7], further comprising: an information acquisition means for acquiring, in the language model, related information from a specialized information database in which predetermined information is stored, based on the prompt generated by the first generation means or a prompt generated from a provisional prompt, and an answer acquisition means for acquiring an answer from the language model using the information acquired by the information acquisition means and the prompt;

[10] The system according to any one of [1] to [7], further comprising: an information acquisition means for acquiring related information from a specialized information database in which predetermined information is stored, based on the generated provisional prompt or a keyword corresponding to an option whose selection has been accepted, a second generation means for generating a prompt based on the information acquired by the information acquisition means and the provisional prompt or the keyword, and an answer acquisition means for acquiring an answer from a language model based on the prompt generated by the second generation means;

[11] The system according to any one of [1] to

[10] , wherein the language model comprises an extraction means for extracting predetermined keywords by referring to a specialized information database in which predetermined information is stored, and the selection means accepts selection of options from options corresponding to the keywords extracted by the extraction means;

[12] The system according to any one of [1] to

[11] , wherein the user terminal comprises an option display means for controlling the display of options, the selection means accepts selection of one or more options from the options controlled to be displayed by the option display means, and the option display means controls to display at least two or more options in a list, and controls to change and display an area displaying each of the options according to information on the number of times the option has been selected by the selection means;

[13] A computer device comprising: a selection means for accepting a selection of one or more options from a plurality of options corresponding to each of a plurality of keywords in response to an input from a user terminal operated by a user; and a first generation means for generating a prompt to be input into a language model, or a provisional prompt for generating the prompt, so that the prompt includes a keyword corresponding to the option selected by the selection means;

[14] A method executed on at least one computer device, comprising: a selection step for accepting a selection of one or more options from a plurality of options corresponding to each of a plurality of keywords in response to an input from a user terminal operated by a user; and a first generation step for generating a prompt to be input into a language model, or a provisional prompt for generating the prompt, so that the prompt includes a keyword corresponding to the option selected by the selection step;

[15] A system comprising at least one computer device, the system comprising: a first acquisition means for acquiring an answer from a language model based on a prompt; a storage means for storing the answer acquired by the first acquisition means in association with information on the prompt corresponding to the answer; a first generation means for generating or acquiring a prompt to be input to the language model or a provisional prompt for generating the prompt; and an identification means for identifying an answer corresponding to information on the prompt from the storage means when information on the prompt or provisional prompt generated or acquired by the first generation means and a prompt satisfying a predetermined condition is stored;

[16] The system according to

[15] , comprising: a second acquisition means for acquiring an answer from a language model based on a prompt generated from the prompt or provisional prompt when information on the prompt generated or acquired by the first generation means and a prompt satisfying a predetermined condition is not stored;

[17] The system according to

[16] , further comprising a second generation means for generating a prompt by adding predetermined information to the provisional prompt, a second acquisition means for acquiring answers from a language model based on the prompt generated by the second generation means, and a storage means for storing at least some of the answers acquired by the second acquisition means in association with information on the prompt corresponding to at least some of the answers;

[18] The system according to any one of

[15] to

[17] , comprising an information acquisition means for acquiring related information from a specialized information database in which predetermined information is stored based on a prompt, a first acquisition means for acquiring an answer from a language model using the related information and the prompt, and an information update means for adding, replacing, and / or deleting predetermined information from the specialized information database, and a deletion means for deleting information on the answer and / or the prompt corresponding to the answer stored in the storage means in response to the addition, replacement, and / or deletion of the predetermined information by the information update means;

[19] A computer device comprising: a first acquisition means for acquiring an answer from a language model based on a prompt, a storage means for storing the answer acquired by the first acquisition means in association with information on the prompt corresponding to the answer, a first generation means for generating or acquiring a prompt to be input to the language model or a provisional prompt for generating the prompt, and an identification means for identifying an answer corresponding to the information on the prompt when the prompt or provisional prompt generated or acquired by the first generation means and information on a prompt that satisfies a predetermined condition are stored in the storage means;

[20] A method executed on at least one computer device, the method comprising: a first acquisition step of acquiring an answer from a language model based on a prompt; a storage step of associating the answer acquired in the first acquisition step with information on the prompt corresponding to the answer and storing the association; a first generation step of generating or acquiring a prompt to be input to the language model or a provisional prompt for generating the prompt; and an identification step of identifying an answer corresponding to the information on the prompt when the prompt or provisional prompt generated or acquired in the first generation step and information on a prompt that satisfies a predetermined condition are stored in the storage step.

[0007] According to the present invention, it is possible to provide a system that allows a user to obtain an answer through simple input, and a system that reduces the number of times a language model outputs an answer.

[0008] 1 is a block diagram showing a configuration of a system according to an embodiment of the present invention. FIG. 2 is a block diagram showing a hardware configuration of a user terminal according to an embodiment of the present invention. FIG. 3 is a block diagram showing a hardware configuration of a server device according to an embodiment of the present invention. FIG. 4 is a diagram showing a flowchart of specialized information update processing according to an embodiment of the present invention. FIG. 5 is a diagram showing a flowchart of keyword setting processing according to an embodiment of the present invention. FIG. 6 is a diagram showing a flowchart of output processing according to an embodiment of the present invention. FIG. 7 is a diagram showing a flowchart of answer acquisition processing according to an embodiment of the present invention. FIG. 8 is a diagram showing a flowchart of another example of answer acquisition processing according to an embodiment of the present invention. FIG. 9 is a diagram showing an example of a display screen according to an embodiment of the present invention. FIG. 10 is a diagram showing an example of a display screen according to an embodiment of the present invention. FIG. 11 is a diagram showing an example of a display screen according to an embodiment of the present invention. FIG. 12 is a diagram showing an example of a display screen according to an embodiment of the present invention. FIG. 13 is a diagram showing a flowchart of display order change processing according to an embodiment of the present invention.

[0009] The following describes embodiments of the present invention, but the present invention is not limited to the following embodiments as long as they do not violate the spirit of the present invention. The order of each process constituting the flowcharts described below is random as long as no contradictions or inconsistencies occur in the process content. Furthermore, it is possible to omit some of the processes constituting the flowcharts or add new processes to each process constituting the flowcharts as long as they do not violate the spirit of the present invention. Furthermore, the device that executes each process constituting the flowcharts can be changed to another device as long as it does not violate the spirit of the present invention. In this case, the process content can be changed so as not to cause contradictions or inconsistencies in the process content.

[0010] The system according to the present invention can assist in searches to obtain answers related to specific information. The specific information is not particularly limited, and the system according to the present invention can be applied in various fields. The following mainly describes the application of the system according to the present invention in the field of long-term care insurance.

[0011] [System Configuration] Fig. 1 is a block diagram showing the configuration of a system according to an embodiment of the present invention. The system 10 includes at least one computer device. The system 10 includes, for example, a user terminal 1 and a server device 2. The system 10 may also include a language model server 3, a specialized information database 4, and an administrator terminal 5. The user terminal 1, the server device 2, the language model server 3, the specialized information database 4, and the administrator terminal 5 are communicably connected to each other via a communication network 6. In the system 10, any of the user terminal 1, the server device 2, and the language model server 3 can function as an information processing device. When any of the devices functions as an information processing device, information is transmitted and received between the user terminal 1, the server device 2, the language model server 3, and the specialized information database 4 as necessary.

[0012] The system 10 may include multiple user terminals 1, multiple server devices 2, multiple specialized information databases 4, and multiple administrator terminals 5. The server devices 2 may function in a distributed manner across multiple computer devices. For example, instead of the server devices 2, a distributed ledger technology such as a blockchain may be used.

[0013] [User Terminal] The user terminal 1 is operated by a user of the system 10. The user terminal 1 is not particularly limited, but examples thereof include a conventional mobile phone, a tablet terminal, a smartphone, a laptop computer, and a desktop computer.

[0014] 2 is a block diagram showing the hardware configuration of a user terminal according to an embodiment of the present invention. The user terminal 1 includes a control unit 11, a RAM 12, a storage unit 13, an input unit 14, a display unit 15, and a communication interface 16, all of which are connected by a bus.

[0015] The control unit 11 is composed of a CPU and a ROM. The control unit 11 executes programs stored in the storage unit 13 and controls the user terminal 1. The RAM 12 is a work area for the control unit 11. The storage unit 13 is a memory area for saving programs and data. In other words, the storage unit 13 functions as a recording medium that stores programs. The control unit 11 performs arithmetic processing based on the programs and data read from the RAM 12 and data input via the input unit 14.

[0016] The display unit 15 has a display screen. The control unit 11 outputs a video signal for displaying an image on the display screen according to the result of the arithmetic processing. Here, the display screen of the display unit 15 may be a touch panel equipped with a touch sensor. In this case, the touch panel functions as the input unit 14. The communication interface 16 can be connected to the communication network 6 wirelessly or via a cable, and can send and receive data to and from other computer devices via the communication network 6. Data received via the communication interface 16 is loaded into the RAM 12, and the control unit 11 performs arithmetic processing.

[0017] 3 is a block diagram showing the hardware configuration of a server device according to an embodiment of the present invention. The server device 2 includes at least a control unit 21, a RAM 22, a storage unit 23, and a communication interface 24, which are connected to each other via a bus.

[0018] The control unit 21 is composed of a CPU and ROM, and executes programs stored in the storage unit 23 to control the server device 2. The control unit 21 also has an internal timer that measures time. The RAM 22 is the work area of ​​the control unit 21. The storage unit 23 is a storage area for saving programs and data. In other words, the storage unit 23 functions as a recording medium that stores programs. The control unit 21 reads the programs and data from the RAM 22, and performs program execution processing based on information received from the user terminal 1 or the language model server 3, etc.

[0019] [Language Model Server] The language model server 3 is a computer device that stores a language model for processing natural language. The hardware configuration of the language model server 3 can be appropriately adapted from the description of the hardware configuration of the server device 2. In this embodiment, the LLM is used as the language model, but a language model other than the LLM may be used as long as it can process natural language.

[0020] The server device 2 may store a language model. When a language model is stored in the server device 2, the system 10 does not need to include the language model server 3. Furthermore, a small-scale language model that is smaller in scale than the LLM may be stored in the server device 2 or the language model server 3. The small-scale language model is a natural language processing model trained based on less data than the LLM.

[0021] [Specialized Information Database] The specialized information database 4 is a database in which predetermined information is stored. In this embodiment, the predetermined information stored in the specialized information database 4 is also referred to as specialized information. The specialized information database 4 is stored in a computer device. The specialized information database 4 may be stored in the server device 2 or the language model server 3. The specialized information database 4 may be stored in a distributed manner across multiple computer devices. The specialized information database 4 may be constructed as a different database for each attribute of the predetermined information.

[0022] The system 10 according to the present embodiment can extract predetermined keywords from a language model and / or output answers by utilizing predetermined information stored in the specialized information database 4. The system 10 can utilize so-called RAG (Retrieval-Augmented Generation).

[0023] [Administrator Terminal] The administrator terminal 5 is operated by an administrator who manages the system 10. The administrator terminal 5 is not particularly limited, but examples thereof include conventional mobile phones, tablet terminals, smartphones, laptops, and desktop computers. The hardware configuration of the administrator terminal 5 can be appropriately adapted from the description of the hardware configuration of the user terminal 1.

[0024] In this embodiment, the program may be stored on a recording medium such as a CD-ROM. In this case, the program stored on the recording medium may be installed in the user terminal 1, the server device 2, or the language model server 3 to execute the predetermined functions. Alternatively, the program may be distributed from a computer device external to the system. In this case, the program distributed from the computer device external to the system may be installed in the user terminal 1, the server device 2, or the language model server 3 to execute the predetermined functions.

[0025] [Specialized Information Update Process] Next, specialized information update process for storing specialized information in the specialized information database 4 will be described.

[0026] The specialized information stored in the specialized information database 4 is not particularly limited, but is preferably information related to the field in which the system 10 is used. For example, in the field of long-term care insurance, the specialized information may be information issued by the Ministry of Health, Labor and Welfare. The specialized information database 4 may be set for each field. The specialized information database 4 may also be set for each department that uses the system 10. For example, information related to departments such as the general affairs department, human resources department, and accounting department may be stored in the specialized information database 4.

[0027] The specialized information stored in the specialized information database 4 may be, for example, text information, image information, video information, sound information, etc. The specialized information stored in the specialized information database 4 may be stored in association with vectorized numerical information obtained by performing vector calculations. The specialized information database 4 may be a vector database.

[0028] Registration of specialized information in the specialized information database 4 may be performed in response to an operation on the administrator terminal 5. Fig. 4 is a flowchart showing specialized information update processing according to an embodiment of the present invention.

[0029] First, the administrator logs in to the system 10 by accessing the server device 2 via a web browser from the administrator terminal 5. Alternatively, the administrator may log in to the system 10 by launching an application program (hereinafter referred to as a dedicated app) downloaded to the administrator terminal 5 and accessing the server device 2.

[0030] When a user of the system 10 logs in to the system, the user may be required to input a user ID, a password, and the like. The password may be set by the user. The password is stored in the system in association with the user ID. The user ID is identification information that can identify the user.

[0031] Next, the administrator terminal 5 accepts input of instruction information through operation by the administrator (step S11). The instruction information is information for storing specialized information in the specialized information database 4. The instruction information includes the specialized information and information regarding whether to add the specialized information to the specialized information database 4, replace the specialized information with already stored specialized information, or delete the specialized information from the specialized information database 4. The instruction information may also include information regarding the file name of the information, comments to be made when the information is displayed on the user terminal 1, etc.

[0032] Next, the administrator terminal 5 transmits an update request to the server device 2 requesting that the specialized information database 4 be updated (step S12). The update request includes instruction information. Upon receiving the update request (step S13), the server device 2 updates the specialized information in the specialized information database 4 (step S14). If the specialized information database 4 is stored in the server device 2, step S14 is executed by the server device 2. If the specialized information database 4 is stored in a computer device other than the server device 2, the server device 2 transmits the update request received in step S13 to the computer device in which the specialized information database 4 is stored. The processing of step S14 is executed by that computer device.

[0033] Here, updating the specialized information refers to a process of adding, replacing, and / or deleting specific information (specialized information) from the specialized information database. The specialized information may be stored in the specialized information database 4 in association with the file name of the information, comments to be made when the information is displayed on the user terminal 1, etc. Furthermore, information indicating that the specific information has been added, replaced, and / or deleted from the specialized information database 4 may be stored in association with information regarding the execution date and time, the administrator who instructed the execution, the file name, comments, etc. The execution date and time may be any of the time when the instruction information is received in step S11, the time when the update request is sent in step S12, the time when the update request is received in step S13, and the time when the specialized information is updated in step S14.

[0034] After executing step S14, the server device 2 deletes the history information in response to the addition, replacement, and / or deletion of the specialized information in step S14 (step S15). The specialized information update process is completed by the above steps S11 to S15.

[0035] The history information stores an answer acquired from a language model in association with information about a prompt corresponding to the answer. In the present embodiment, the system 10 can acquire an answer from a language model based on a prompt. "Acquiring an answer from a language model" refers to inputting a prompt into the language model, causing the language model to output an answer. The system 10 also stores an answer acquired from a language model in association with information about the prompt corresponding to the answer. The information about the prompt may include, for example, the prompt itself input into the language model, information about keywords and / or text information included in the prompt, information about the answer level included in the prompt, information about the output format of the answer included in the prompt, and a temporary prompt used to generate the prompt. The history information may also store attribute information about the user who input the prompt in association with information about the answer or prompt. The attribute information may include, for example, the user's job title, years of service, gender, age, etc.

[0036] In this embodiment, a question or instruction input into a language model to obtain an answer from the language model is defined as a "prompt," and information generated in the process of generating a prompt based on input and / or selection at the user terminal 1 is defined as a "temporary prompt."

[0037] The history information is stored, for example, in a database in the storage unit 23 of the server device 2. In step S15, the server device 2 can delete information about the answers and / or prompts corresponding to the answers stored in the database in the storage unit 23.

[0038] Note that "deleting history information" in step S15 may mean completely erasing the information to be deleted from the database of the storage unit 23, or may mean setting a deletion flag on the information to be deleted. Information that has been flagged for deletion is excluded from the history information that is the subject of determination in step S61 or S71, which will be described later.

[0039] In step S15, the history information to be deleted may be all history information stored in the database, or may be history information related to the specialized information updated in step S14. History information related to specialized information may be, for example, history information associated with a prompt that includes the specialized information, history information associated with an answer that includes the specialized information, or history information associated with an answer that references the specialized information.

[0040] In the above description, the processes of steps S11 and S12 are executed by the administrator terminal 5, but they may also be executed by the user terminal 1. Furthermore, instead of the processes of steps S11 to S13, the process of step S14 may be executed in response to the satisfaction of a predetermined condition. For example, when addition, replacement, and / or deletion of information is detected in a predetermined database, addition, replacement, and / or deletion of information may be executed in the specialized information database 4 so as to correspond to the information.

[0041] [Keyword Setting Process] Next, a description will be given of a keyword setting process for setting keywords corresponding to options displayed on the user terminal 1. The keyword setting process may be executed after the specialized information update process is executed. The keyword setting process may also be executed in response to an operation input to the administrator terminal 5.

[0042] In this embodiment, a keyword is text information indicating a specific content. A keyword may be a single word or may include multiple words. The system 10 allows the administrator terminal 5 to operate and register keywords to be displayed as options on the user terminal 1 in the storage unit 23. When registering a keyword, attributes such as a major category, a medium category, and a small category may be set for each keyword (two attributes, a major category and a small category, may be used, and the number of attributes is not particularly limited. Attributes represented by numbers such as a first category, a second category, and a third category may also be used). For example, the major category, the medium category, and the small category may be higher-level concepts, lower-level concepts than the major category, and lower-level concepts than the medium category, respectively.

[0043] Furthermore, among the words constituting the prompt, the upstream concept for identifying the content of the prompt may be referred to as a major category, the downstream concept from the major category may be referred to as a medium category, and the downstream concept from the medium category may be referred to as a small category. For example, if "home care" or "home nursing" is a keyword for the major category, "problem," "fee," and "method" may be keywords for the medium and small categories. By registering attributes for keywords in this manner, for example, multiple options for the major category are displayed on the user terminal 1, and selecting at least one of these options displays multiple options for the medium category. Selecting at least one of the options for the displayed medium category on the user terminal 1 displays multiple options for the small category, and it becomes possible to select at least one of these options.

[0044] Furthermore, when registering a keyword, the system 10 can register related keywords related to each keyword in the storage unit 23. The system 10 can register multiple related keywords for one keyword. Furthermore, when registering this related keyword, the system 10 can register related keywords related to this related keyword in the storage unit 23. In this case, multiple related keywords can also be registered for one related keyword. By registering related keywords in this manner, for example, multiple options are displayed on the user terminal 1, and when at least one of these options is selected, options corresponding to the related keywords registered for the keyword corresponding to the selected option are displayed. Furthermore, when at least one of these options is selected, options corresponding to the related keywords registered for the related keyword corresponding to the selected option are displayed.

[0045] 5 is a flowchart showing the keyword setting process according to the embodiment of the present invention. First, the administrator logs in to the system 10 from the administrator terminal 5. The process of logging in to the system 10 is the same as the process of updating specialized information, and therefore will not be repeated here.

[0046] Next, the administrator terminal 5 transmits a keyword setting request to the server device 2, requesting that a keyword be set by an operation of the administrator (step S21).

[0047] The keyword setting request sent in step S21 may include a prompt regarding what kind of keywords to set. For example, the prompt included in the keyword setting request may be, "Refer to the information stored in the specialized information database 4 for long-term care insurance, and set keywords by dividing into major categories, medium categories, and minor categories. Set the keywords for the medium categories to correspond to the keywords for the major categories and to subdivide each keyword. Set the keywords for the minor categories to correspond to the keywords for the medium categories and to subdivide each keyword."

[0048] The keyword setting request may include information about the term or field of the keyword to be set. For example, in step S21, the administrator inputs the term or field for which the keyword is to be set on the administrator terminal 5. Alternatively, in step S21, the administrator may input the specialized information database 4 for which the keyword is to be set on the administrator terminal 5. When the input information is transmitted to the server device 2 as a keyword setting request, the server device 2 may generate a prompt regarding what keyword to set based on the input information. For example, the server device 2 stores a template for a prompt regarding what keyword to set, and the server device 2 can generate the above-described prompt by embedding information about the input term or field or the field set in the specialized information database 4.

[0049] When the server device 2 receives the keyword setting request (step S22), it uses the language model to extract keyword information related to the keyword (step S23).

[0050] The processing of step S23 will be described. In step S23, the server device 2 transmits a keyword setting request to the language model server 3. Upon receiving the keyword setting request, the language model server 3 causes the language model to refer to the specialized information database 4 to extract predetermined keywords. More specifically, the language model server 3 inputs a prompt included in the keyword setting request or a prompt generated based on the keyword setting request into the language model, causing the language model to refer to the specialized information database 4 and extract predetermined keywords.

[0051] The "predetermined keyword" is a keyword that may be displayed in an option for generating a prompt or a provisional prompt in the output process described below. The "predetermined keyword" is a keyword that has a predetermined relationship with the term or field input to the administrator terminal 5. Examples of the predetermined relationship include synonymous / neighbor relationships, superordinate / subordinate concept relationships, part / whole relationships, related concepts / related themes, properties / characteristics / attributes, procedures / methods / processes, purposes / effects / roles, issues / problems / backgrounds, related laws / systems, and use examples / specific cases. The predetermined keyword may also be a keyword having attributes such as a major category, a medium category, or a minor category. The relationship between the major category and the term or field input to the administrator terminal 5, the relationship between the major category and the medium category, or the relationship between the medium category and the minor category may have the above-mentioned predetermined relationship.

[0052] The server device 2 stores the keyword information extracted in step S23 (step S24), and deletes the history information (step S25). The keyword setting process is completed by the above steps S21 to S25.

[0053] The keyword information includes text information indicating the keyword and information about the keyword category, including information about the category to which the keyword belongs and the keyword to which the keyword is associated.

[0054] The history information in step S25 has the same meaning as the history information in step S15. The history information deleted in step S25 may be all history information stored in the database, or may be history information related to the keyword information stored in step S24. History information related to keyword information is, for example, information associated with information related to the keyword corresponding to the keyword information, or history information associated with a prompt including the keyword corresponding to the keyword information.

[0055] The process of step S25 may be omitted. If the language model is stored in the server device 2, step S23 is executed in the server device 2.

[0056] Note that the keyword setting process may be performed in step S24 instead of or in addition to the processes of steps S21 to S23, by inputting keyword information to the administrator terminal 5. For example, the administrator may be able to set a keyword by inputting information about a keyword that the administrator has created himself or herself to the administrator terminal 5.

[0057] After the processing of step S23, the extracted keyword information may be controlled to be displayed on the administrator terminal 5. The administrator can check the displayed keywords and change, add, and / or delete the keywords by operating the administrator terminal 5. The input keyword information may be transmitted from the administrator terminal 5 to the server device 2, thereby executing the processing from step S24 onwards.

[0058] In response to the execution of the specialized information update process, a keyword setting process may be executed, and in response to the execution of the specialized information update process, instruction information prompting the execution of the keyword setting process may be displayed on the administrator terminal 5. In addition, when the specialized information update process is executed, a process of deleting the keyword information stored in the server device 2 may be executed, and the processes of steps S21 to S25 may be executed.

[0059] [Output Processing] Next, a description will be given of output processing in which a user displays a predetermined answer using the system 10. Fig. 6 is a diagram showing a flowchart of the output processing according to this embodiment.

[0060] First, the user logs in to the system 10 from the user terminal 1. The operation of logging in to the system 10 is omitted here because it is similar to the operation of the administrator terminal 5 in the specialized information update process. It is preferable that the user registers attribute information about the user in the system 10 in advance, in association with the user ID.

[0061] When logging in to the system 10, the user terminal 1 displays a screen for inputting a question (step S31). In step S31, the system 10 controls the user terminal 1 to display options. At least one option is displayed on the display screen in step S31. One option corresponds to one keyword. Multiple options may be displayed on the display screen in step S31. Each of the multiple options corresponds to a different keyword. Details of the display screen will be described later. The user terminal 1 accepts input of one option from the options displayed on the screen (step S32).

[0062] The user terminal 1 transmits the input information input in step S32 to the server device 2 (step S33). The input information in step S33 includes information related to the options selected in step S32. The server device 2 receives the input information (step S34). Step S34 is a process of accepting the selection of one or more options from multiple options corresponding to each of multiple keywords in response to the input from the user terminal 1.

[0063] The server device 2 identifies an option to be displayed next (step S35). In step S35, the server device 2 refers to a database in which keywords are stored to identify an option to be displayed next. Specifically, the server device 2 identifies a keyword associated with the keyword corresponding to the option accepted in step S34. The identified keyword is the keyword corresponding to the option to be displayed next.

[0064] Next, the server device 2 generates a temporary prompt to be displayed on the user terminal 1 from the information about the options received in step S34 (step S36). In step S36, the temporary prompt is generated to generate a prompt to be input into the language model so as to include a keyword corresponding to the option selected in step S32. The method of generating the temporary prompt in step S36 is not particularly limited.

[0065] In step S36, the temporary prompt may be generated, for example, by embedding keywords corresponding to the options selected in step S32 into a preset template. The template may vary depending on the number of options selected and / or whether or not text information, described below, has been input. That is, in step S34, the template to be referenced is identified depending on the number of options selected on the user terminal 1. The server device 2 can generate the temporary prompt by inserting keywords corresponding to the options into predetermined positions in the template.

[0066] For example, examples of fixed phrases include "Tell me about (keyword corresponding to first option) in the long-term care insurance system," and "Tell me about (keyword corresponding to second option) in (keyword corresponding to first option) in long-term care insurance." The first option is the option selected first, and the second option is the option selected second.

[0067] Alternatively, in step S36, a method of generating a temporary prompt may be to set up a table in which temporary prompts are stored in advance in association with keywords corresponding to the options. The server device 2 may refer to the table to identify the temporary prompt corresponding to the keyword.

[0068] The server device 2 transmits option information related to the option identified in step S35 and the temporary prompt generated in step S36 to the user terminal 1 (step S37). The user terminal 1 receives the option information and the temporary prompt (step S38) and displays them on the screen (step S39). In step S39, the user terminal 1 is controlled to display the temporary prompt generated in step S36. The options displayed in step S39 are options corresponding to the keywords associated with the option selected in step S32.

[0069] The processing of steps S32 to S39 is repeated any number of times depending on the content of the user's question. In other words, the system 10 can accept the selection of one or more options from the options controlled to be displayed in steps S31 and S39. The number of times steps S32 to S39 are repeated is set in advance by operating the administrator terminal 5. It is preferable that the administrator set keyword items depending on the number of repetitions.

[0070] After the repeated processing of steps S32 to S39 is completed and the user selects an option corresponding to the keyword they wish to select, the user terminal 1 accepts input of text information through user operation (step S40). In step S40, it is preferable that a word is input as the text information. In step S40, information regarding the answer level and / or information regarding the output format of the answer may also be accepted. The user terminal 1 transmits the input information entered in step S40 to the server device 2 (step S41). The input information in step S41 includes information regarding the option selected in step S32 and the text information entered in step S40. The input information in step S41 may also include information regarding the answer level and / or information regarding the output format. The server device 2 receives the input information (step S42). Step S42 is a process of accepting input of text information in response to an input from the user terminal.

[0071] Next, the server device 2 generates a provisional prompt to be displayed on the user terminal 1 based on the keyword corresponding to the option selected in step S32 and the text information input in step S40 (step S43). The process of step S43 is executed in the same manner as the process of step S36.

[0072] The server device 2 transmits the temporary prompt generated in step S43 to the user terminal 1 (step S44). The user terminal 1 receives the temporary prompt (step S45) and displays the temporary prompt on the screen (step S46).

[0073] The user checks the temporary prompt, and if the user determines that the question he or she wants to ask is expressed correctly, operates the user terminal 1, which then transmits an answer request to the server device 2 (step S47). The answer request includes information about the temporary prompt. The answer request may include information about keywords included in the temporary prompt, text information, and information about the user's attributes. When the server device 2 receives the answer request (step S48), it executes an answer acquisition process to acquire answer information about the answer (step S49).

[0074] When the server device 2 acquires the answer information in step S49, it transmits the answer information to the user terminal 1 (step S50). The user terminal 1 receives the answer information (step S51). The answer information transmitted in step S50 is answer information corresponding to the answer request transmitted from the user terminal 1 in step S47. The user terminal 1 displays an answer on the screen based on the received answer information (step S52). The answer information acquired in step S49 and the answer information displayed in step S52 do not have to match. The answer information displayed in step S52 may be a part of the answer information acquired in step S49. The server device 2 stores the answer acquired in step S49 as history information in the storage unit 23 (step S53). The output process is completed by the processing of steps S31 to S53.

[0075] The history information in step S53 has the same meaning as the history information in step S15. Regarding step S53, storage of the history information may be executed only when answer information is acquired in step S70 or S91 described below. Furthermore, in step S53, the server device 2 may store information related to keywords corresponding to the selected options in the user terminal 1. The information related to the keywords may be stored in association with the date and time of selection. The server device 2 can calculate the number of times a predetermined keyword has been selected within a predetermined period.

[0076] After the processing of step S52, input of evaluation information for evaluating the answer may be executed in response to an operation on the user terminal 1. The answer information transmitted in step S50 may include information requesting the user to evaluate the answer. In step S52, information for the user to evaluate the answer is displayed on the display screen. The evaluation information may be a graded evaluation of whether the answer was satisfactory, or may be information that accepts comments about the answer in the form of text information. The evaluation information is stored in the storage unit 23 of the server device 2. The evaluation information may be stored in the storage unit 23 of the server device 2 in association with history information corresponding to the evaluation information. The evaluation information can be controlled to be displayed on the administrator terminal 5.

[0077] The order of the processes of steps S35 and S36 may be reversed. Furthermore, the processes of steps S35 and S36 and / or step S43 may be executed by the user terminal 1. In this case, the processes of steps S33, S34, S37 and S38 and / or steps S41, S42, S44 and S45 may be omitted.

[0078] The temporary prompt displayed in step S39 may be a prompt. In this case, steps S40 to S46, as well as step S63 (described later) or steps S83 to S85 (described later) are omitted. The system 10 may generate a prompt so as to include a keyword corresponding to the option selected in step S32. The temporary prompt displayed in step S46 may be a prompt. In this case, step S63 (described later) or steps S83 to S85 (described later) are omitted. The system 10 may generate a prompt based on the keyword corresponding to the option selected in step S32 and the text information, etc., input of which was accepted in step S40.

[0079] [Answer Acquisition Processing] Next, the answer acquisition processing in step S49 will be described. Fig. 7 is a flowchart of the answer acquisition processing according to this embodiment. In the answer acquisition processing shown in Fig. 7, technical information is acquired by the language model server 3.

[0080] First, the server device 2 determines whether information relating to the temporary prompt received in step S48 and a prompt that satisfies a predetermined condition is stored (step S61). As described above, information relating to the prompt is stored as history information in the server device 2. In step S61, the server device 2 compares the temporary prompt with information relating to the prompt included in the history information to determine whether information relating to the temporary prompt and a prompt that satisfies a predetermined condition is stored.

[0081] The "predetermined condition" in step S61 may be, for example, that the information about the temporary prompt and the prompt match or are similar. The information about the temporary prompt and the prompt to be compared may include the temporary prompt itself, keywords and / or text information contained in the temporary prompt, attributes of the user who entered the options, the answer level, the output format, etc.

[0082] A match between the provisional prompt and the information about the prompt means that the comparison target is a complete match between the provisional prompt and the information about the prompt. A similarity between the provisional prompt and the information about the prompt means that the comparison target is similar between the provisional prompt and the information about the prompt. For example, if the comparison target is keywords, the server device 2 may determine that the comparison target is similar if a predetermined threshold or more of keywords (e.g., two or more out of three) match. If the comparison target is the provisional prompt itself, the server device 2 may determine that the comparison target is similar if the vector similarity between the provisional prompt and the prompt is a predetermined threshold or more (e.g., 0.5 or more).

[0083] When the comparison targets are keywords and / or text information, the server device 2 may determine whether the order in which the keywords and / or text information were selected or entered matches, in addition to the content of the keywords and / or text information. In other words, the server device 2 may determine whether the content and / or order of the keywords and / or text information included in the provisional prompt matches or is similar to that of the keywords and / or text information included in the prompts stored in the history information.

[0084] If information about the temporary prompt and a prompt that satisfies the predetermined condition is stored (YES in step S61), the server device 2 acquires answer information by identifying an answer corresponding to the information about the prompt from the history information (step S62).On the other hand, if information about the temporary prompt and a prompt that satisfies the predetermined condition is not stored (NO in step S61), the server device 2 generates a prompt from the temporary prompt (step S63).

[0085] In step S63, the server device 2 generates a prompt by adding predetermined information to the provisional prompt. The predetermined information may be, for example, information about a keyword, answer level, and / or output format corresponding to an option not selected by the user in step S32 and / or step S40. Furthermore, the predetermined information may preferably be information about a prompt that is not stored in the server device 2 as history information.

[0086] In step S63, the server device 2 identifies information about a prompt that is set in the system 10 but is not stored in the server device 2. The server device 2 then generates a temporary prompt by connecting the identified information to information of the same type as the identified information using a selective conjunction. Alternatively, the server device 2 generates a temporary prompt by replacing the identified information with information of the same type as the identified information in the temporary prompt. The server device 2 then generates a prompt by combining the temporary prompt received in step S48 with the generated temporary prompt.

[0087] In addition, information of the same type may be, for example, in the case of keywords, keywords of the same attribute; in the case of answer levels, information set as an answer level; and in the case of output formats, information set as an output format.

[0088] For example, if the temporary prompt received in step S48 is "Please tell me about keyword B of keyword A. Please output in output format X," the following process is executed in step S63. If output formats X, Y, and Z are set in the system 10, the information about the keyword and output format included in the temporary prompt is "keyword A, keyword B, and output format X," so the server device 2 adds information about output format Y and output format Z to the temporary prompt.

[0089] The server device 2 generates a prompt such as, for example, "Please tell me about keyword B of keyword A. Please output in output format X, Y, or Z." Alternatively, the server device 2 generates a prompt such as, for example, "Please tell me about keyword B of keyword A. Please output in output format X. Please tell me about keyword B of keyword A. Please output in output format Y. Please tell me about keyword B of keyword A. Please output in output format Z."

[0090] The server device 2 transmits the prompt generated in step S63 to the language model server 3 (step S64). Upon receiving the prompt (step S65), the language model server 3 acquires specialized information as related information from the specialized information database 4 based on the prompt (step S66). In step S66, the received prompt is input into the language model, and the specialized information is output as a response. That is, in step S66, the language model can acquire related information from the specialized information database 4. The related information may be, for example, information including a keyword included in the prompt, or information similar to the prompt (for example, information having a value similar to a vectorized value of the prompt).

[0091] In step S66, the language model server 3 can use a known method for the process of acquiring related information from the specialized information database 4. For example, when the related information is information that includes a keyword included in the prompt, the language model server 3 may acquire information that includes the keyword included in the prompt from the specialized information database 4. Alternatively, the language model server 3 may acquire information that includes a keyword corresponding to the option selected in step S32 from the specialized information database 4.

[0092] Furthermore, for example, when the related information is information having a vectorized value that is similar to the vectorized value of the prompt, the language model server 3 can acquire the related information as follows. For example, the language model server 3 performs vector calculation on the prompt received in step S65 to vectorize the prompt. Furthermore, the specialized information stored in the specialized information database 4 has been vectorized by performing vector calculation. The server device 2 uses the vectorized prompt and the vectorized information in the specialized information database 4 to acquire information that provides a value that is similar to the value obtained from the vectorized prompt. In other words, when values ​​obtained from vectorized sentences are similar between text A and text B, text A and text B are determined to be related. The related information may also be similar vectorized values.

[0093] In the present embodiment, "approximate" means, for example, that the absolute value of the difference between the values ​​is equal to or less than a predetermined threshold (or less than the predetermined threshold). The predetermined threshold can be set as appropriate.

[0094] After acquiring the specialized information in step S66, the language model server 3 uses the specialized information and the prompt to output an answer from the language model (step S67). In step S67, the specialized information and the prompt are input to the language model, and an answer is output. The language model server 3 transmits answer information related to the output answer to the server device 2 (step S68). The server device 2 receives the answer information (step S69) and acquires the answer information (step S70).

[0095] Next, the server device 2 associates the answer information with information about the prompt (step S71). In step S63, the server device 2 generates a prompt that outputs multiple answers within one answer. Therefore, in step S71, the server device 2 executes a process of associating at least some of the acquired answers with information about the prompt corresponding to at least some of the answers. For example, the server device 2 associates a first part included in the answer with information about the prompt corresponding to the first part, and associates a second part included in the answer with information about the prompt corresponding to the second part.

[0096] The process of identifying which part of the answer acquired in step S70 relates to which keyword, answer level, and / or output format can be performed using a known method. For example, the system 10 may perform additional learning, such as fine-tuning, on the language model regarding how to output an answer. Alternatively, the system 10 may generate a prompt in step S63 to include a provisional prompt that specifies the output.

[0097] The answer information and the information about the prompt associated in step S70 are stored as history information in step S53. In other words, the system 10 can store multiple pieces of history information by outputting an answer from the language model once. The answer acquisition process is completed by the above steps S61 to S71.

[0098] Next, another example of the answer acquisition process in step S49 will be described. Fig. 8 is a flowchart showing another example of the answer acquisition process according to an embodiment of the present invention. The answer acquisition process shown in Fig. 8 differs from the answer acquisition process shown in Fig. 7 in that specialized information is acquired by the server device 2.

[0099] First, the server device 2 determines whether information relating to the temporary prompt received in step S48 and a prompt that satisfies a predetermined condition is stored (step S81). If information relating to the temporary prompt and a prompt that satisfies the predetermined condition is stored (YES in step S81), the server device 2 acquires answer information by identifying an answer corresponding to the information relating to the prompt from the history information (step S82). The processing in steps S81 and S82 is the same as that in steps S61 and S62, and therefore a description thereof will be omitted.

[0100] If the server device 2 does not store information related to the temporary prompt received in step S48 and a prompt that satisfies the predetermined conditions (NO in step S81), the server device 2 generates a temporary prompt (step S83). The process of step S83 can be the same as the process of step S63, except that the generated "prompt" is a "temporary prompt."

[0101] The server device 2 acquires specialized information as related information from the specialized information database 4 based on the temporary prompt generated in step S83 or the keyword corresponding to the option selected and accepted in step S32 (step S84).

[0102] In step S84, a known method can be used for the process in which the server device 2 acquires related information from the specialized information database 4. For example, the process in step S84 can be the same as the process in step S66 in which the language model server 3 acquires related information from the specialized information database 4, except that the "prompt" is replaced with a "temporary prompt" and the "language model server 3" is replaced with the "server device 2."

[0103] When the specialized information is acquired in step S84, the server device 2 generates a prompt based on the information acquired in step S84 and the keyword corresponding to the temporary prompt received in step S48 or the option selected in step S32 (step S85).

[0104] In step S85, the server device 2 can generate a prompt by adding the specialized information acquired in step S84 to the temporary prompt received in step S48. Also, in step S85, the server device 2 can generate a prompt to be input to the language model so as to include the keyword corresponding to the option selected in step S32 and the text information input in step S40.

[0105] The server device 2 transmits the prompt generated in step S85 to the language model server 3 (step S86). The language model server 3 receives the prompt (step S87). The language model server 3 outputs an answer by inputting the prompt into a language model (step S88). The language model server 3 transmits answer information regarding the output answer to the server device 2 (step S89). The server device 2 receives the answer information (step S90) and acquires the answer information (step S91).

[0106] The server device 2 associates the answer information with information about the prompt (step S92). The process of step S92 can be the same as the process of step S71. The process of steps S81 to S92 above completes the answer acquisition process.

[0107] In the above, the temporary prompts to be input to the language model are generated by the processing of steps S31 to S48, but the temporary prompts used in the answer acquisition process may be acquired from another computer device. That is, instead of the processing of steps S31 to S48, the answer acquisition process may be executed by acquiring temporary prompts input from another computer device.

[0108] Furthermore, the processing of step S63 or S83 may be omitted. In this case, the processing of step S71 or S92 is also omitted. Furthermore, in this case, the temporary prompt received in step S48 is used as the prompt in the answer acquisition processing shown in FIG. 7. The prompt used in the answer acquisition processing may be obtained from one input in another computer device.

[0109] The processes of steps S63 to S71 and S53, or steps S83 to S92 and S53, may be executed in response to an operation on the administrator terminal 5. For example, the above processes may be executed after the keyword setting process. At this time, answers corresponding to temporary prompts that combine the set keywords are stored as history information. When answers corresponding to all combinations of the set keywords, answer levels, and output formats have been stored as history information, the processes of steps S63 to S71 or steps S83 to S92 may be omitted in the output process executed in response to an operation on the user terminal 1.

[0110] In the above-described answer acquisition process, related specialized information is acquired in step S66 or S84, but step S66 or S84 may be omitted. If step S84 is omitted, step S85 is also omitted. In this case, the temporary prompt displayed in step S39 or S46 may be input as a prompt to the language model. The system 10 can acquire an answer from the language model based on the prompt generated in step S36 or S43 or a prompt generated from the temporary prompt generated in step S36 or S43.

[0111] [Explanation of Display Screen] Next, the display screen according to this embodiment will be explained. Figures 9 to 13 are diagrams showing examples of the display screen according to this embodiment. Figure 9 shows the display screen 100a in step S31.

[0112] The display screen 100a is a screen for a user to input a question to the system 10. The display screen 100a has a first option display area 101, a second option display area 102, a third option display area 103, a free word input area 104, and a generation prompt display area 110. The display screen 100a may also have an answer level selection area 105 and / or an output format selection area 106.

[0113] The first option display area 101 is an area for displaying the first option that the user initially selects. It is preferable that the first option is always selected by the user. A plurality of first options 30, 31, 32, 33, and 34 are displayed in the first option display area 101. In addition, first options 30a to 30j are displayed so as to correspond to the first option 30. Each of the keywords corresponding to the first options 30a to 30j is associated with the keyword corresponding to the first option 30.

[0114] Similarly, first options 31a to 31d corresponding to first option 31, first options 32a to 32c corresponding to first option 32, and first options 33a to 33c corresponding to first option 33 are displayed.

[0115] In the present embodiment, among the multiple options displayed, the keyword associated with the first option is different from the keyword associated with the second option. That is, on the display screen 100a, the keyword associated with the keyword "home care" corresponding to the first option 30 is different from the keyword "facility care" corresponding to the first option 31. Note that among the multiple options displayed, there may be options in which some or all of the keywords associated with the options overlap.

[0116] In this embodiment, each option is displayed in a manner that allows the corresponding keyword to be understood. In Figures 9, 10, and 12, text indicating the content of the keyword corresponding to the option is displayed in a rectangular frame.

[0117] The second option display area 102 is an area for displaying a second option that the user selects second. The third option display area 103 is an area for displaying a third option that the user selects third. On the display screen 100a, options corresponding to the second option display area 102 and the third option display area 103 are not displayed. The free word input area 104 is an area for the user to input text information (also called a free word) in step S40.

[0118] The answer level selection area 105 is an area for displaying options related to the answer level selected by the user. Level options 105a to 105d are displayed in the answer level selection area 105. The system 10 pre-sets the level options in correspondence with keywords related to the answer levels. The system 10 may also store temporary prompts corresponding to the keywords related to the answer levels. The keywords related to the answer levels may be, for example, keywords related to the level of the answer, such as "easy to understand," "concise," "without using technical terms," ​​"so that even an elementary school student can understand," or "explain as if an expert," or keywords related to the expression of the answer, such as "in the polite form" or "in the polite form."

[0119] The output format selection area 106 is an area for displaying options related to the output format selected by the user. The output format selection area 106 displays output format options 106a to 106d. The output format options 106a to 106d are options from which the user selects the output format of the answer in step S40. The system 10 pre-sets the output format options in correspondence with keywords related to the output format. The system 10 may also store temporary prompts corresponding to keywords related to the output format. Keywords related to the output format may be, for example, keywords related to the type of answer to be output, such as "bullet list," "table format," "FAQ format," or "unspecified," or keywords related to the number of characters to be output, such as "within 300 characters."

[0120] The generated prompt display area 110 is an area for displaying the temporary prompt generated in step S36 or S43.

[0121] The display screen 100a may have icons 121 to 124. When icon 121 is selected, an answer request is sent to the server device 2 in step S47. The temporary prompt displayed in the generated prompt display area 110 when icon 121 is selected becomes the temporary prompt included in the answer request. When icon 122 is selected, the content selected or displayed on the display screen 100a is deleted. When icon 123 is selected, the first option displayed in the first option display area 101 is deleted.

[0122] 10 shows the display screen 100b in step S39 after the first option is selected. In the following explanation, the contents explained on the display screen 100a will be omitted to avoid duplication.

[0123] Display screen 100b is a display screen that is displayed after first option 30a is selected on display screen 100a. In first option display area 101 on display screen 100b, "Home care," which corresponds to the selected first option 30a, is displayed. In second option display area 102 on display screen 100b, second options 40a to 40c are displayed. Furthermore, the first option is not displayed on display screen 100b.

[0124] The second options 40a to 40c each correspond to a keyword associated with the keyword "home care" corresponding to the first option 30a.

[0125] Also, the temporary prompt generated in step S36, "Tell me about home nursing care under the long-term care insurance system," is displayed in generated prompt display area 110. As shown in Fig. 10, the temporary prompt displayed in generated prompt display area 110 displays "home nursing care" in bold. In this embodiment, system 10 controls so that the keyword corresponding to the option selected in step S32 is displayed in a different format from the other text included in the generated temporary prompt.

[0126] The different manner is preferably a manner in which the keyword is displayed more prominently than other text, for example, by making the keyword bold, blinking, changing the color of the keyword, changing the size of the keyword, etc.

[0127] In this embodiment, the system 10 may perform control so that the temporary prompt displayed in the generated prompt display area 110 cannot be edited from the user terminal 1. Alternatively, the system 10 may perform control so that the temporary prompt displayed in the generated prompt display area 110 can be edited from the user terminal 1. The processing from step S47 onwards is performed using the temporary prompt edited by operating the user terminal 1.

[0128] When icon 121 is selected on display screen 100b, an answer corresponding to the provisional prompt displayed in generated prompt display area 110 is output. Specifically, when icon 121 is selected, the processes from step S47 onward are executed. Figure 11 corresponds to display screen 100c at step S52 after the answer has been output, with steps S40 to S46 omitted from display screen 100b shown in Figure 10.

[0129] The display screen 100 c has a condition display area 130 , a questionnaire input area 140 , and an answer display area 150 .

[0130] The condition display area 130 displays information about the conditions for outputting an answer. The condition display area 130 displays conditions 131 and a generation prompt 132. The condition 131 displays options selected by the user and / or text information entered by the user.

[0131] The generation prompt 132 displays the prompt that was input to the language model to output an answer. Note that the prompt displayed in the generation prompt 132 does not have to match the prompt that was actually input to the language model. The prompt displayed in the generation prompt 132 may be the same as the provisional prompt displayed in the generation prompt display area 110.

[0132] The questionnaire input area 140 is an area for inputting evaluation information. The questionnaire input area 140 is provided with a satisfaction level input area 141 and a reason input area 142. The user can operate the user terminal 1 to input the satisfaction level of the answer in the satisfaction level input area 141. The user can also operate the user terminal 1 to input the reason for the evaluation in the reason input area 142.

[0133] At least a portion of the answer information acquired in step S49 is displayed in the answer display area 150. Information that can identify the specialized information that was acquired from the specialized information database 4 and referenced when acquiring the answer may be displayed in the answer display area 150. The information that can identify the specialized information may be the specialized information itself, a URL to the specialized information, or a comment when the specialized information is displayed on the user terminal 1.

[0134] FIG. 12 shows the display screen 100d in step S46 after the first and second options have been selected and text information has been input.

[0135] Display screen 100d is a display screen that is displayed after selecting second option 40a on display screen 100b and then entering "problem" as a free word. A first option display area 101 on display screen 100d displays "home care," which corresponds to the selected first option 30a, and a second option display area 102 displays "physical care," which corresponds to the selected second option 40a. Furthermore, a third option display area 103 on display screen 100d displays third options 50a to 50f. Furthermore, display screen 100b does not display the first and second options.

[0136] The third options 50a to 50f each correspond to a keyword associated with the keyword "physical care" corresponding to the second option 40a.

[0137] Additionally, the generated prompt display area 110 displays the temporary prompt generated in step S43, "Tell me about the problems with physical care in home care under the long-term care insurance system." As shown in FIG. 12 , the temporary prompt displayed in the generated prompt display area 110 displays "home care," "physical care," and "problems" in bold. In this embodiment, the system 10 controls the display of the keywords and text information corresponding to the option selected in step S32 in a manner different from other text included in the generated temporary prompt. Preferably, the different manner is a manner in which the keywords and text information are displayed more prominently than the other text, as described above.

[0138] When icon 121 is selected on display screen 100d, an answer corresponding to the provisional prompt displayed in generation prompt display area 110 is output. Specifically, when icon 121 is selected, the processes from step S47 onward are executed. Figure 13 shows display screen 100e in step S52 after an answer request is sent from display screen 100d displayed in Figure 12 and an answer is output. The components displayed on display screen 100e are the same as those on display screen 100c. An answer corresponding to the prompt displayed in generation prompt 132 is displayed in answer display area 150 on display screen 100e.

[0139] [Display Order Change Processing] Next, a description will be given of a display order change processing for changing, in accordance with predetermined information, the area in which options are displayed on the user terminal 1. Fig. 14 is a diagram showing a flowchart of the display order change processing according to this embodiment.

[0140] The timing of execution of the display order change process is not particularly limited, but may be executed in response to a request from the administrator terminal 5, or may be executed at predetermined intervals (for example, one month). Furthermore, the display order change process may be executed for all options, or may be executed for a predetermined range of options. The predetermined range of options refers to options associated with the same keyword (for example, the first options 30a to 30j), options displayed on the same display screen, etc.

[0141] First, the server device 2 identifies the number of times each option has been selected during a predetermined period for which the display order change process is to be executed (step S91). The predetermined period is not particularly limited, but may be, for example, one month, one week, one day, etc. The predetermined period may be the period from the last time the display order change process was executed until the next time the process is executed.

[0142] Next, the server device 2 changes the area in which each option is displayed in accordance with the information relating to the specified number of times (step S92). The information relating to the number of times is, for example, the number of times the option was selected during a predetermined period. The server device 2 stores the changed display area in association with the option (step S93). The display order change process is completed by the processing of steps S91 to S93.

[0143] Changing the display area in step S93 is a concept that includes changing the display order on the same display screen, changing the options displayed on the first display screen to be displayed on the second display screen, causing the options displayed on the first display screen to be further displayed on the second display screen, and displaying further options.

[0144] For example, step S93 may change the position of the first option 30j to the position of the first option 30b on the display screen 100a. Furthermore, for example, step S93 may change the positions of the first options 33a to 33c associated with the first option 33 on the display screen 100a to positions higher than the first option 30. Furthermore, for example, step S93 may further display the first option 30b higher than the other first options on the display screen 100a. Furthermore, for example, step S93 may change the second option 40a that was displayed on the display screen 100b so that it is displayed on the display screen 100a. In this case, the second option 40a may be displayed near the first option 30a, and selecting the second option 40a on the display screen 100a may result in the selection of both the first option 30a and the second option 40a.

[0145] Note that the display mode of the options may be controlled to differ depending on the information regarding the number of times specified in step S91. For example, the server device 2 may display options that have been selected more frequently in a predetermined period in a more prominent manner than other options. Furthermore, the server device 2 may display options that have been selected more frequently in a predetermined period in red, options that have been selected in the middle order in blue, and options that have been selected less frequently in black. In this case, the server device 2 may or may not change the display area of ​​the options.

[0146] In this embodiment, the system 10 controls the display of at least two or more options in a list (see FIGS. 9, 10, and 12). By the processes of steps S91 to S93, the system 10 can control the display of each option in step S31, S39, or S46 by changing the area in which each option is displayed, depending on the information regarding the number of times the option was selected in step S32.

[0147] In this embodiment, the system 10 controls the display screen to display the next option each time an option is selected, but it may also be controlled to display, for example, the first option, the second option, and the third option on a single display screen.

[0148] The system 10 may also control the user terminal 1 to display search history information related to the search history searched by the user operating the user terminal 1. The search history information is stored by associating keywords corresponding to selected options, input text information, information about the answer level, information about the output format, the user ID, and / or the search date and time with one another. The search history information may be displayed in a list on the user terminal 1. Furthermore, by selecting one piece of search history information by operating the user terminal 1, the keywords, etc. corresponding to the search history information may be displayed in a selected state on the display screen 100a.

[0149] Note that the system 10 may control the user terminal 1 to display a screen for selecting a specialized information database 4 and / or a language model server 3 before step S31. The system 10 may display a search screen similar to a portal site before step S31. The specialized information database 4 and / or the language model server 3 may be constructed for each field in which a user asks a question, or for each product or service provided to the user. For example, the specialized information database 4 and / or the language model server 3 are associated with a field, product, or service. The user can select the specialized information database 4 and / or the language model server 3 to be used to output an answer in the system 10. The selection of the specialized information database 4 and / or the language model server 3 may be performed by selecting the field, product, or service. Furthermore, the system 10 may search for and extract a specific specialized information database 4 and / or language model server 3 on the screen for selecting the specialized information database 4 and / or the language model server 3 in response to an operation of the user terminal 1. The system 10 may also control the user terminal 1 to display a list of selectable predetermined specialized information databases 4 and / or language model servers 3 .

[0150] On the display screen of the user terminal 1, the specialized information database 4 is displayed in a manner that allows the information stored in the specialized information database 4 to be understood. For example, attributes of the information stored in the specialized information database 4 may be displayed. For example, information attributes such as "nursing care insurance field," "personnel recruitment information," "accounting information," and "customer service" are displayed on the user terminal 1. By selecting the attribute of the information about which the user wants to ask a question, the selected specialized information database 4 and / or language model server 3 are used to execute output processing and answer output processing.

[0151] Note that the process of accepting input of text information in step S40 does not necessarily have to be provided. If the process of accepting input of text information is not provided, the processes of steps S40 to S46 are omitted. Alternatively, options regarding the answer level and / or output format are input, and the processes of steps S40 to S46 are executed. In this case, the free word input area 104 is not displayed on the display screen 100. At this time, the user is limited to inputting only the options displayed on the display screen 100. This allows the system 10 to limit questions input into the language model to only keywords intended by the administrator who manages the options, and to prevent the language model from outputting answers regarding unrelated matters. In this way, the system can provide answers to the user by limiting the content of the question input by the user.

[0152] In this embodiment, the system 10 stores evaluation information. The administrator can display the evaluation information on the administrator terminal 5 and review the template phrases used to generate temporary prompts for answers with low evaluations. The administrator can also consider problems and areas for improvement based on the evaluation information.

[0153] In addition, the computer device or language model that extracts keyword information in step S23 of the keyword setting process, the computer device or language model that acquires specialized information in steps S66 and S84, and the computer device or language model that outputs answers in steps S67 and S88 may be different computer devices or language models.

[0154] Although the above describes an embodiment in which the specialized information database 4 is constructed and the specialized information database 4 is referenced to extract keyword information and acquire specialized information, the processes for extracting keyword information and acquiring specialized information are not limited to this. For example, in step S23, the language model may search one or more websites based on the prompt included in the keyword setting request or the prompt generated by the server device 2, and extract predetermined keywords from the websites. For example, in step S66 or step S84, the language model may search one or more websites based on the generated prompt, and acquire related information from the websites.

[0155] The above describes an aspect in which the system 10 supports a search for obtaining an answer related to specific information. However, the process of supporting the generation of prompts by selecting options in the system 10 is not limited to searches, and can be used to support the generation of prompts for a language model to produce various outputs. For example, the system 10 can be used to generate prompts for a language model to summarize or proofread a sentence, or to generate prompts for a language model to generate sentences or images. The sentences or images generated based on the language model may be, for example, materials such as proposal materials, analytical reports for marketing, or program code.

[0156] In this way, the system includes at least one computer device, a selection means for accepting a selection of one or more options from a plurality of options corresponding to each of a plurality of keywords in response to an input from a user terminal operated by a user, and a first generation means for generating a prompt to be input to a language model, or generating a provisional prompt for generating the prompt, so as to include a keyword corresponding to the option selected by the selection means. This can assist the user in inputting a prompt to be input to the language model. Furthermore, by preparing options corresponding to keywords in advance, the burden on the user when generating a prompt can be reduced. Furthermore, the user can obtain an output from the language model with simple input. For example, the user can obtain an answer to a question from the language model by inputting options.

[0157] Furthermore, by providing the system with a prompt display unit that controls the display of the prompt or provisional prompt generated by the first generating unit on the user terminal, the user can confirm the prompt or provisional prompt generated based on the selected option. Furthermore, if the prompt or provisional prompt is different from the user's intention, the user can reselect an option.

[0158] Furthermore, in this way, the system includes an option display means for controlling the user terminal to display options, the selection means accepts selection of one or more options from the options controlled to be displayed by the option display means, and when one option is selected from the multiple options by the selection means, the option display means controls to further display options corresponding to keywords associated with the selected one option, thereby making it possible to provide more detailed information about the content of the question the user asks the language model. Furthermore, in this way, among the multiple options displayed, the keywords associated with a first option and the keywords associated with a second option are different, making it possible to display the option intended by the user.

[0159] Furthermore, by controlling the prompt display means to display the keyword corresponding to the option selected by the selection means in a manner different from other text contained in the generated prompt or provisional prompt, the user can confirm that the keyword selected by the user is included in the prompt or provisional prompt.

[0160] In this way, the system is provided with an input means for accepting input of text information in response to input from the user terminal, and the first generation means generates the prompt or the provisional prompt based on the keyword corresponding to the option selected and accepted by the selection means and the text information input and accepted by the input means, thereby allowing the user to flexibly input the content of the question they want to ask in addition to input using pre-set options.

[0161] Furthermore, in this way, the first generation means generates the prompt or the provisional prompt by embedding keywords corresponding to the options selected by the selection means into a predetermined template, thereby reducing the processing burden on the computer device for generating the prompt or provisional prompt.

[0162] Furthermore, in this way, the system is equipped with an answer acquisition means for acquiring an answer from a language model based on the prompt generated by the first generation means or a prompt generated from a provisional prompt, thereby making it possible to output an answer from the language model.

[0163] Furthermore, in this way, the system includes an information acquisition means for acquiring related information from a specialized information database storing predetermined information based on the prompt generated by the first generation means or a prompt generated from a provisional prompt in a language model, and an answer acquisition means for acquiring an answer from the language model using the information acquired by the information acquisition means and the prompt, thereby making it possible to output an answer based on the predetermined information from the language model. Also, in this way, the system includes an information acquisition means for acquiring related information from a specialized information database storing predetermined information based on the generated provisional prompt or a keyword corresponding to an option selected by the system, a second generation means for generating a prompt based on the information acquired by the information acquisition means and the provisional prompt or the keyword, and an answer acquisition means for acquiring an answer from the language model based on the prompt generated by the second generation means, thereby making it possible to output an answer based on the predetermined information from the language model. This makes it possible to acquire an answer that includes information updated since the learning of the language model was completed.

[0164] In this way, the system is equipped with an extraction means that extracts specified keywords by having the language model refer to a specialized information database, and the selection means accepts the selection of an option from options corresponding to the keywords extracted by the extraction means, thereby reducing the burden on the administrator of setting keywords.

[0165] Furthermore, in this way, the system includes an option display means for controlling the user terminal to display options, a selection means for accepting the selection of one or more options from the options controlled to be displayed by the option display means, and the option display means for controlling the display of at least two or more options in a list, and the area for displaying each option is controlled to be changed according to information regarding the number of times the option has been selected by the selection means, thereby making it possible to change the area for displaying options according to the selection history of the option.Furthermore, options that are frequently selected during a search can be changed to be displayed in an area that is easy for the user to select.

[0166] In this way, the system includes at least one computer device, a first acquisition means for acquiring an answer from a language model based on a prompt, a storage means for storing the answer acquired by the first acquisition means in association with information related to the prompt corresponding to the answer, a first generation means for generating or acquiring a prompt to be input to the language model or a provisional prompt for generating the prompt, and an identification means for identifying an answer corresponding to the information related to the prompt from the storage means when the prompt or provisional prompt generated or acquired by the first generation means and information related to a prompt that satisfies a predetermined condition are stored. This reduces the number of times an answer is output from the language model. Furthermore, reducing the number of times the language model is used can speed up the response time for outputting an answer. Furthermore, reducing the number of times the language model is used can reduce usage fees for using the language model.

[0167] Furthermore, in this way, when the system does not store information about the prompt or provisional prompt generated or acquired by the first generation means and a prompt that satisfies a specified condition, the system is provided with a second acquisition means that acquires an answer from a language model based on a prompt generated from the prompt or provisional prompt, thereby making it possible to output an answer to the user even when the answer is not stored in the system.

[0168] Furthermore, the system may include a second generation means for generating a prompt by adding predetermined information to the provisional prompt, a second acquisition means for acquiring answers from a language model based on the prompt generated by the second generation means, and a storage means for storing at least some of the answers acquired by the second acquisition means in association with information on the prompt corresponding to at least some of the answers, thereby dividing the answers output from the language model and storing them as answers corresponding to multiple questions. This allows the system to store multiple answers by outputting a single answer, thereby reducing the number of times answers need to be output from the language model.

[0169] In this way, the system is equipped with an information acquisition means that acquires related information from a specialized information database in which specified information is stored based on a prompt, and a first acquisition means that acquires an answer from a language model using the related information and the prompt, and is equipped with an information update means that adds, replaces, and / or deletes specified information from the specialized information database, and a deletion means that deletes information about the answer stored in the storage means and / or the prompt corresponding to the answer in response to the addition, replacement, and / or deletion of specified information by the information update means, thereby making it possible to prevent answers output based on past specified information from being provided to the user.

[0170] 1 User terminal 2 Server device 3 Language model server 4 Specialized information database 5 Administrator terminal 6 Communication network 10 System 11 Control unit 12 RAM 13 Storage unit 14 Input unit 15 Display unit 16 Communication interface 21 Control unit 22 RAM 23 Storage unit 24 Communication interface 101 First option display area 102 Second option display area 103 Third option display area 104 Free word input area 105 Answer level selection area 105a to 105d Level options 106 Output format selection area 106a to 106d Output format options 110 Generation prompt display area 121 to 124 Icons 130 Condition display area 131 Conditions 132 Generation prompt 140 Questionnaire input area 141 Satisfaction level input area 142 Reason input area 150 Answer display area

Claims

1. A system comprising at least one computer device, comprising: a selection means for accepting the selection of one or more options from a plurality of options corresponding to each of a plurality of keywords in response to input from a user terminal operated by a user; and a first generation means for generating a prompt to be input into a language model so as to include a keyword corresponding to the option selected by the selection means, or for generating a provisional prompt for generating the prompt.

2. The system according to claim 1, further comprising prompt display means for controlling the user terminal to display the prompt or provisional prompt generated by the first generation means.

3. The system according to claim 1 or 2, further comprising: an option display means for controlling the display of options in the user terminal; a selection means for accepting the selection of one or more options from the options controlled to be displayed by the option display means; and when one option is selected from the plurality of options by the selection means, the option display means for controlling the display of options to further display options corresponding to keywords associated with the selected one option.

4. The system according to claim 3, wherein a keyword associated with a first option and a keyword associated with a second option of the displayed multiple options are different.

5. The system according to claim 2, wherein the prompt display means controls so that the keyword corresponding to the option selected by the selection means is displayed in a manner different from other text included in the generated prompt or provisional prompt.

6. A system as described in claim 1 or 2, further comprising an input means for accepting input of text information in response to input from the user terminal, wherein a first generation means generates the prompt or the provisional prompt based on keywords corresponding to options selected by the selection means and the text information input by the input means.

7. A system as described in claim 1 or 2, wherein the first generating means generates the prompt or the provisional prompt by embedding a keyword corresponding to the option selected by the selecting means into a predetermined template.

8. The system according to claim 1 or 2, further comprising: answer acquisition means for acquiring an answer from a language model based on the prompt generated by the first generation means or a prompt generated from a provisional prompt.

9. The system according to claim 1 or 2, further comprising: an information acquisition means for acquiring related information from a specialized information database storing predetermined information based on the prompt generated by the first generation means or a prompt generated from a provisional prompt in the language model; and an answer acquisition means for acquiring an answer from the language model using the information acquired by the information acquisition means and the prompt.

10. The system described in claim 1 or 2, comprising: an information acquisition means for acquiring related information from a specialized information database in which specified information is stored, based on the generated temporary prompt or a keyword corresponding to an option for which selection has been accepted; a second generation means for generating a prompt based on the information acquired by the information acquisition means and the temporary prompt or the keyword; and an answer acquisition means for acquiring an answer from a language model based on the prompt generated by the second generation means.

11. A system as claimed in claim 1 or 2, comprising an extraction means for extracting predetermined keywords by referring to a specialized information database in which predetermined information is stored using a language model, and a selection means for accepting a selection of an option from options corresponding to the keyword extracted by the extraction means.

12. A system as described in claim 1 or 2, comprising an option display means for controlling the display of options in the user terminal, wherein the selection means accepts the selection of one or more options from the options controlled to be displayed by the option display means, and the option display means controls to display at least two or more options in a list, and controls to change and display the area displaying each of the options according to information regarding the number of times the option has been selected by the selection means.

13. A computer device comprising: a selection means for accepting the selection of one or more options from a plurality of options corresponding to each of a plurality of keywords in response to input from a user terminal operated by a user; and a first generation means for generating a prompt to be input into a language model so as to include a keyword corresponding to the option selected by the selection means, or for generating a provisional prompt for generating the prompt.

14. A method executed on at least one computer device, comprising: a selection step of accepting the selection of one or more options from a plurality of options corresponding to each of a plurality of keywords in response to input from a user terminal operated by a user; and a first generation step of generating a prompt to be input into a language model, or generating a provisional prompt for generating the prompt, so that the prompt includes keywords corresponding to the options selected in the selection step.

15. A system comprising at least one computer device, comprising: a first acquisition means for acquiring an answer from a language model based on a prompt; a storage means for storing the answer acquired by the first acquisition means in association with information related to the prompt corresponding to the answer; a first generation means for generating or acquiring a prompt to be input to the language model or a provisional prompt for generating the prompt; and an identification means for identifying an answer corresponding to the information related to the prompt from the storage means when the prompt or provisional prompt generated or acquired by the first generation means and information related to a prompt that satisfies predetermined conditions are stored.

16. The system according to claim 15, further comprising: a second acquisition means for acquiring an answer from a language model based on a prompt or provisional prompt generated or acquired by the first generation means and a prompt generated from the provisional prompt when information about a prompt satisfying a predetermined condition is not stored.

17. The system described in claim 16, further comprising: a second generation means for generating a prompt by adding predetermined information to the provisional prompt; a second acquisition means for acquiring answers from a language model based on the prompt generated by the second generation means; and a storage means for storing at least some of the answers acquired by the second acquisition means in association with information about the prompt corresponding to at least some of the answers.

18. A system according to any one of claims 15 to 17, comprising: an information acquisition means for acquiring related information from a specialized information database in which specified information is stored based on a prompt; a first acquisition means for acquiring an answer from a language model using the related information and the prompt; an information update means for adding, replacing, and / or deleting specified information from the specialized information database; and a deletion means for deleting information related to an answer and / or a prompt corresponding to the answer stored in the storage means in response to the addition, replacement, and / or deletion of specified information by the information update means.

19. A computer device comprising: a first acquisition means for acquiring an answer from a language model based on a prompt; a storage means for storing the answer acquired by the first acquisition means in association with information related to the prompt corresponding to the answer; a first generation means for generating or acquiring a prompt to be input into the language model or a provisional prompt for generating the prompt; and an identification means for identifying an answer corresponding to the information related to the prompt when the prompt or provisional prompt generated or acquired by the first generation means and information related to a prompt that satisfies predetermined conditions are stored in the storage means.

20. A method executed on at least one computer device, comprising: a first acquisition step of acquiring an answer from a language model based on a prompt; a storage step of associating and storing the answer acquired in the first acquisition step with information related to the prompt corresponding to the answer; a first generation step of generating or acquiring a prompt to be input into the language model or a provisional prompt for generating the prompt; and an identification step of identifying an answer corresponding to the information related to the prompt when the prompt or provisional prompt generated or acquired in the first generation step and information related to a prompt that satisfies predetermined conditions are stored in the storage step.

Images