Information processing system, information processing method, and program

Abstract

To provide an information processing system and so on which can reduce a work load upon machine learning.SOLUTION: The present invention is directed to an information processing system having at least one processor according to an embodiment thereof. In the information processing system, the processor acquires, in its history acquiring step, acquires a response history indicating an inquiry from a user and a response to the inquiry from a person in charge. In a learning instruction step, the processor gives an artificial intelligence module instructions to carry out machine learning using at least learning data using the inquiry indicated by the acquired response history as an input and the response indicated by the response history as an output.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing system, an information processing method, and a program. [Background technology]

[0002] Patent Document 1 discloses a technology for extracting one or more answer candidate sentences from multiple answer candidate sentences using a trained model trained using positive example training answer sentences to a training query sentence and negative example training answer sentences sampled according to the frequency distribution of the training answer sentences. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-124824 Summary of the Invention [Problem to be solved by the invention]

[0004] Creating query and response data for machine learning to realize a query response function using an artificial intelligence module requires a large workload.

[0005] In view of the above circumstances, the present invention provides an information processing system and the like that can reduce the workload during machine learning. [Means for solving the problem]

[0006] According to one aspect of the present invention, there is provided an information processing system including at least one processor. In this information processing system, the processor acquires a response history indicating inquiries from users and responses to the inquiries by personnel in a history acquisition step. In a learning instruction step, the processor instructs an artificial intelligence module to perform machine learning using at least learning data in which the inquiries indicated in the acquired response history are used as input and the responses indicated in the response history are used as output.

[0007] According to this aspect, the workload during machine learning can be reduced. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a diagram illustrating an example of the overall configuration of a response support system 1. FIG. [Figure 2] 2 is a diagram illustrating an example of a hardware configuration of a server device 10. FIG. [Figure 3] FIG. 2 is a diagram illustrating an example of a hardware configuration of a person in charge terminal 30. [Figure 4] FIG. 10 is an activity diagram illustrating an example of a history accumulation process. [Figure 5] FIG. 10 is a diagram illustrating an example of a displayed person-in-charge screen. [Figure 6] FIG. 2 is a diagram illustrating an example of a response history database DB1. [Figure 7] FIG. 10 is a diagram illustrating an example of a displayed response history screen. [Figure 8] FIG. 10 is an activity diagram illustrating an example of machine learning processing. [Figure 9] FIG. 2 is a diagram illustrating an example of a template database DB2. [Figure 10] FIG. 10 illustrates an example of exclusion processing. [Figure 11] FIG. 10 is an activity diagram illustrating an example of a draft generation process. [Figure 12] FIG. 10 illustrates an example of an addition process. [Figure 13] FIG. 10 is a diagram showing an example of a displayed answer draft. [Figure 14] FIG. 10 is an activity diagram illustrating an example of an additional learning process. [Figure 15] FIG. 2 is a diagram illustrating an example of a product database DB3. [Figure 16] FIG. 10 is an activity diagram showing another example of the additional learning process. [Figure 17] FIG. 10 is a diagram illustrating an example of a weight table. [Figure 18]FIG. 10 is a diagram illustrating another example of a weight table. DETAILED DESCRIPTION OF THE INVENTION

[0009] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the accompanying drawings. Various features shown in the following embodiments can be combined with each other.

[0010] Incidentally, the program for realizing the software appearing in this embodiment may be provided as a non-transitory computer-readable medium, or may be provided so that it can be downloaded from an external server, or may be provided so that the program is started on an external computer and its functions are realized on a client terminal (so-called cloud computing).

[0011] In this embodiment, the term "unit" may also include, for example, a combination of hardware resources implemented by a circuit in the broad sense and software information processing that can be specifically realized by these hardware resources. In addition, various types of information are handled in this embodiment, and this information may be represented by, for example, physical values of signal values representing voltages and currents, high and low signal values as a group of binary bits consisting of 0 or 1, or quantum superposition (so-called quantum bits), and communication and calculations may be performed on a circuit in the broad sense.

[0012] In addition, a circuit in the broad sense is a circuit realized by at least appropriately combining a circuit, circuitry, a processor, a memory, etc. That is, it includes an application specific integrated circuit (ASIC), a programmable logic device (e.g., a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)), etc.

[0013] <Embodiment 1> 1. System Configuration The system configuration according to the first embodiment will be described below. Fig. 1 is a diagram showing an example of the overall configuration of a response support system 1. Fig. 1 shows an overview of each device included in the response support system 1 and the users who use those devices. Each overview will be explained as needed, with reference to other figures.

[0014] The response support system 1 is an information processing system that executes information processing such as response support processing that supports tasks of responding to inquiries from users. In the following, when simply referring to an "inquiry," it refers to an inquiry from a user. An inquiry is, for example, an inquiry about a commercial product. Commercial products include goods and services. In this case, the user is a purchaser or user of the commercial product. It is assumed that the commercial product has an inquiry desk (also called customer service, support desk, help desk, etc.) that responds to inquiries. In addition to inquiries about commercial products, inquiries may also include inquiries about stores, inquiries about delivery dates, inquiries about procedures, etc.

[0015] The response support system 1 includes a communication line 2, a server device 10, an AI device 20, a staff member terminal 30, and an administrator terminal 40. The communication line 2 is not particularly limited, but may be configured, for example, by the Internet. The communication line 2 may also include a local area network, a mobile communication network, a VPN (Virtual Private Network), etc. The communication line 2 mediates the exchange of data between devices connected to the line. In the example of FIG. 1, the server device 10 and the AI device 20 are connected to the communication line 2 by wire, and the staff member terminal 30 and the administrator terminal 40 are connected wirelessly. Note that the connection of each device to the communication line 2 may be wired or wireless.

[0016] The server device 10 is an information processing device that executes information processing such as response support processing. The server device 10 stores a response history database DB1, a template database DB2, and a product database DB3. The response history database DB1 stores a history of responses to inquiries. The template database DB2 stores templates of sentences used in writing responses to inquiries. The product database DB3 stores detailed information about products to be provided to users.

[0017] The server device 10 also receives inquiries from users and executes a process of transmitting a response to the user who made the inquiry. The following describes a case where the server device 10 functions as an email server and inquiries and responses are made by email.

[0018] The AI device 20 is an information processing device that executes information processing using AI (Artificial Intelligence) technology. The AI device 20 includes an artificial intelligence module 200. The artificial intelligence module 200 is a module that has been adjusted (tuned) to realize a predetermined function using AI technology.

[0019] The artificial intelligence module 200 has a natural language processing model whose accuracy has been improved by machine learning using a large-scale data set called LLM (Large Language Models), for example, and realizes a sentence generation function that can generate natural sentences. The artificial intelligence module 200 uses this sentence generation function to generate a draft of a sentence that answers a query (hereinafter referred to as an "answer draft").

[0020] The staff member terminal 30 is a terminal whose user is an inquiry staff member who responds to inquiries, and is, for example, a personal computer. The staff member terminal 30 displays a screen for responding to inquiries and accepts operations from the inquiry staff. The administrator terminal 40 is a terminal whose user is a system administrator who manages the response support system 1, and is, for example, a personal computer. The administrator terminal 40 displays a management screen for managing the system and accepts operations from the system administrator.

[0021] The server device 10 executes a display process for displaying images on the person in charge terminal 30 and the administrator terminal 40, and an authentication process for authenticating users (inquiry persons and system administrators) who use the person in charge terminal 30 and the administrator terminal 40.

[0022] The server device 10 performs processes such as generating and transmitting an HTML (Hyper Text Markup Language) file as display processing, and displays a web page showing a system screen on the person in charge terminal 30 and the administrator terminal 40. Note that an application program for using the response support system 1 may be installed in the person in charge terminal 30 and the administrator terminal 40, and the server device 10 may perform processes such as generating and transmitting display data in that application as display processing. The server device 10 controls the display on the person in charge terminal 30 and the administrator terminal 40 by performing these display processes.

[0023] The server device 10 stores authentication information (user ID, password, etc.) for authenticating users who use the response support system 1, such as inquiry personnel and system administrators, and authenticates users who input the authentication information. By authenticating users, the server device 10 can restrict access to data and assign identification information to data entered by users to make the data identifiable.

[0024] 2. Hardware Configuration The hardware configuration according to the first embodiment will be described below. 2 is a diagram showing an example of the hardware configuration of server device 10. Server device 10 includes a control unit 11, a storage unit 12, a communication unit 13, and a bus 14. Bus 14 electrically connects the various units included in server device 10.

[0025] (Control unit 11) The control unit 11 has at least one processor. The at least one processor may be configured by, for example, a central processing unit (CPU), a micro processing unit (MPU), a graphics processing unit (GPU), one or more integrated circuits, one or more discrete circuits, or a combination thereof (not shown).

[0026] The control unit 11 is a computer that realizes various functions related to the response assistance system 1 by reading out predetermined programs stored in the storage unit 12. In other words, information processing by software stored in the storage unit 12 is specifically realized by the control unit 11, which is an example of hardware, and can be executed as each functional unit included in the control unit 11. Note that the control unit 11 is not limited to being a single unit, and it may be implemented with multiple control units 11 for each function. It may also be a combination of these.

[0027] (Storage unit 12) The memory unit 12 stores various pieces of information defined above. This can be implemented, for example, as a storage device such as a solid state drive (SSD) or a hard disk drive (HDD) that stores various programs and the like related to the response assistance system 1 executed by the control unit 11, or as a memory such as a random access memory (RAM) that stores temporarily required information (arguments, arrays, etc.) related to the program calculations. The memory unit 12 stores various programs, variables, etc. related to the response assistance system 1 executed by the control unit 11.

[0028] (Communications Department 13) The communication unit 13 is configured by a communication module. The communication module may be a wireless communication module conforming to standards such as IEEE802.11a / b / g / n / ac / ax, LTE, 5G, or 6G, or may be a wired communication module conforming to standards such as IEEE802.3. The communication unit 13 is configured to be able to transmit various electrical signals from the server device 10 to external components. The communication unit 13 is also configured to be able to receive various electrical signals from the external components to the server device 10. More preferably, the communication unit 13 has a network communication function, which allows various information to be communicated between the server device 10 and external devices via the communication line 2.

[0029] 2 has the same hardware configuration as the server device 10. In the following description of the AI device 20, only the control unit 21 is assigned a different reference numeral from the control unit 11 of the server device 10.

[0030] 3 is a diagram showing an example of the hardware configuration of the person in charge terminal 30. The person in charge terminal 30 includes a control unit 31, a memory unit 32, a communication unit 33, an input unit 34, an output unit 35, and a bus 36. The bus 36 electrically connects the various units included in the person in charge terminal 30. The control unit 31, the memory unit 32, and the communication unit 33 are similar hardware to the control unit 11, the memory unit 12, and the communication unit 13 shown in FIG. 2, although the specifications, model, etc. may be different.

[0031] (Input unit 34) The input unit 34 has keys, buttons, a touch screen, a mouse, etc., and receives input from the user. The input unit 34 may also have a microphone and have the function of receiving voice input from the user.

[0032] (Output unit 35) The output unit 35 has a display, a speaker, etc., and displays visual information generated in a manner that is visible to the user, such as a screen, an image, an icon, or text, on the display surface of the display, and outputs sound including voice.

[0033] 3 has the same hardware configuration as the person in charge terminal 30. In the explanation of the manager terminal 40, only the control unit 41 is assigned a different reference numeral from the control unit 31 of the person in charge terminal 30.

[0034] 3. Information Processing The following describes the information processing according to the embodiment. In the following description, the server device 10, the AI device 20, the person in charge terminal 30, and the manager terminal 40 are described as the subjects of each information processing, but the information processing is executed by at least one processor included in the response support system 1, i.e., the processor included in the control unit of each device.

[0035] The response support system 1 executes the above-described response support process. The response support process includes a history accumulation process, a machine learning process, a draft generation process, and an additional learning process. The history accumulation process is a process for accumulating a history of responses to inquiries from users.

[0036] Fig. 4 is an activity diagram showing an example of history accumulation processing. The history accumulation processing shown in Fig. 4 is started when the server device 10 receives an inquiry email from a user. First, the server device 10 receives the inquiry email from the user and transmits screen data showing a person in charge screen representing the received inquiry email to the person in charge terminal 30 (activity A11). The person in charge terminal 30 displays the person in charge screen shown by the transmitted screen data (activity A12).

[0037] Fig. 5 is a diagram showing an example of a displayed person in charge screen. The person in charge screen C1 shown in Fig. 5 displays the character string "Please respond to the inquiry," a display field D11 for the inquiry source, a display field D12 for the subject, a display field D13 for the inquiry content, a field D14 for inputting an answer, a response history display button B11, a create answer draft button B12, and a send button B13.

[0038] Display field D11 displays the user name or email address of the user who made the inquiry. In FIG. 5, the user name "User A" is displayed. Display field D12 displays the subject of the inquiry. In FIG. 5, the subject "Regarding Product α" is displayed. Display field D13 displays the body of the email as the content of the inquiry. Input field D14 is an area where the person in charge of the inquiry can enter a response to the inquiry, and FIG. 5 shows a state in which no response has yet been entered.

[0039] The response history display button B11 is an operation image for displaying the response history to past inquiries. The answer draft generation button B12 is an operation image for instructing the server device 10 to generate an answer draft. The generation of an answer draft will be described later. The send button B13 is an operation image for sending the answer entered by the inquiry person in the input field D14 to the user who made the inquiry.

[0040] 4, a case will be described where the response history display button B11 is operated. When the response history display button B11 is operated, the staff member terminal 30 accepts it as an operation to display the response history screen, and transmits instruction data indicating the inquiry content (inquiry source, inquiry subject, and inquiry text) and an instruction to display the response history to the server device 10 (activity A13). Upon receiving the instruction data, the server device 10 reads the response history from the response history database DB1 in accordance with the instruction indicated in the received instruction data.

[0041] Fig. 6 is a diagram showing an example of the response history database DB1. The response history database DB1 shown in Fig. 6 stores categories, inquiry personnel, inquiry emails, inquiry dates and times, response emails, response dates and times, response types, reference history, and the like, in association with one another. "Category" stores items that share common knowledge and know-how for responding to inquiries, and in the example of Fig. 6, the names of the products that are the subject of inquiries (such as "product α" and "product β"). In addition to products, categories also store "store," "delivery date," or "procedures," etc.

[0042] "Inquiry Person" stores the identification information (person name or user ID) of the person in charge of the inquiry who responded to the inquiry. "Inquiry Email" stores the email indicating the content of the inquiry sent by the user. "Inquiry Date and Time" stores the date and time the inquiry email was received. "Response Email" stores the email indicating the content of the response sent to the user by the person in charge of the inquiry. "Response Date and Time" stores the date and time the response email was sent.

[0043] "Answer type" stores the type of answer. Answer types are defined to classify similar answers, and examples of stored types include "first answer," "apology," "thanks," and "resolved." Two or more types may be stored as answer types. "Referenced history" stores the history of response history (inquiry emails, response emails, etc.) being referenced by the inquiry person. Referencing the response history means that the inquiry person refers to the response history as a reference when writing an answer, and this will be explained in more detail later with reference to Figure 7. In the example of Figure 6, the number of times the response history has been referenced (hereinafter referred to as "reference count") is stored as the referenced history.

[0044] When reading the response history, the server device 10 first determines the category of the inquiry and the type of response required for the inquiry from the inquiry content indicated by the instruction data (activity A14). For example, if the inquiry content contains a specific keyword, the server device 10 determines the category corresponding to the keyword. For example, if the inquiry content (particularly the subject and reply email) contains the name of a product, the server device 10 determines that product as the category of the inquiry.

[0045] Furthermore, the server device 10 determines the category as "store" if the inquiry contains keywords that are often used when searching for a store, such as "store," "location," or "where," determines the category as "delivery date" if the inquiry contains keywords that are often used when asking about delivery dates, such as "when," "what day," or "pick up," and determines the category as "procedure" if the inquiry contains keywords that are often used in procedures, such as "procedure," "selection," or "decision." Note that multiple categories may be determined for one inquiry.

[0046] The server device 10 also determines the response type from the inquiry content as follows: For example, if the inquiry is from a source that has not existed in the past, the server device 10 determines the response type to be "initial response." If the inquiry content contains keywords indicating a malfunction or complaint, such as "initial defect," "breakdown," and "not working," the server device 10 determines the response type to be "apology," and if the inquiry content does not contain these keywords, the server device 10 determines the response type to be "thanks" (including thanks for the inquiry). If the inquiry content contains keywords indicating the end of the inquiry, such as "resolved," "resolved," and "thank you," the server device 10 determines the response type to be "resolved."

[0047] The server device 10 reads, for example, the inquiry email, inquiry date and time, reply email, and reply date and time as a response history from the response history database DB1 (activity A15). At that time, the server device 10 reads out response history whose category and reply type determined in activity A14 are the same as the inquiry content. Next, the server device 10 generates a response history screen that displays the read response history so that it can be referenced, and transmits screen data showing the generated response history screen to the staff terminal 30 (activity A16). The staff terminal 30 displays the response history screen indicated by the transmitted screen data (activity A17).

[0048] Fig. 7 is a diagram showing an example of a displayed response history screen. In addition to the display fields and buttons shown in Fig. 5, the response history screen C2 shown in Fig. 7 also displays a response history display field D21 and a response history switching button B21. The display field D21 displays the response history (inquiry email, inquiry date and time, reply email, and reply date and time) read out in activity A15. When the switching button B21 is operated, the agent terminal 30 switches the response history to be displayed in the display field D21 (activity A21).

[0049] Switching of response histories may be performed by transmitting multiple response histories to the person in charge terminal 30 in advance, or by transmitting the next response history from the server device 10 each time a switch is made. When the response history is switched, the person in charge terminal 30 notifies the server device 10 of the displayed response history. Note that even if a response history is displayed, if it is determined that there is no point in referring to it, the person in charge terminal 30 immediately switches to the next response history, and therefore the person in charge terminal 30 may notify the server device 10 of only those response histories that have been displayed for a time exceeding a threshold as response histories that have actually been referred to.

[0050] The server device 10 updates the reference history of the notified response history in the response history database DB1 (in the example of FIG. 6, the number of references is incremented by one) (activity A22). Note that the reference history is not limited to the number of references, and the reference time, reference date, identification information of the person in charge of the referenced inquiry, or the number of characters actually used in the text of the response (which can be determined by comparing the text) may also be stored in the response history database DB1 as the reference history.

[0051] The person in charge of the inquiry inputs a response to the inquiry in the input field D14 while referring to the displayed response history. The person in charge terminal 30 accepts the input of the response by the person in charge of the inquiry (activity A23). After the person in charge of the inquiry has finished inputting the response, the person in charge operates the send button B13. When the send button B13 is operated, the person in charge terminal 30 transmits the response input in the input field D14 to the user who made the inquiry (activity A24).

[0052] When the staff member terminal 30 transmits the response, it generates response information indicating the response based on the transmitted response (activity A25). The staff member terminal 30 generates, as response information, information including, for example, the identification information of the person in charge of the inquiry, the inquiry email, the inquiry time, the response email, and the response time. The staff member terminal 30 transmits the generated response information to the server device 10. Based on the transmitted response information, the server device 10 determines the type of response in the same manner as above (activity A26). The server device 10 stores and saves the determination result together with the response information as a response history in the response history database DB1 (activity A27).

[0053] The history accumulation process is executed as described above. Next, a machine learning process is executed in which the artificial intelligence module 200 performs machine learning using the response history accumulated by the history accumulation process. Fig. 8 is an activity diagram showing an example of machine learning processing. The machine learning processing shown in Fig. 8 is started when the system administrator performs an operation (learning start operation) to instruct the administrator terminal 40 to start machine learning. The machine learning referred to here is assumed to be initial machine learning that is performed in a state where a learning model has not yet been generated.

[0054] The server device 10 receives an instruction to start machine learning through a learning start operation (activity A31). Next, the server device 10 acquires response histories from the response history database DB1 (activity A32). In the example of FIG. 8, the server device 10 acquires all response histories stored in the response history database DB1. Next, the server device 10 performs a process of excluding specific entries (also referred to as "specific entries") from the responses to the inquiries indicated by the acquired response histories.

[0055] The specific description items are items that are predetermined as items that are inappropriate as training data to be used in machine learning by the artificial intelligence module 200 that generates the answer draft. The specific description items are, for example, standard phrases, signatures, advertisements, etc. that are included in every answer. The following describes the case where standard phrases are excluded as the specific description items.

[0056] The fixed phrases include, for example, a greeting (such as "Thank you for your continued support") and a self-introduction to be written at the beginning of a reply, and a closing statement (such as "Thank you in advance") to be written at the end of a reply. These specific items to be written are, for example, created in advance as templates by a system administrator and stored in the template database DB2. The server device 10 reads the templates from the template database DB2 (activity A33).

[0057] Fig. 9 is a diagram showing an example of the template database DB2. In the template database DB2 shown in Fig. 9, answer types and templates are stored in association with each other. In addition to the answer types explained in Fig. 6, answer types include "second time or later" and "additional information". Since commonly used fixed phrases may differ depending on the answer type, a template is created for each answer type.

[0058] For example, the answer type "first reply" is associated with the templates "(beginning of the sentence) Thank you for your inquiry," "(end of the sentence) Please feel free to contact us if you have any other questions," and "(end of the sentence) Thank you in advance." "(beginning of the sentence)" indicates a sentence used at the beginning of an answer, and "(end of the sentence)" indicates a sentence used at the end of an answer. Furthermore, the answer type "second time or later" is associated with the templates "(beginning of the sentence) Thank you for your continued support," "(beginning of the sentence) I will answer your question," and "(end of the sentence) Is this resolved?"

[0059] In addition, templates for the beginning and end of a sentence are associated with each answer type. Note that the templates shown in FIG. 9 are only examples, and sentences other than these may be stored as templates. Furthermore, multiple templates may be associated with one answer type. Furthermore, one template may be associated with multiple answer types. Furthermore, the beginning sentence and the ending sentence may be associated separately.

[0060] The server device 10 compares the acquired response history with the read template (activity A34). If the answer indicated by the response history contains a sentence that is identical to or similar to the template, the server device 10 excludes the sentence as a specific entry from the answer (activity A35). For example, if a certain percentage or more of the characters contained in the sentence are the same as the template, the server device 10 determines that the sentence is similar to the template.

[0061] For example, the server device 10 determines that sentences such as "Thank you very much for your inquiry," "Thank you for your inquiry," and "Thank you for your inquiry" are similar to the template "Thank you for your inquiry," and excludes them from the response as specific items. The server device 10 executes an exclusion process for excluding specific items from all of the acquired response histories.

[0062] FIG. 10 is a diagram illustrating an example of the exclusion process. FIG. 10 shows a response sentence E11 that reads, "Thank you very much for your inquiry. Regarding the operation method of product α, it is... Please feel free to contact us if you have any other questions." In this case, the server device 10 identifies the sentences "Thank you very much for your inquiry" and "Please feel free to contact us if you have any other questions" as specific information items, as shown in sentence E12, because they are similar to the "initial response" template shown in FIG. 9. Then, the server device 10 generates sentence E13, "Regarding the operation method of product α, it is..." as a sentence after the exclusion process, excluding the specific information items.

[0063] Next, the server device 10 acquires the referenced history associated with each response history in the response history database DB1 (activity A41). Subsequently, the server device 10 performs a weighting process for each response history based on the acquired referenced history (activity A42). The response history is used as training data for machine learning by the artificial intelligence module 200, which will be described later. Specifically, machine learning is performed using the subject and inquiry email as input (the subject is optional) and the reply email as output. The input and output are also referred to as question and answer, example question and answer, etc., and the training data is also referred to as learning data or training data, etc.

[0064] Weighting processing is a process in which, for example, the more highly rated a response history is by an inquiry staff member or a user, the greater its influence as training data in machine learning. The degree of evaluation of a response history is represented by the reference history described in the description of FIG. 7. In the example of FIG. 8, the server device 10 assigns a heavier weight to an answer because the more times it is referenced, the higher the evaluation by the inquiry staff member. For example, the server device 10 assigns a heavier weight to a response history by increasing the number of times it is used as training data in the artificial intelligence module 200.

[0065] Note that the weighting method in machine learning is not limited to this. For example, some artificial intelligence modules allow the weighting of training data to be specified by parameters. In such cases, the server device 10 may execute a process of specifying the weight to be assigned using the parameters as the weighting process. Furthermore, other well-known methods for weighting training data may also be used.

[0066] Furthermore, the reference history used for weighting is not limited to the number of references, but may also be the above-mentioned reference time, reference date, identification information of the referenced inquiry staff, or the number of characters actually used in the answer text. In this case, the server device 10 weights the answer heavier, for example, the longer the reference time or the more characters used, since it can be said that the answer is more highly rated by the inquiry staff. Also, the closer the reference date is to the present, the higher the weight is, since it can be said that the answer is more recently rated. Also, the higher the position of the referenced inquiry staff, the higher the weight is, since it can be said that the answer is more highly rated by a more capable inquiry staff.

[0067] Then, the server device 10 instructs the artificial intelligence module 200 to start machine learning using the acquired response history as training data (activity A43). At this time, the server device 10 instructs the artificial intelligence module 200 to perform machine learning for each query category. The server device 10 also instructs the artificial intelligence module 200 to perform weighted machine learning using a weighting process. Specifically, the server device 10 performs this instruction by transmitting to the AI device 20 instruction data indicating the acquired response history and the above-mentioned machine learning instructions (for each category, including weighting).

[0068] The AI device 20, following the instructions indicated by the transmitted instruction data, causes the artificial intelligence module 200 to perform machine learning using the response history indicated by the instruction data as training data (activity A44). At this time, the AI device 20 also performs the machine learning by weighting as instructed for each query category. By performing the machine learning, the AI device 20 generates a learning model for each query category (activity A45). The AI device 20 associates the generated learning models with the corresponding categories and stores them (activity A46).

[0069] The machine learning process is executed as described above. Next, a draft generation process is executed in which the artificial intelligence module 200 generates a draft of an answer to the query using the learning model generated by the machine learning process.

[0070] Fig. 11 is an activity diagram showing an example of a draft generation process. The draft generation process shown in Fig. 11 is started when the server device 10 receives an inquiry email from a user, similar to the history accumulation process shown in Fig. 4. First, activities A11 and A12 shown in Fig. 4 are executed, and the person in charge screen shown in Fig. 5 is displayed.

[0071] 11, the person in charge of the inquiry operates the answer draft generation button B12. The person in charge terminal 30 then accepts this operation as an operation for generating an answer draft, and transmits instruction data indicating the inquiry content (inquiry source, inquiry subject, and inquiry text) and an instruction to generate an answer draft to the server device 10 (activity A51). The instruction data in this case is a so-called prompt. Upon receiving the instruction data, the server device 10 transmits instruction data to the AI device 20 instructing it to generate a draft of an answer corresponding to the inquiry content indicated by the instruction data, in accordance with the instruction indicated in the received instruction data (activity A52).

[0072] The AI device 20 determines the category of the query content indicated by the transmitted instruction data using the method described in the explanation of FIG. 6 (activity A53). Next, the AI device 20 reads out the learning model associated with the determined category (activity A54). The AI device 20 inputs the query content to the artificial intelligence module 200, and generates an answer draft using the read learning model (activity A55). This learning model uses answers that exclude specific description items as training data as described in FIG. 8, so an answer draft that does not include specific description items is generated.

[0073] The AI device 20 transmits the answer draft generated by the artificial intelligence module 200 to the server device 10 together with the query content used to generate the answer draft. The server device 10 determines the type of answer to the query target from the transmitted query content using the method described in the description of FIG. 6 (activity A61). Next, the server device 10 reads out a template associated with the determined answer type from the template database DB2 (activity A62). Subsequently, the server device 10 executes an addition process to add specific information indicated in the read template to the transmitted answer draft (activity A63).

[0074] FIG. 12 is a diagram illustrating an example of the addition process. FIG. 12 shows an answer draft E21 generated by the artificial intelligence module 200, which states, "Regarding the operation method of product α, it is..." If the answer type "first answer" is determined, the server device 10 generates an answer draft E22 after the addition process by adding a template associated with "first answer" to the answer draft E21, i.e., a template beginning with "Thank you for your inquiry" and ending with "Please feel free to contact us if you have any other questions" and "Thank you in advance." The server device 10 transmits the generated answer draft to the person in charge terminal 30. The person in charge terminal 30 displays the transmitted answer draft on the person in charge screen (activity A64).

[0075] FIG. 13 is a diagram showing an example of a displayed reply draft. On the person in charge screen C3 shown in FIG. 13, the generated reply draft E22 is displayed in the reply input field D14 shown in FIG. 5. The reply draft E22 is a sentence in which specific information has been added to the reply draft E21 generated by inputting the subject E31 and the inquiry content E32 into the artificial intelligence module 200. The person in charge of the inquiry makes corrections to the displayed reply draft E22 if necessary. The person in charge terminal 30 accepts the corrections (activity A65), and the reply to the user is completed. After that, the operations of activities A22 to A24 shown in FIG. 4 are performed, the reply is sent to the user, and the response history is saved. The draft generation process is executed as described above.

[0076] As described above, the server device 10 executes a history acquisition step of acquiring a response history indicating inquiries from users and responses from staff to those inquiries (activity A32). The server device 10 also executes a learning instruction step of instructing the artificial intelligence module 200 to perform machine learning using at least learning data in which the inquiries indicated in the acquired response history are used as input and the responses indicated in the response history are used as output (activity A43).

[0077] The server device 10 also executes an input step of inputting a user inquiry to the artificial intelligence module 200 (activity A52). In the machine learning process shown in Fig. 8, the artificial intelligence module 200 has already performed machine learning based on a response history, which is a history of inquiries from users and responses by personnel to those inquiries. This machine learning is machine learning that uses at least learning data (teacher data) in which the inquiries indicated in the response history are input and the responses indicated in the response history are output.

[0078] The response history is information that is accumulated in large quantities in the work of an inquiry manager responding to inquiries. By performing machine learning using such response history, the response support system 1 does not need to create large amounts of learning data for machine learning. Therefore, the workload during machine learning can be reduced compared to when the acquired response history is not used for machine learning. Note that the learning data may include manually created inquiry and response sentences. Even in this case, the workload during machine learning can be reduced because the effort of creating sentences is eliminated for the acquired response history.

[0079] Furthermore, when a new inquiry is made by a user (activity A11), the server device 10 inputs the inquiry to the artificial intelligence module 200 that has already executed machine learning (input step of activity A52), and executes an output step of outputting the answer output by the artificial intelligence module 200 to which the inquiry was input in the input step as an answer to the inquiry (activity A63). According to this embodiment, the effort required to answer an inquiry can be reduced compared to when all answers are created manually.

[0080] Furthermore, inquiries and responses are made by email in the response support system 1. According to this embodiment, the effort required for responding by email can be reduced compared to when an answer draft is not generated by the artificial intelligence module 200 when making inquiries and responding by email.

[0081] The server device 10 also executes a reference acquisition step (activity A41) that acquires a reference history (for example, the referenced history shown in FIG. 6) indicating other answers that the enquirer referred to when creating the answer. The server device 10 then executes a learning instruction step that instructs the artificial intelligence module 200 to perform machine learning on the answer indicated by the acquired response history by assigning a weight according to the history of references to the answer indicated by the acquired reference history (activities A42 and A43).

[0082] In the above example, the referenced history is the number of times the response history was referenced, but this is not limited to this. For example, the server device 10 may use the reference time (the time the response history was displayed) as the referenced history, and assign a weight to the longer the reference time. Furthermore, the server device 10 may use the number of characters or the proportion of the entire text of the portion of the referenced response history that was adopted in the actual answer as the referenced history, and assign a weight to the larger the number of characters or the proportion. Furthermore, the more recently the response history was referenced, the more weight may be assigned. In either case, the quality of the answer output by the artificial intelligence module 200 can be improved compared to when weighting based on the reference history is not performed. Furthermore, the server device 10 may set the weight to 0 so that the response history is not used in machine learning.

[0083] The server device 10 also executes an exclusion step of excluding specific entries (for example, the template shown in Fig. 9) from the answer indicated by the acquired response history (activity A35).The server device 10 also executes a learning instruction step of instructing the artificial intelligence module 200 to perform machine learning using the answer indicated by the acquired response history, from which the specific entries have been excluded, as learning data (activity A43).

[0084] The server device 10 also executes an output step (activity A63) in which it inputs an inquiry to the artificial intelligence module 200, and adds written matters to the answer output by the artificial intelligence module 200 and outputs the result as a response to the inquiry. Note that the specific written matters are not limited to the above-mentioned templates, and may be, for example, a signature or advertisement. Since these specific written matters are often written at the end of a reply email, the server device 10 may, for example, exclude the contents after the "end of sentence" template as a signature or advertisement. In either case, it is possible to eliminate the influence of the specific written matters on the response.

[0085] <Variation: Additional Learning> The response support system 1 may perform additional learning to improve the accuracy of product descriptions in response drafts generated by the draft generation process or to increase the amount of information. Additional learning is the process of learning new information while preventing the forgetting of previously learned information (destructive forgetting), and is also called continuous learning. When performing additional learning, techniques such as fine tuning, transfer learning, or distillation can be used.

[0086] Fig. 14 is an activity diagram showing an example of additional learning processing. The additional learning processing shown in Fig. 14 is started when the system administrator performs an operation (additional learning operation) to instruct additional learning on the administrator terminal 40. When the additional learning operation is performed, the server device 10 acquires product information (detailed information about the product) from the product database DB3 (activity A71).

[0087] Fig. 15 is a diagram showing an example of product database DB3. Product database DB3 shown in Fig. 15 stores product names, product descriptions, product images, option information, link information, etc., in association with one another. Option information is information on options (accessories, additional functions, etc.) related to the product. Link information is information on links to web pages containing information related to the product. It is assumed that product database DB3 stores the latest product information.

[0088] Next, the server device 10 transmits one of the acquired product information and instruction data instructing additional learning as the product information input to the AI device 20 (activity A72). When the instruction data instructing additional learning is transmitted, the AI device 20 first reads out the learning model to be subjected to additional learning, i.e., the learning model to which the product indicated by the transmitted product information is associated as the category of the inquiry (activity A73).

[0089] Next, the AI device 20 causes the artificial intelligence module 200 to perform additional learning on the read learning model using the transmitted product information as input (activity A74). By performing this additional learning, the AI device 20 updates the learning model so that the input product information can be incorporated into the answer draft (activity A75). The AI device 20 saves the updated learning model in this way (activity A76).

[0090] The additional learning process is performed as described above for one product, and is repeated the same number of times as the number of products indicated by the read product information. The learning model updated by the additional learning process is used by the artificial intelligence module 200 in the draft generation process shown in FIG.

[0091] The additional learning process is executed, for example, before the inquiry clerk starts using the artificial intelligence module 200. In this case, the server device 10 executes, for example, an information acquisition step of acquiring target information related to the subject of the inquiry from source data about the subject (activity A71). For example, the server device 10 uses the product database DB3 shown in FIG. 15 as source data and acquires product information stored in the product database DB3 as target information. Then, the server device 10 executes a learning instruction step of instructing the artificial intelligence module 200 to perform additional learning using learning data that uses the acquired target information as input (activity A72).

[0092] The source data may be a webpage article about a product or a response about the product. The target information may be a procedural guide, an instruction manual that describes how to troubleshoot a problem, or the like. In either case, the response history contains only the part of the target information that was inquired about. In contrast, the target information contains more comprehensive information about the product, the procedure, or how to troubleshoot a problem. Therefore, the response can contain more comprehensive information than if machine learning using the target information were not performed.

[0093] The additional learning process may also be performed after the inquiry specialist begins using the artificial intelligence module 200. Fig. 16 is an activity diagram showing another example of the additional learning process. The additional learning process shown in Fig. 16 is executed in a situation where new response history is accumulated and new products are provided after the person in charge of inquiries starts using the artificial intelligence module 200.

[0094] First, the server device 10 executes an information acquisition step of acquiring target information related to the target of the query from source data for that target (activity A81). The server device 10 acquires, for example, product information as target information (activity A81). Next, the server device 10 acquires a response history (activity A82), similar to activity A32 shown in FIG. 8. In activities A81 and A82, the server device 10 may acquire only the response history and product information added since the initial machine learning or the previous additional learning.

[0095] Next, the server device 10 adds the acquired target information for the inquiry target indicated by the acquired response history to the inquiry (activity A83). For example, if the inquiry email contains a link related to the product that is the subject of the inquiry, the server device 10 adds product information that is the same as or similar to the information published at the link destination to the inquiry. For example, if the inquiry email contains a link to one of multiple functions that the product has, the server device 10 adds product information that explains that function to the inquiry.

[0096] Furthermore, if the query is about one function but only the name of that function is included in the query, the server device 10 adds product information that explains that function to the query. Furthermore, if the name of the function is not stated but a phrase that appears in the description of that function is included in the query, the server device 10 adds product information that explains that function to the query. In this way, if the server device 10 can determine the specific matters being inquired about regarding the inquired product, it adds information from the product information that explains those matters in detail to the query.

[0097] Then, the server device 10 executes a learning instruction step of instructing the artificial intelligence module 200 to perform additional learning using the query with the added target information as input and the data with the response to the query as output as learning data (activity A84). According to this aspect, the content of the query used as training data is more detailed than when the target information is not added to the query, making the intent of the query clearer and improving the accuracy of the response.

[0098] The server device 10 may perform the method of adding target information to a query and performing additional learning before the use of the artificial intelligence module 200 is started. In this case, the query to which the target information is added is a query indicated by an already accumulated response history. The server device 10 may also perform the method of simply inputting target information into the artificial intelligence module 200 and performing additional learning after the use of the artificial intelligence module 200 is started. The server device 10 may also perform additional learning by adding target information to a response rather than a query, or by adding target information to both the query and the response. In either case, the accuracy of the response can be improved compared to when additional learning is not performed.

[0099] <Variation: Ability Information> The response support system 1 may use ability information that indicates the level of an inquiry person's ability to respond (hereinafter referred to as "response ability"). The response ability of an inquiry person is expressed, for example, by the number of years of experience, title, and qualifications held by the inquiry person. The longer the number of years of experience, the higher the response ability. The higher the title, the higher the response ability. The more specialized the qualification, the wider the required field of expertise, or the more difficult it is to obtain, the higher the response ability.

[0100] The server device 10 executes a staff member acquisition step of acquiring capability information indicating the level of response capability of the staff member who created the response. The capability information is, for example, registered in advance in a staff member database that stores information about the staff member in association with the identification information of the staff member. When acquiring the response history, the server device 10 also acquires the identification information of the staff member who responded, and acquires the capability information registered in association with the acquired identification information.

[0101] Next, the server device 10 executes a learning instruction step of instructing the artificial intelligence module 200 to perform machine learning by assigning a weight to the answer indicated by the acquired response history according to the level of the answering ability of the person in charge of creating the answer indicated by the acquired ability information. The server device 10 performs this weighting using a weight table that associates the level of answering ability indicated by the ability information with a weight.

[0102] FIG. 17 is a diagram showing an example of a weight table. In the weight table TB4 shown in FIG. 17, ability information, answering ability, and weight are associated with each other. The ability information includes years of experience, job title, and qualifications. For example, if the years of experience are "N1 year or more," the answering ability is "high" and the weight is "W1." If the years of experience are "N2 years or more but less than N1 year," the answering ability is "medium" and the weight is "W2." If the years of experience are "less than N2 years," the answering ability is "low" and the weight is "W3" (weights are W1>W2>W3).

[0103] The titles "T11, T12" have a "high" answering ability and a weight of "W1." The titles "T21, T22" have a "medium" answering ability and a weight of "W2." The titles "T31, T32" have a "low" answering ability and a weight of "W3." The qualifications "Q11, Q12, ..." have a "high" answering ability and a weight of "W1." The qualifications "Q21, Q22, ..." have a "medium" answering ability and a weight of "W2." The qualifications "Q31, Q32, ..." have a "low" answering ability and a weight of "W3."

[0104] The server device 10 refers to the weighting table TB4 and instructs the artificial intelligence module 200 to perform machine learning by assigning weights associated with the acquired capability information to the answers. The learning model generated by this machine learning assigns a heavier weight to answers from enquirers with higher response capabilities, so the quality of the answers output by the artificial intelligence module 200 can be improved compared to when weights based on capability information are not assigned.

[0105] The capability information may be used as the reference history described above. In this case, the server device 10 acquires the capability information of the inquiry person as a reference history indicating other answers that the inquiry person referred to when creating a response. The server device 10 then instructs the artificial intelligence module 200 to perform machine learning on the answers indicated in the acquired response history by weighting the answers indicated by the acquired capability information so that the higher the answering ability of the inquiry person indicated by the acquired capability information, the heavier the weight is. In this case, response history referenced by an inquiry person with higher answering ability is more likely to include better answers, so the quality of the answers output by the artificial intelligence module 200 can be improved compared to when weighting based on capability information as a reference history is not performed.

[0106] <Variation: Resolution> The degree to which the inquiry is resolved by the response of the inquiry specialist (hereinafter referred to as "resolution degree") may be used. In this case, the server device 10 executes a determination step of determining the degree to which the inquiry is resolved by the response indicated by the acquired response history as the resolution degree. Then, the server device 10 executes a learning instruction step of instructing the artificial intelligence module 200 to perform machine learning by assigning a weight to the response indicated by the acquired response history according to the determined resolution degree of the response.

[0107] The server device 10 determines the degree of solution by, for example, extracting keywords indicating the degree of solution from the user's reply email to the answer. Alternatively, the server device 10 may send an email to the user who made the inquiry requesting them to fill out a satisfaction survey, and determine the degree of solution based on the results of the returned survey. The server device 10 performs the weighting using a weight table that associates solution level information indicating the degree of solution with the level of solution and a weight.

[0108] Fig. 18 is a diagram showing another example of a weight table. The weight table TB5 shown in Fig. 18 associates resolution level information, resolution levels, and weights. The resolution level information includes keywords included in the reply email and survey results. For example, the keywords in the reply email, "KW11, KW12, ...", have a resolution level of "high" and a weight of "W1". Furthermore, the keywords in the reply email, "KW21, KW22, ...", have a resolution level of "medium" and a weight of "W2", and the keywords in the reply email, "KW31, KW32, ...", have a resolution level of "low" and a weight of "W3" (weights are W1>W2>W3).

[0109] Additionally, the survey results for "AN11, AN12, ..." have a "high" degree of resolution and a weight of "W1." The survey results for "AN21, AN22, ..." have a "medium" degree of resolution and a weight of "W2," and the survey results for "AN31, AN32, ..." have a "low" degree of resolution and a weight of "W3."

[0110] The server device 10 instructs the artificial intelligence module 200 to perform machine learning by referencing the weight table TB5 and assigning weights associated with the user's reply email to the answer or the user's questionnaire results to the answer. The learning model generated by this machine learning assigns a heavier weight to answers with higher resolution, so the resolution level of the output answer can be improved compared to when no weight based on the resolution level is assigned. Note that the resolution level may be determined using a classifier that realizes a function of classifying the resolution level from the reply email or questionnaire results using machine learning.

[0111] <Modifications: Other examples> In response support system 1, inquiries and responses are made via email, but this is not limiting and may also be done via chat, or by exchanging messages via a web page or app function. In either case, by performing machine learning using at least a response history including responses created by an inquiry person in the course of actual work as learning data, the workload during machine learning can be reduced compared to when such response history is not used in machine learning.

[0112] <Example of variation: Variation of composition> The configuration (overall configuration, hardware configuration, functional configuration, etc.) shown in FIG. 1 and other figures is an example, and other configurations may be used as long as they are not inconvenient for implementation. For example, the server device 10 may be distributed across two or more devices, or may be provided in the form of SaaS (Software as a Service) or a cloud computing system. Furthermore, the information processing performed by the server device 10 may be collectively executed by the staff terminal 30. In short, as long as the necessary information processing is executed by the entire response support system 1, the devices that execute that information processing may have any configuration.

[0113] The output destination of information or data (hereinafter referred to as "information, etc.") may be another device, a display, a memory unit (including an internal memory unit and an external memory unit), an email address, an account of another system, etc. Acquisition of information, etc. includes acquiring information, etc. generated by the device itself, as well as acquiring information, etc. transmitted from another device. The table, etc. (table, database, etc.) in which parameters are associated is not limited to the illustrated table, etc., and the number of parameters may be reduced or increased. Furthermore, information, etc. corresponding to parameters may be obtained using a mathematical formula, a conditional formula, etc., without using a table, etc.

[0114] The above-described embodiments are information processing devices such as the server device 10 and the person in charge terminal 30, and information processing systems such as the response support system 1 including the server device 10 and the person in charge terminal 30, but may also be information processing methods. The information processing methods include the same steps as those executed by the information processing system. The above-described embodiments may also be programs. The programs cause a computer to execute the same steps as those executed by the information processing system.

[0115] <Additional Notes> Furthermore, it may be provided in the following aspects.

[0116] (1) An information processing system having at least one processor, wherein the processor, in a history acquisition step, acquires a response history indicating inquiries from users and responses from personnel to the inquiries, and in a learning instruction step, instructs an artificial intelligence module to perform machine learning using at least learning data in which the inquiries indicated in the acquired response history are used as input and the responses indicated in the response history are used as output.

[0117] According to this aspect, the workload during machine learning can be reduced.

[0118] (2) In the information processing system described in (1) above, in the output step, when a new inquiry is made by a user, the processor inputs the inquiry into the artificial intelligence module that has already performed the machine learning, and outputs the answer output by the artificial intelligence module as the answer to the inquiry.

[0119] According to this aspect, it is possible to reduce the time and effort required to respond to inquiries.

[0120] (3) In the information processing system described in (1) or (2) above, the processor performs the inquiry and the response by email.

[0121] According to this embodiment, the time and effort required for replying by e-mail can be reduced.

[0122] (4) In the information processing system described in any one of (1) to (3) above, in the information acquisition step, the processor acquires target information regarding the target of the query from source data about the target, and in the learning instruction step, instructs the artificial intelligence module to perform additional learning using learning data that has the acquired target information as input.

[0123] According to this embodiment, more comprehensive information can be included in the response.

[0124] (5) In the information processing system described in any one of (1) to (4) above, in the information acquisition step, the processor acquires target information regarding the target of the query from source data regarding the target of the query, and in the learning instruction step, instructs the artificial intelligence module to perform additional learning using as input the target information acquired for the target of the query indicated by the acquired response history added to the query, and data that outputs a response to the query as the learning data.

[0125] According to this embodiment, the accuracy of the answer can be improved.

[0126] (6) In the information processing system described in any one of (1) to (5) above, in the reference acquisition step, the processor acquires a reference history indicating other answers that the person in charge referred to when creating the answer, and in the learning instruction step, instructs the artificial intelligence module to perform machine learning on the answer indicated by the acquired response history by assigning a weight according to the history of reference to the answer indicated by the acquired reference history.

[0127] According to this embodiment, the quality of the answers output by the artificial intelligence module can be improved.

[0128] (7) In the information processing system described in any one of (1) to (6) above, in the person-in-charge acquisition step, the processor acquires ability information indicating the level of the answering ability of the person in charge who created the answer, and in the learning instruction step, instructs the artificial intelligence module to perform machine learning on the answer indicated by the acquired response history by assigning a weight according to the level of the answering ability of the person in charge who created the answer indicated by the acquired ability information.

[0129] According to this embodiment, the answer output by the artificial intelligence module can be improved.

[0130] (8) In the information processing system described in any one of (1) to (7) above, in the determination step, the processor determines the degree to which the inquiry is resolved by the answer indicated by the acquired response history as a resolution level, and in the learning instruction step, instructs the artificial intelligence module to perform machine learning by assigning a weight to the answer indicated by the acquired response history according to the resolution level determined for the answer.

[0131] According to this aspect, the degree of resolution provided by the output answer can be improved.

[0132] (9) In the information processing system described in (2) above, in the excluding step, the processor excludes specific information from the answer indicated by the acquired response history, in the learning instruction step, instructs the artificial intelligence module to perform machine learning using the answer indicated by the acquired response history from which the information has been excluded as the learning data, and in the output step, inputs the query to the artificial intelligence module, and outputs the answer output by the artificial intelligence module plus the information as the answer to the query.

[0133] According to this embodiment, it is possible to eliminate the influence of specific entries on the response.

[0134] (10) An information processing system having a processor, wherein in an input step, the processor inputs a user's inquiry to an artificial intelligence module, and the artificial intelligence module has performed machine learning based on a response history, which is a history of inquiries from users and replies to the inquiries by personnel, and the machine learning is machine learning that uses at least learning data that inputs the inquiries indicated in the response history and outputs the answers indicated in the response history, and in an output step, outputs the answer output by the artificial intelligence module to which the inquiry was input in the input step as the answer to the inquiry.

[0135] According to this aspect, the workload during machine learning can be reduced.

[0136] (11) An information processing method, comprising the steps of the information processing system according to any one of (1) to (10) above.

[0137] According to this aspect, the workload during machine learning can be reduced.

[0138] (12) A program that causes a computer to execute each step of the information processing system according to any one of (1) to (10) above.

[0139] According to this aspect, the workload during machine learning can be reduced. Of course, this is not the case. Furthermore, the above-described embodiments and modifications may be combined in any desired manner.

[0140] Finally, while various embodiments of the present invention have been described, these are presented by way of example only and are not intended to limit the scope of the invention. The novel embodiments may be embodied in various other forms, and various omissions, substitutions, and modifications may be made without departing from the spirit of the invention. The embodiments and their modifications are intended to be included within the scope and spirit of the invention, as well as within the scope of the inventions and their equivalents as defined in the appended claims. [Explanation of symbols]

[0141] 1: Response support system 2: Communication line 10: Server device 11: Control section 20:AI device 21: Control unit 30: Person in charge terminal 31: Control unit 40: Administrator terminal 41: Control unit 200: Artificial Intelligence Module

Claims

1. An information processing system comprising at least one processor, the processor: In the history acquisition step, a response history indicating inquiries from users and responses to the inquiries by personnel is acquired; In the learning instruction step, the artificial intelligence module is instructed to perform machine learning using at least learning data in which the query indicated by the acquired response history is used as input and the answer indicated by the response history is used as output. Information processing system.

2. 2. The information processing system according to claim 1, the processor: In the output step, when a new inquiry is made by a user, the inquiry is input to the artificial intelligence module that has already performed the machine learning, and the answer output by the artificial intelligence module is output as the answer to the inquiry. Information processing system.

3. 2. The information processing system according to claim 1, the processor: The inquiry and the response are made by email. Information processing system.

4. 2. The information processing system according to claim 1, the processor: In the information acquisition step, object information relating to the object of the query is acquired from source data for the object of the query; In the learning instruction step, the artificial intelligence module is instructed to perform additional learning using learning data that has the acquired target information as an input. Information processing system.

5. 2. The information processing system according to claim 1, the processor: In the information acquisition step, object information relating to the object of the query is acquired from source data for the object of the query; In the learning instruction step, the artificial intelligence module is instructed to perform additional learning using, as input, the target information acquired for the target of the inquiry indicated by the acquired response history, added to the inquiry, and using data that is an answer to the inquiry as output as the learning data. Information processing system.

6. 2. The information processing system according to claim 1, the processor: In the reference acquisition step, a reference history indicating other answers that the person in charge referred to when creating the answer is acquired, In the learning instruction step, the artificial intelligence module is instructed to perform machine learning by assigning a weight to the answer indicated by the acquired response history according to a history of reference to the answer indicated by the acquired reference history. Information processing system.

7. 2. The information processing system according to claim 1, the processor: In the person-in-charge acquisition step, ability information indicating the level of the answering ability of the person-in-charge who created the answer is acquired, In the learning instruction step, the artificial intelligence module is instructed to perform machine learning by assigning a weight to the answer indicated by the acquired response history in accordance with the level of the answering ability of the person in charge who created the answer indicated by the acquired ability information. Information processing system.

8. 2. The information processing system according to claim 1, the processor: In the determination step, a degree to which the inquiry is resolved by the response indicated by the acquired response history is determined as a resolution degree; In the learning instruction step, the artificial intelligence module is instructed to perform machine learning by assigning a weight to the answer indicated by the acquired response history according to the determined degree of solution. Information processing system.

9. 3. The information processing system according to claim 2, the processor: In the excluding step, specific descriptions are excluded from the answers indicated by the acquired response history; In the learning instruction step, the artificial intelligence module is instructed to perform machine learning using answers indicated by the acquired response history, excluding the recorded items, as the learning data; In the output step, the query is input to the artificial intelligence module, and the answer output by the artificial intelligence module is added with the description items and output as the answer to the query. Information processing system.

10. An information processing system including a processor, the processor: In the input step, a user's query is input to the artificial intelligence module. The artificial intelligence module has performed machine learning based on a response history, which is a history of inquiries from users and responses by personnel to the inquiries; the machine learning is machine learning using at least learning data in which a query indicated by the response history is input and an answer indicated by the response history is output, In the output step, the answer output by the artificial intelligence module to which the query was input in the input step is output as the answer to the query. Information processing system.

11. An information processing method, comprising: The information processing system according to any one of claims 1 to 10, Information processing methods.

12. A program, A computer is caused to execute each step of the information processing system according to any one of claims 1 to 10. program.

Images