Response support apparatus, response support method, and response support program

Abstract

To take care of the mind of an operator corresponding to an inquiry.SOLUTION: An input part 21 receives conversation information between a user of a station and an operator corresponding to an inquiry from the user. The emotion estimation unit 28 estimates, from the conversation information received by the input unit 21, the user's emotion at the time of handling by the operator. If the user's emotion is a positive emotion, the emotion estimation unit 28 outputs the estimated user's emotion to the operator terminal used by the operator.SELECTED DRAWING: Figure 22

Description

[Technical Field]

[0001] The present disclosure relates to a technology for assisting in responding to inquiries at stations. [Background technology]

[0002] In recent years, the number of unmanned stations has been increasing. It is necessary to create an environment where various people can use each station without inconvenience. To realize such an environment, it is desirable to be able to respond to inquiries at station counters remotely even if there is no station staff at the station. It is also desirable to be able to respond to inquiries at station counters via telework, etc. However, remotely responding to a wide variety of inquiries from multiple stations and multiple ticket offices is a complex and difficult task, so there is a need to support operators in responding to inquiries.

[0003] Patent Document 1 describes that if an inquiry input at a station information terminal is a standard inquiry, a standard reply is given, and if the inquiry is not a standard inquiry, the terminal is connected to a staff member terminal. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2004-240807 Summary of the Invention [Problem to be solved by the invention]

[0005] Even with the technology described in Patent Document 1, operators will still have to respond to non-routine inquiries from multiple stations and multiple ticket windows. It is these non-routine inquiries that are difficult to respond to and require support. Furthermore, operators who handle a variety of inquiries can be stressed, so psychological support is also required. The present disclosure aims to provide mental care for operators. [Means for solving the problem]

[0006] A response assistance device according to the present disclosure includes: an input unit that receives conversation information between a station user and an operator who responds to an inquiry from the user; an emotion estimation unit that estimates the emotion of the user at the time of the operator's response from the conversation information received by the input unit, and if the emotion of the user is positive, outputs the estimated emotion of the user to an operator terminal used by the operator; Equipped with. [Effects of the Invention]

[0007] In the present disclosure, the positive emotions of the user are notified to the operator, thereby increasing the motivation of the operator to work. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a configuration diagram of a response support device 10 according to a first embodiment. [Figure 2] 3 is a flowchart showing the operation of the response support device 10 according to the first embodiment. [Figure 3] FIG. 2 is an explanatory diagram of the operation of the response support device 10 according to the first embodiment. [Figure 4] 10 is a flowchart showing the operation of the response support device 10 according to the second embodiment. [Figure 5] FIG. 10 is an explanatory diagram of the operation of the response support device 10 according to the second embodiment. [Figure 6] 10 is a flowchart showing the operation of the response support device 10 according to the third embodiment. [Figure 7] FIG. 10 is an explanatory diagram of the operation of the response support device 10 according to the third embodiment. [Figure 8] 10 is a flowchart showing the operation of the response support device 10 according to the fourth embodiment. [Figure 9] FIG. 10 is an explanatory diagram of the operation of the response support device 10 according to the fourth embodiment. [Figure 10] 10 is a flowchart showing the operation of the response support device 10 according to the fifth embodiment. [Figure 11] FIG. 10 is an explanatory diagram of the operation of the response support device 10 according to the fifth embodiment. [Figure 12] FIG. 10 is a configuration diagram of a response support device 10 according to a first modification. [Figure 13] 10 is a flowchart showing the operation of the response support device 10 according to the first modification. [Figure 14] FIG. 10 is an explanatory diagram of the operation of the response support device 10 according to the first modification. [Figure 15] FIG. 13 is a configuration diagram of a response support device 10 according to a sixth embodiment. [Figure 16] 13 is a flowchart showing the operation of the response support device 10 according to the sixth embodiment. [Figure 17] FIG. 20 is an explanatory diagram of the operation of the response support device 10 according to the sixth embodiment. [Figure 18] FIG. 13 is a configuration diagram of a response support device 10 according to a seventh embodiment. [Figure 19] 13 is a flowchart showing the operation of the response support device 10 according to the seventh embodiment. [Figure 20] FIG. 20 is a diagram showing a list screen of inquiries according to the seventh embodiment. [Figure 21] FIG. 11 is a configuration diagram of a response support device 10 according to a third modification. [Figure 22] FIG. 13 is a configuration diagram of a response support device 10 according to an eighth embodiment. [Figure 23] 13 is a flowchart showing the operation of the response support device 10 according to the eighth embodiment. [Figure 24] FIG. 20 is a configuration diagram of a response support device 10 according to a ninth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Embodiment 1 ***Configuration Description*** The configuration of a response support device 10 according to the first embodiment will be described with reference to FIG. The response support device 10 is a computer. The response assistance device 10 includes hardware such as a processor 11, a memory 12, a storage 13, and a communication interface 14. The processor 11 is connected to other hardware via signal lines and controls the other hardware.

[0010] The processor 11 is an IC that performs processing. IC stands for Integrated Circuit. Specific examples of the processor 11 include a CPU, a DSP, and a GPU. CPU stands for Central Processing Unit. DSP stands for Digital Signal Processor. GPU stands for Graphics Processing Unit.

[0011] The memory 12 is a storage device that temporarily stores data. Specific examples of the memory 12 include SRAM and DRAM. SRAM stands for Static Random Access Memory. DRAM stands for Dynamic Random Access Memory.

[0012] The storage 13 is a storage device that stores data. A specific example of the storage 13 is an HDD. HDD is an abbreviation for Hard Disk Drive. The storage 13 may also be a portable recording medium such as an SD (registered trademark) memory card, CompactFlash (registered trademark), NAND flash, a flexible disk, an optical disk, a compact disk, a Blu-ray (registered trademark) disk, or a DVD. SD is an abbreviation for Secure Digital. DVD is an abbreviation for Digital Versatile Disk.

[0013] The communication interface 14 is an interface for communicating with an external device. Specific examples of the communication interface 14 include Ethernet (registered trademark), USB, HDD, etc. It is an MI (registered trademark) port. USB is an abbreviation for Universal Serial Bus. HDMI is an abbreviation for High-Definition Multimedia Interface. The response support device 10 is connected to an operator terminal 51, a remote counter terminal 52, and the like via a communication interface 14.

[0014] The response support device 10 includes, as functional components, an input unit 21, a model control unit 22, and an output unit 23. The functions of the functional components of the response support device 10 are realized by software. The storage 13 stores a program that realizes the function of each functional component of the response support device 10. This program is read into the memory 12 by the processor 11 and executed by the processor 11. In this way, the function of each functional component of the response support device 10 is realized.

[0015] The storage 13 stores a response generation model 31. The response generation model 31 may be stored in a storage device external to the response support device 10, instead of in the storage 13.

[0016] 1 shows only one processor 11. However, there may be a plurality of processors 11, and the plurality of processors 11 may cooperate to execute programs that realize the respective functions.

[0017] ***Explanation of Operation*** The operation of the response support device 10 according to the first embodiment will be described with reference to FIGS. The operation procedure of the response support device 10 according to the embodiment 1 corresponds to the response support method according to the embodiment 1. Moreover, the program that realizes the operation of the response support device 10 according to the embodiment 1 corresponds to the response support program according to the embodiment 1.

[0018] (Step S11: Input processing) The input unit 21 accepts input of inquiry information 41 from a user at a ticket counter in a station. Specifically, an operator is called from a remote counter terminal 52 or the like installed at a station counter via an operator terminal 51 or the like used by the operator, and a conversation takes place between the user and the operator. The conversation is conducted by voice or text. The text data of the conversation between the user and the operator is the inquiry information 41. If the conversation is conducted by voice, the voice is converted into text to generate the inquiry information 41. Note that the inquiry information 41 may consist only of the user's remarks from the conversation. Furthermore, the user may input the content of the inquiry before calling an operator from the remote terminal 52. In this case, the information representing the input content of the inquiry is the inquiry information 41.

[0019] (Step S12: Model control processing) The model control unit 22 inputs the inquiry information 41 received in step S11 into the response generation model 31, which is a learning model. The response generation model 31 is given in advance a station work manual 42, which is a manual for station work. In other words, the response generation model 31 is trained using the station work manual 42 as learning data. Alternatively, the response generation model 31 is able to access the station work manual 42. The station work manual 42 also describes the work to be done in response to inquiries. When inputting the inquiry information 41, the model control unit 22 also inputs to the response generation model 31 instructions to generate a response plan 43 for the inquiry information 41 and to specify a grounds description 44 that serves as the grounds for the response plan 43 in the station service manual 42. In other words, the model control unit 22 inputs as a prompt The user inputs inquiry information 41 and also inputs instructions to generate a response proposal 43 and specify a grounds description 44. Then, the model control unit 22 acquires the proposed response 43 output by the response generation model 31 and the grounds description 44 that is the grounds for the proposed response 43 in the station service manual 42.

[0020] The learning model is what is known as generative AI or LLM. AI stands for Artificial Intelligence. LLM stands for Large Language Model. The learning model may be constructed using algorithms such as BERT and GPT. BERT is a Bidirectional Encoder Representations from Transformers. GPT stands for Generative Pretrained Transformer. A learning model may be constructed by combining multiple algorithms, including these algorithms. The learning model may be a trainable model, where the learning model is trained using operator responses. When the learning model is a generative AI, it is assumed that the learning model is stored in a storage device external to the response support device 10.

[0021] (Step S13: Output processing) The output unit 23 outputs the proposed response 43 and the reasoning statement 44 acquired in step S12. Specifically, the output unit 23 outputs the proposed response 43 and the rationale description 44 to the operator terminal 51 or the like used by the operator, and displays them. This allows the operator to respond to the user by referring to the proposed response 43 and the rationale description 44. Here, when displaying the grounds statement 44, instead of displaying an excerpt of the grounds statement 44 in the station service manual 42, the page of the grounds statement 44 in the station service manual 42 may be displayed. This allows the operator to check the statements before and after the grounds statement 44, which may allow for a more appropriate response.

[0022] ***Effects of the First Embodiment*** As described above, the response support device 10 according to the first embodiment uses the response generation model 31 to acquire the response plan 43 corresponding to the inquiry information 41 and the rationale description 44 of the response plan 43 in the station service manual 42. By using the response generation model 31, it is possible to generate the response plan 43 even for an unstructured inquiry. Furthermore, since the rationale description 44 of the response plan 43 in the station service manual 42 is obtained, the operator can provide an appropriate response.

[0023] Embodiment 2 The second embodiment differs from the first embodiment in that additional information 45 is input. In the second embodiment, this difference will be explained, and explanation of the same points will be omitted.

[0024] ***Explanation of Operation*** The operation of the response support device 10 according to the second embodiment will be described with reference to FIGS. The process in step S23 is the same as the process in step S13 in FIG.

[0025] (Step S21: Input processing) The input unit 21 receives input of additional information 45 in addition to the inquiry information 41. Here, the input unit 21 may receive the additional information 45 directly, or may receive information that can identify the additional information 45. This will be described in detail later.

[0026] (Step S22: Model control processing) The model control unit 22 inputs the additional information 45 received in step S21, in addition to the inquiry information 41, to the response generation model 31. Then, the model control unit 22 acquires the response proposal 43 and the rationale description 44 output by the response generation model 31.

[0027] The additional information 45 includes information such as counter information indicating the location of the counter, train operation information, train fare information, equipment information, personal information of the user, ticket information, lost property information, map information, and event information.

[0028] The counter information indicates the location of the counter where the user is making an inquiry. The location of the counter is, for example, the name of the station where the counter is installed and the name of the ticket gate. The counter information can be identified from the remote counter terminal 52 to which the operator terminal 51 is connected. Therefore, when the additional information 45 is counter information, the input unit 21 accepts input of terminal identification information from the remote counter terminal 52 to which the operator terminal 51 is connected. Then, the input unit 21 identifies the counter information from the terminal identification information. In other words, when a user inputs an inquiry and an operator responds, the input unit 21 acquires terminal identification information from the remote counter terminal 52 to which the operator terminal 51 of the responding operator is connected. Then, the input unit 21 identifies the counter information from the identification information of the remote counter terminal 52 by referring to predetermined correspondence information between the identification information of the remote counter terminal 52 and the counter information. In this way, the input unit 21 acquires the identified counter information as additional information 45 in addition to the inquiry information 41. By taking the ticket counter information into consideration, the response generation model 31 is able to output a response proposal 43 and a reasoning statement 44 specific to the station or ticket gate such as the nearest exit.

[0029] In addition to the ticket counter information, the name of the user's destination station and the names of any intermediate stations may be input as additional information 45. For example, when a user inquires about their destination, input unit 21 may acquire the name of the user's destination station and the names of any intermediate stations as additional information 45. When input unit 21 recognizes an inquiry including the name of the destination station and the names of any intermediate stations by speech recognition, it is conceivable that input unit 21 may use the recognized name of the destination station and the names of any intermediate stations as additional information 45. By taking into consideration the name of the user's destination station and the names of any intermediate stations, response generation model 31 becomes able to output a route as response proposal 43.

[0030] The train operation information indicates the operation status, such as whether the train is operating normally or whether there is a delay or cancellation. The train operation information can be obtained from a system that manages train operations. Therefore, when the additional information 45 is operation information, the input unit 21 accepts input of the operation information from the system that manages train operations. For example, when receiving an inquiry from a user, the input unit 21 accepts input of a category indicating that the inquiry is about train operation information. When the category indicating that the inquiry is about train operation information is input in addition to the inquiry information 41, the input unit 21 acquires operation information from a system that manages train operations as additional information 45. Furthermore, the input unit 21 may recognize the content of the inquiry from the user, and when it is recognized as an inquiry about train operations, it may set the train operation information as additional information 45. Note that the input unit 21 specifies the information to be acquired and the acquisition source by referring to information that indicates the information to be acquired and the acquisition source in advance according to the content of the inquiry. The method of recognizing the content of the inquiry may be any method, such as using AI. By taking into account train operation information, the response generation model 31 is able to output a proposed response 43 that takes into account the operation status and a reasoning statement 44. For example, the response generation model 31 is able to output a route or boarding time that takes into account the operation status as the proposed response 43.

[0031] The train fare information is information that indicates the train fare for each section of travel. The train fare information can be obtained from a fare management system. Therefore, when the additional information 45 is fare information, the input unit 21 accepts input of the fare information from the fare management system. For example, when the input unit 21 receives an inquiry from a user, the input unit 21 may input information about train fare information. The input unit 21 accepts input of a category that the inquiry is an inquiry about train fare information in addition to the inquiry information 41. When the category that the inquiry is about train fare information is input, the input unit 21 acquires fare information from the fare management system as additional information 45. Furthermore, the input unit 21 may recognize the content of the inquiry from the user, and if it is recognized as an inquiry about train fares, it may set fare information as additional information 45. By taking train fare information into consideration, the response generation model 31 is able to output a proposed response 43 and a reasoning statement 44 that take into account competing sections. A competing section is a section where multiple railway companies' lines run. For competing sections, fares may be set lower in consideration of the fares of other companies. It is possible to output an appropriate proposed response 43 even for inquiries about fares that have such special circumstances.

[0032] The equipment information is information managed by the railway company about equipment installed at stations that can be used by users. Specifically, the equipment information is data managed by the railway company about equipment such as automatic ticket gates, ticket issuing machines, and fare adjustment machines. The equipment information can be acquired from the equipment management system. Therefore, when the additional information 45 is equipment information, the input unit 21 accepts input of the equipment information from the equipment management system. For example, when receiving an inquiry from a user, the input unit 21 receives input of a category indicating that the inquiry is about being unable to pass through the ticket gate. When the category indicating that the inquiry is about being unable to pass through the ticket gate is input in addition to the inquiry information 41, the input unit 21 acquires device information about the automatic ticket gate from the device management system as additional information 45. The input unit 21 recognizes the content of the inquiry from the user, and when it is recognized that the inquiry is about being unable to pass through the ticket gate, it may acquire device information about the automatic ticket gate as additional information 45. By taking device information into consideration, the response generation model 31 can output a proposed response 43 and a rationale statement 44 according to the status of the device operated by the user. For example, when a user inquires that they were unable to pass through a ticket gate, the proposed response 43 and rationale statement 44 can be output taking into account causes such as insufficient balance or no entry history.

[0033] The user's personal information is information such as the user's age and gender. The user's personal information is input by the user using the remote counter terminal 52. The input unit 21 then accepts the input of the user's personal information from the remote counter terminal 52. The input unit 21 may acquire the user's personal information from the remote counter terminal 52 by using the recognition results of an image from a camera installed in the remote counter terminal 52, information on an IC card read by the remote counter terminal 52, the results of voice recognition, etc. By taking into account the user's personal information, the response generation model 31 can output a response proposal 43 and a reasoning statement 44 according to characteristics such as age and gender. For example, it can output a response proposal 43 and a reasoning statement 44 that are easy to understand for children.

[0034] Ticket information is information about tickets held by a user. Tickets include passenger tickets and limited express tickets. Ticket information includes the fare, the identification information of the train the user plans to board, the seat number, etc. The ticket information is input by the user using the remote ticket counter terminal 52. At this time, the user may manually input the information written on the ticket, or the information on the ticket may be read using a scanner. Also, if the ticket information is stored on an IC card, the ticket information may be input by having a card reader installed in the remote ticket counter terminal 52 read the ticket information from the IC card. Therefore, the input unit 21 accepts input of ticket information from the remote ticket counter terminal 52. By taking ticket information into consideration, the response generation model 31 is able to output a proposed response 43 and a statement of reasons 44 that take into consideration the contents of the ticket in hand.

[0035] The lost item information is information about items lost by passengers on trains. The lost item information includes the classification of the lost item and information about the train on which the passenger was riding when the item was lost. The lost item information is input by the user using the remote terminal 52. The input unit 21 receives the input of the lost item information from the remote terminal 52. For example, when input unit 21 receives an inquiry from a user, it accepts input of a category indicating that the inquiry is about losing belongings. When the category indicating that the inquiry is about losing belongings is input in addition to inquiry information 41, input unit 21 acquires lost item information from remote counter terminal 52 as additional information 45. Input unit 21 recognizes the content of the inquiry from the user, and may acquire lost item information as additional information 45 when it is recognized that the inquiry is about losing belongings. By taking into consideration the lost item information, the response generation model 31 can output a proposed response 43 indicating specific contact information, etc., and a reason description 44. Specific contact information may be the name of a station or a police station, etc.

[0036] The map information indicates a map of the area around the station. The map information can be acquired from an external map management server or the like. Therefore, the input unit 21 accepts input of the map information from the external map management server or the like. At this time, the input unit 21 may accept input of map information corresponding to the location of the counter indicated by the counter information. For example, when receiving an inquiry from a user, the input unit 21 receives input of a category indicating that the inquiry is for local guidance. When the category indicating that the inquiry is for local guidance is input in addition to the inquiry information 41, the input unit 21 acquires map information from a map management server or the like as additional information 45. The input unit 21 may recognize the content of the inquiry from the user, and acquire map information as additional information 45 when it is recognized that the inquiry is for local guidance. The response generation model 31 takes into consideration map information, making it possible to output a response plan 43 showing specific routes and a reasoning statement 44.

[0037] The event information is information about events taking place in the vicinity of the station. The event information includes information such as the location and time of the event. The event information can be acquired from an external event management server or the like. Therefore, the input unit 21 accepts input of the event information from the event management server. By taking event information into consideration, the response generation model 31 is able to output a proposed response 43 such as route guidance that avoids congestion and a description of the reasons 44.

[0038] ***Effects of the Second Embodiment*** As described above, the response support device 10 according to the second embodiment inputs the additional information 45 in addition to the inquiry information 41 to the response generation model 31. This makes it possible to acquire a more appropriate response plan 43 and rationale description 44. As a result, it becomes possible for the operator to provide an appropriate response.

[0039] Embodiment 3 The third embodiment differs from the first and second embodiments in that a past response history 46 is provided to the response generation model 31. In the third embodiment, this difference will be explained, and explanation of the same points will be omitted. In the third embodiment, a case where a modification is made to the first embodiment will be described. However, it is also possible to make modifications to the second embodiment.

[0040] ***Explanation of Operation*** The operation of the response support device 10 according to the third embodiment will be described with reference to FIGS. The processes in steps S31 and S33 are the same as those in steps S11 and S13 in FIG.

[0041] (Step S32: Model control processing) As in the first embodiment, the model control unit 22 inputs inquiry information 41 to the response generation model 31. In the third embodiment, the response generation model 31 is provided in advance with a past response history 46 in addition to a station service manual 42. That is, the response generation model 31 is trained using the response history 46 as training data. Alternatively, the response generation model 31 is able to access the response history 46. The response history 46 is a combination of inquiries from users in the past and responses given to the users in response to the inquiries. That is, the response history 46 is a combination of inquiry information 41 and responses given by operators to users in response to inquiries in the inquiry information 41, and is stored in the storage 13 of FIG. 1. The response content may be the response plan 43 output by the response generation model 31. However, the response content may be the response plan 43 modified by an operator, or may be the response content created by an operator completely different from the response plan 43. Then, the model control unit 22 acquires the proposed response 43 and the reasoning description 44 output by the response generation model 31.

[0042] ***Effects of the Third Embodiment*** As described above, in the response support device 10 according to the third embodiment, the past response history 46 is provided to the response generation model 31. This makes it possible to generate a proposed response 43 based on the response content given by the operator to the user when a similar inquiry has been made in the past. For example, when an inquiry is made from a ticket counter at a new station, it becomes possible to generate a proposed response 43 based on the response history of similar inquiries made in the past at other similar stations. This makes it possible to obtain a more appropriate proposed response 43 and rationale description 44. As a result, it becomes possible for the operator to give an appropriate response.

[0043] Embodiment 4 The fourth embodiment differs from the first to third embodiments in that a classification 47, which is a premise for generating a proposed response 43, is input to a response generation model 31. In the fourth embodiment, this difference will be explained, and explanations of the same points will be omitted. In the fourth embodiment, a case where a modification is made to the first embodiment will be described. However, modifications can also be made to the second and third embodiments.

[0044] ***Explanation of Operation*** The operation of the response support device 10 according to the fourth embodiment will be described with reference to FIGS. The process in step S43 is the same as the process in step S13 in FIG.

[0045] (Step S41: Input processing) The input unit 21 receives input of the classification 47 in addition to the inquiry information 41. Here, the input unit 21 may directly receive the classification 47, or may receive information that can identify the classification 47. This will be described in detail later.

[0046] (Step S42: Model control processing) The model control unit 22 inputs the classification 47 received in step S41 to the response generation model 31 in addition to the inquiry information 41. The response generation model 31 then generates a proposed response 43 and a rationale statement 44 based on the classification 47. The model control unit 22 then acquires the proposed response 43 and rationale statement 44 output by the response generation model 31.

[0047] The classification 47 is information that is a premise for generating the response plan 43. The classification 47 is at least one of the attributes of the station, the user, and the inquiry information 41.

[0048] The attributes of a station include its size, location, type, etc. The size is determined by the number of passengers, etc. The size of a station can be divided into, for example, large stations, medium-sized stations, small stations, etc. In addition, unmanned stations without station staff may also be included in the classification of station size. Location indicates whether the area is a commuter town, an office district, or a tourist destination. Type indicates whether the station is a transfer station or not. Station attributes may also include the number of exits, etc. The attributes of the station can be identified from the remote counter terminal 52 connected to the operator terminal 51. Therefore, the input unit 21 receives input of terminal identification information from the remote counter terminal 52 connected to the operator terminal 51. Then, the input unit 21 identifies the attributes of the station from the terminal identification information.

[0049] For example, when providing road guidance, the route to be guided may differ between an office district and a tourist spot. In an office district, guidance may be given to the shortest route, whereas in a tourist spot, guidance may be given to a route that passes through stores or points of interest.

[0050] The attributes of the user include walking status, purpose of use, etc. Walking status can be classified into normal walking, wheelchair use, etc. Purpose of use can be classified into tourist, business person, etc. The attributes of the user are input by the user using the remote counter terminal 52. The input unit 21 then accepts the input of the user attributes from the remote counter terminal 52. The input unit 21 may acquire image data obtained by photographing the user using a camera provided in the remote counter terminal 52, and identify the attributes of the user from the image data. The input unit 21 may also identify the attributes of the user from the results of identifying the voice input by the user using the remote counter terminal 52.

[0051] For example, when providing directions, the route to be guided varies depending on the walking state of the user. Therefore, by generating a response plan 43 based on the walking state of the user, an appropriate response plan 43 can be generated.

[0052] The attributes of the inquiry information 41 include inquiries specific to a ticket counter, inquiries specific to a station, inquiries common to all lines, general inquiries, and the like. The attributes of the inquiry information 41 are input by the user using the remote terminal 52. For example, before connecting to an operator, the user is prompted to select a category of the inquiry content, thereby inputting the attributes of the inquiry information 41. The input unit 21 then accepts input of the attributes of the inquiry information 41 from the remote terminal 52.

[0053] For example, if the inquiry is common to the line, even if an appropriate response example cannot be obtained at the counter where the inquiry originated, if a response example to a similar inquiry is obtained at another station along the line, it will be possible to generate a response proposal 43 based on that response example.

[0054] ***Effects of the Fourth Embodiment*** As described above, the response support device 10 according to the fourth embodiment inputs the classification 47, which is a premise for generating the response plan 43, to the response generation model 31. Then, the response generation model 31 generates the response plan 43 based on the classification 47. This makes it possible to obtain a more appropriate response plan 43 and rationale description 44. As a result, the operator can provide an appropriate response.

[0055] Embodiment 5. The fifth embodiment differs from the first to fourth embodiments in that response plan 43 is generated by separating out parts for which it is difficult to generate response plan 43. In the fifth embodiment, this difference will be explained, and explanation of the same points will be omitted. In the fifth embodiment, a case where a modification is made to the first embodiment will be described. Modifications to forms 2 to 4 may also be made.

[0056] ***Explanation of Operation*** The operation of the response support device 10 according to the fifth embodiment will be described with reference to FIGS. The processes in steps S51 and S53 are the same as those in steps S11 and S13 in FIG.

[0057] (Step S52: Model control processing) In addition to the inquiry information 41, the model control unit 22 inputs to the response generation model 31 classification instructions 48 for classifying response plans 43 according to the difficulty of automatically generating the response plans 43. The classification instructions 48 are commands to classify and generate response plans 43 according to the difficulty of automatically generating the response plans 43. Then, the response generation model 31 determines the difficulty of automatic generation for each content of the response plan 43. Then, the response generation model 31 generates the response plan 43 by dividing it into parts with high difficulty and parts with low difficulty. In addition, the response generation model 31 identifies the grounds description 44 that serves as the grounds for the response plan 43. Then, the model control unit 22 acquires the proposed response 43 and the reasoning description 44 output by the response generation model 31.

[0058] A method for determining the difficulty will be described. The response generation model 31 can determine the difficulty based on the likelihood that the information has been changed. Therefore, the model control unit 22 may instruct the model control unit 22 to determine the difficulty based on the likelihood that the information has been changed. For example, suppose an inquiry is made regarding a route to a destination station. At this time, suppose the response generation model 31 uses the response history 46 to generate a proposed response 43 that includes transfer stations, etc., as well as the departure time of the train to be boarded. Here, if there are changes to the train schedule due to the influence of a typhoon, etc., it is highly likely that the train departure time has also changed. Therefore, in this case, the response generation model 31 generates a response generation model 31 that distinguishes between transfer stations, etc., and the departure time of the train to be boarded. For example, classification means highlighting parts of the information that are likely to have been changed to draw attention to them. Alternatively, for parts of the information that are likely to have been changed, information after the change may be separately included in the response plan 43. As a specific example, if a change has occurred in the train schedule and a new temporary schedule has been obtained, information about the new schedule may be included in the response plan 43.

[0059] Note that when there is a change in facilities or a change in the bus schedule, the response generation model 31 may learn that a change has occurred. This makes it easier for the response generation model 31 to determine whether or not there is a high possibility that information has been changed. Furthermore, as described in the third embodiment, there is a case where the response plan 43 is generated using the response history 46. In this case, for example, it becomes easier to determine that there is a high possibility that information related to the bus schedule has been changed in the response plan 43 that uses the response history 46 before the change in the bus schedule.

[0060] The response generation model 31 may also be trained using training data in which parts that are not universal and may be subject to change are explicitly labeled, making it easier for the response generation model 31 to determine whether or not the information is likely to have changed.

[0061] The response generation model 31 may determine that the difficulty is high when sufficient information is not available, regardless of the likelihood that the information has been changed.

[0062] ***Effects of the Fifth Embodiment*** As described above, the response support device 10 according to the fifth embodiment generates the response plan 43 by separating the parts for which it is difficult to generate the response plan 43. This prompts the operator to check the contents of the response plan 43, thereby preventing the user from giving an incorrect response.

[0063] ***Other Configurations*** <Variation 1> In the fifth embodiment, the response generation model 31 determines the difficulty of generating a response plan 43. However, a functional component that determines the difficulty of generating a response plan 43 may be provided separately.

[0064] The configuration of the response support device 10 according to the first modification will be described with reference to FIG. 1 in that the response support device 10 includes a classification control unit 24 as a functional component. The function of the classification control unit 24 is realized by software, similar to the other functional components.

[0065] The operation of the response support device 10 according to the first modification will be described with reference to FIGS. The process of step S61 is the same as the process of step S51 in FIG.

[0066] (Step S62: Model control processing) As in the first embodiment, the model control unit 22 inputs inquiry information 41 to the response generation model 31. Then, the model control unit 22 acquires the response proposal 43 and the rationale description 44 output by the response generation model 31. In other words, the response generation model 31 generates the response proposal 43 that is not classified by difficulty.

[0067] (Step S63: Segmentation control process) The classification control unit 24 classifies the response proposals 43 acquired in step S62 according to the difficulty of automatically generating the response proposals 43. The method of determining the difficulty is the same as that used by the response generation model 31, and the determination is made based on the likelihood that the information has been changed, etc.

[0068] (Step S64: Output processing) The output unit 23 outputs the rationale description 44 acquired in step S62 and the response proposal 43 classified in step S63.

[0069] Embodiment 6 The sixth embodiment differs from the third embodiment in that the response generation model 31 is trained using the response history 46. In the sixth embodiment, this difference will be explained, and explanation of the same points will be omitted.

[0070] ***Configuration Description*** The configuration of a response support device 10 according to the sixth embodiment will be described with reference to FIG. 1 in that the response assistance device 10 includes a learning unit 25 as a functional component. The function of the learning unit 25 is realized by software, similar to the other functional components.

[0071] ***Explanation of Operation*** The operation of the response support device 10 according to the sixth embodiment will be described with reference to FIGS. The processing from step S71 to step S73 is the same as the processing from step S31 to step S33 in FIG.

[0072] (Step S74: Learning process) The learning unit 25 uses, as learning data, a pair of the inquiry information 41 input in step S71 and the response given by the operator to the user in response to the proposed response 43 output in step S73, and causes the response generation model 31 to learn the pair. In other words, the learning unit 25 causes the response generation model 31 to learn the newly obtained response history 46 as learning data. The content of the response from the operator to the user is acquired from the conversation between the operator and the user. In this case, if a description in the station service manual 42 that is the basis for the content of the operator's response is identified, the learning unit 25 may also include the description that is the basis in the learning data.

[0073] In this example, the response generation model 31 is trained every time an inquiry is received. However, pairs of inquiry information 41 and response contents for a certain period of time may be accumulated as training data, and the response generation model 31 may be trained all at once.

[0074] ***Effects of the Sixth Embodiment*** As described above, the response support device 10 according to the sixth embodiment uses the content of the response given by the operator to the inquiry to train the response generation model 31. As a result, the more the operator responds, the more appropriate the response proposal 43 and the rationale description 44 the response generation model 31 can output.

[0075] Embodiment 7 The seventh embodiment differs from the first to sixth embodiments in that it selects a query to be dealt with preferentially. In the seventh embodiment, this difference will be explained, and explanation of the same points will be omitted. In the seventh embodiment, a case where a modification is made to the first embodiment will be described. However, modifications can also be made to the second to sixth embodiments.

[0076] There are cases where a large number of inquiries are made from multiple stations. The number of operators is limited, and users may have to wait for an operator to be assigned. In this case, it is necessary to appropriately allocate the inquiries to operators. In the seventh embodiment, a method will be described for selecting inquiries to be handled with priority and appropriately allocating the inquiries to operators.

[0077] ***Configuration Description*** The configuration of the response support device 10 according to the seventh embodiment will be described with reference to FIG. 1 in that the response support device 10 includes an inquiry distribution unit 26 as a functional component. The function of the inquiry distribution unit 26, like the other functional components, is realized by software.

[0078] ***Explanation of Operation*** The operation of the response support device 10 according to the seventh embodiment will be described with reference to FIG. The process shown in FIG. 19 is executed as a preliminary step to the process shown in FIG.

[0079] (Step S81: Selection information reception process) The input unit 21 receives input of selection information. The selection information includes, for each currently occurring inquiry, the language used in the inquiry, the category of the inquiry content, and the response waiting time. The selection information also includes the idle time of each agent. The language used for the inquiry and the category of the inquiry are input by the user before making the inquiry from a remote ticket window terminal 52 or the like installed at a ticket window in the station. The language used for the inquiry and the category of the inquiry may be identified from the recognition results of speech recognition of the user's inquiry. The response waiting time is input by counting the elapsed time since the inquiry was made using a timer. The idle time is the time that has elapsed since the operator finished responding to the inquiry, which is counted by a timer and entered.

[0080] (Step S82: Urgency Degree Identification Processing) The inquiry distribution unit 26 determines the level of urgency for each currently occurring inquiry from the category of the inquiry content included in the selection information received in step S81. Here, it is assumed that the level of urgency is determined in advance for each category.

[0081] (Step S83: Priority determination process) The inquiry distribution unit 26 determines the priority of each currently occurring inquiry based on the urgency determined in step S82 and the response waiting time included in the selection information. The priority indicates the degree to which the operator will preferentially respond to the inquiry. For example, the inquiry distribution unit 26 converts the urgency and the response waiting time into evaluation values. Here, the higher the urgency, the higher the evaluation value. Also, the longer the response waiting time, the higher the evaluation value. The inquiry distribution unit 26 determines the priority for each inquiry so that the larger the sum of the evaluation value calculated from the urgency and the evaluation value calculated from the response waiting time, the higher the priority.

[0082] (Step S84: Operator identification process) The inquiry distribution unit 26 identifies an operator who can handle each currently occurring inquiry based on the language used in the inquiry and the category of the inquiry content included in the selection information. Here, it is assumed that an operator capable of handling each language used in the inquiry is determined in advance, and an operator capable of handling each category of the inquiry content is determined in advance. For each currently occurring inquiry, the inquiry distribution unit 26 identifies an operator capable of handling both the language used in the inquiry and the category of the inquiry content. For example, for each station staff member, the station staff member's name and ID, work experience information, and skill information are stored in storage 13 as station staff information. The work experience information indicates the categories that the station staff member has handled in the past. For example, the work experience information indicates that the station staff member has experience handling transportation ICs and tourism-related matters. The skill information indicates the languages ​​that can be spoken. The inquiry distribution unit 26 identifies operators who can handle each category of inquiry content by referring to the work experience information. The inquiry distribution unit 26 identifies operators who can handle the language used in the inquiry by referring to the skill information.

[0083] (Step S85: Allocation process) The inquiry allocating unit 26 sorts the operators in descending order of idle time. Then, the inquiry allocating unit 26 sets each operator as a target operator in the sorted order. Of the inquiries identified in step S84 as operators that the target operator can handle, the inquiry allocating unit 26 allocates the inquiry with the highest priority determined in step S83 to the target operator.

[0084] (Step S86: Confirmation process) The inquiry distribution unit 26 notifies the operator of the inquiry distributed in step S85, and causes the operator to start responding to the inquiry. 20, a list screen of currently occurring inquiries is displayed on the operator terminal 51 used by each operator. The inquiry distribution unit 26 displays a confirmation button in the column of the inquiry distributed to a certain operator on the list screen displayed on the operator terminal 51 of that operator, and notifies the operator of the allocated inquiry. When the operator operates the operator terminal 51 and presses the confirmation button, a response to the inquiry begins.

[0085] That is, the inquiry distribution unit 26 treats each inquiry as a target inquiry, identifies an operator who can handle the target inquiry, and selects an operator to handle the target inquiry from the operators who can handle the target inquiry. During this selection, the inquiry distribution unit 26 selects the target operator in descending order of idle time, and distributes to the target operator the inquiries with the highest priority among the inquiries that the target operator can handle. Then, the inquiry distribution unit 26 displays a confirmation button in the column of the inquiry distributed to the target operator on the inquiry list screen displayed on the operator terminal 51 used by the target operator.

[0086] Here, the inquiry distribution unit 26 displays a confirmation button in the column of the inquiry with the highest priority. However, the inquiry distribution unit 26 may display a confirmation button in the column of one or more inquiries with a priority higher than a preset standard, and allow the operator to select one of the inquiries as the corresponding inquiry.

[0087] ***Effects of the Seventh Embodiment*** As described above, the response support device 10 according to the seventh embodiment selects inquiries to be dealt with preferentially and appropriately assigns the inquiries to operators, thereby improving the user's satisfaction with the response to the inquiries.

[0088] ***Other Configurations*** <Variation 2> In the seventh embodiment, in step S85, the inquiry distribution unit 26 distributes the inquiry only to the operators identified as available in step S84. However, if there are no available operators, the inquiry distribution unit 26 may distribute the inquiry to operators who are not identified as available.

[0089] In the seventh embodiment, it is assumed that an operator who can handle each inquiry category is determined in advance. In this case, it is possible to determine that an operator can handle a category for which the operator has handled a standard amount of inquiries in the past. By allocating inquiries as in the second modification, it is also possible to gradually increase the number of categories that an operator can handle.

[0090] <Variation 3> For some inquiries, fixed answers 32 may be prepared so that the answer can be automatically given without the intervention of an operator. In this case, as shown in Fig. 21, the response support device 10 includes a response determination unit 27 as a functional component. The response determination unit 27 recognizes the content of the inquiry information 41 received by the input unit 21. The method for recognizing the content of the inquiry information 41 may be any method, such as using AI. The response determination unit 27 determines whether the content of the recognized inquiry information 41 is an inquiry for which a fixed answer 32 is prepared. If the content of the inquiry information 41 is an inquiry for which a fixed answer 32 is prepared, the response determination unit 27 outputs the fixed answer 32 corresponding to the content of the inquiry information 41 to the remote counter terminal 52, etc. On the other hand, if the content of the inquiry information 41 is not an inquiry for which a fixed answer 32 is prepared, the response determination unit 27 causes an operator to handle the inquiry. Specifically, the response determination unit 27 causes the inquiry allocation unit 26 to assign an operator.

[0091] If the content of the inquiry information 41 is not an inquiry for which a fixed answer 32 is prepared, the response determination unit 27 may input the inquiry information 41 to the model control unit 22 to acquire a proposed response 43. If the reliability of the proposed response 43 is higher than a reference value, the response determination unit 27 outputs the proposed response 43 to the remote counter terminal 52 or the like. On the other hand, if the reliability of the proposed response 43 is equal to or lower than the reference value, the response determination unit 27 causes the inquiry allocation unit 26 to assign an operator.

[0092] The response determination unit 27 automatically provides the operator with a response for the inquiry content for which there is no fixed response 32. The response determination unit 27 may input a determination result as to whether or not an automatic response was possible. If it is input that an automatic response is possible, the response determination unit 27 may add the response by the operator to the storage 13 as a fixed response 32 for the inquiry content of the inquiry information 41. This allows the fixed responses 32 to be gradually enriched.

[0093] The response generation model 31 may have an operator input an evaluation of the response plan 43. The response generation model 31 can improve the accuracy of calculating the reliability of the response plan 43 by learning the response plan 43 and the evaluation by the operator. This can improve the accuracy of the response determination unit 27 in determining whether to automatically respond with the response plan 43 or to have an operator respond.

[0094] <Variation 4> A terminal for making an inquiry may be prepared for each purpose of use. In this case, the inquiry distribution unit 26 determines the priority from the terminal that has sent the inquiry. For example, a terminal that can be used only for categories that require an inquiry urgently may be prepared. In this case, the inquiry distribution unit 26 does not determine the urgency based on the category, and treats inquiries from this terminal as having a high urgency. The inquiry distribution unit 26 then determines the priority so that the priority is high. This makes it possible to determine the priority simply and appropriately. Furthermore, by preparing a terminal that can be used only for categories that require an inquiry urgently, it is possible to prevent an inquiry that is urgently needed from being delayed because a terminal is not available.

[0095] <Variation 5> In the seventh embodiment, the longer the response waiting time, the higher the evaluation value and the higher the priority. It is also possible to set a higher priority for inquiries from terminals that are expected to have a long response waiting time. Specifically, the inquiry distribution unit 26 calculates the priority as described in the seventh embodiment, and then corrects the priority to be higher for inquiries from terminals that are expected to have a long response waiting time.

[0096] At this time, the inquiry distribution unit 26 predicts the response waiting time for each terminal. Specifically, the inquiry distribution unit 26 predicts the response waiting time for each terminal from the length of the queue or the number of people waiting. For example, the inquiry distribution unit 26 determines a unit waiting time for each unit length of the queue, and predicts the response waiting time from the queue length and the unit waiting time. Alternatively, the inquiry distribution unit 26 determines a unit waiting time per person, and predicts the response waiting time from the number of people waiting and the unit waiting time. Here, the input unit 21 receives images captured by a camera around the terminal. The inquiry allocator 26 identifies the length of the queue or the number of people waiting for processing at the terminal from the images.

[0097] The inquiry allocating unit 26 may predict the response waiting time from the attributes of the people waiting, in addition to the length of the queue or the number of people waiting. For example, if the people waiting include someone who needs assistance, the inquiry allocating unit 26 may predict the response waiting time by assuming that it will take longer for that person to be served than for others. Furthermore, if the category or content of the inquiry of the person waiting can be identified, the inquiry distribution unit 26 may predict the response waiting time taking into consideration the category or content of the inquiry of the person waiting. For example, it is conceivable that a weight is defined for each category or content of the inquiry, and the inquiry distribution unit 26 predicts the response waiting time by multiplying the unit waiting time by the weight.

[0098] In addition, when predicting the response waiting time from only the queue length, For example, if the queue is less than 2 meters, the priority will not be adjusted, and if the queue is between 2 meters and 5 meters, the priority will be increased by one level, etc. Similarly, when predicting response waiting time based only on the number of waiting people, it can be said that priority is determined based on the number of waiting people. For example, if the number of waiting people is less than two, the priority may not be adjusted, if the number of waiting people is between two and five, the priority may be increased by one level, etc.

[0099] <Variation 6> The priority may be increased if the inquiry is from a user who has a ticket and can be given preferential treatment. Specifically, the inquiry distribution unit 26 calculates the priority as described in the seventh embodiment, and then corrects the priority to be higher if the inquiry is from a user who has a ticket. For example, tickets that allow inquiries to be handled on a priority basis can be sold for a fee, and inquiries from users who purchase the tickets can be handled on a priority basis. The tickets can be recorded on IC cards, and when the IC card with the ticket recorded on it is held over a terminal, the inquiry distribution unit 26 corrects the priority to be higher. This makes it possible for inquiries to be handled on a priority basis by purchasing a ticket if you are in a hurry.

[0100] Embodiment 8 The eighth embodiment differs from the first to seventh embodiments in that it estimates the emotion of at least one of the user and the operator. In the eighth embodiment, this difference will be explained, and explanation of the same points will be omitted. In the eighth embodiment, a case where a modification is made to the first embodiment will be described. However, modifications can also be made to the second to seventh embodiments.

[0101] ***Configuration Description*** The configuration of the response support device 10 according to the eighth embodiment will be described with reference to FIG. 1 in that the response support device 10 includes, as a functional component, an emotion deduction unit 28. The function of the emotion deduction unit 28 is realized by software, similar to the other functional components.

[0102] ***Explanation of Operation*** The operation of the response support device 10 according to the eighth embodiment will be described with reference to FIG. The process shown in Fig. 23 is a process executed as a subsequent stage of the process shown in Fig. 2. The process shown in Fig. 23 may be executed after a certain period of time, such as one day, has passed, with all of the processes shown in Fig. 2 that have been executed during that period as targets. For example, the process shown in Fig. 23 may be executed when an operator finishes his or her work for the day, with each of the processes shown in Fig. 2 that have been executed by that operator that day as targets.

[0103] (Step S91: Conversation information reception process) The input unit 21 accepts input of conversation information between a user and an operator. The conversation information includes the conversation between the user and the operator, which is the inquiry information 41 accepted in step S11 of Fig. 2, and information on the conversation consisting of the response made by the operator by referring to the proposed response 43 etc. in step S13 of Fig. 2 and the user's reply to the response. When the user and the operator are having a conversation by voice, the input unit 21 converts the conversation information into text.

[0104] (Step S92: User emotion determination process) The emotion estimation unit 28 determines whether the emotion of the user is in a positive state or a negative state from the conversation information received in step S91. Specifically, the emotion deduction unit 28 determines whether the user's emotion is positive or negative based on the content of the conversation information uttered by the user and, in the case of a voice conversation, the tone and speed of the user's voice. In this case, the emotion deduction unit 28 may assign a score to the user's emotion instead of determining whether the user's emotion is positive or negative. For example, the emotion deduction unit 28 assigns a score such that the more positive the user's emotion, the higher the score, and the more negative the user's emotion, the lower the score. In this case, the emotion estimation unit 28 may control the score by referring to modifiers of the words uttered by the user. For example, if the word of gratitude is accompanied by the modifier "very," it may be possible to assign a higher score. As a specific example, it may be possible to assign a higher score to the words "Thank you very much. You were very helpful" than to the words "Thank you very much. You were very helpful."

[0105] (Step S93: Evaluation output process) The feeling estimation unit 28 outputs the feeling of the user determined in step S92. Specifically, when the user's emotion is positive, the emotion deduction unit 28 outputs the fact that the emotion was positive to the operator terminal 51. Note that when a score is assigned to the user's emotion, the emotion deduction unit 28 may output the score. The emotion deduction unit 28 may convert the score into a star mark representing the user's satisfaction level, and output and display it on the operator terminal 51. At this time, the emotion deduction unit 28 may also output words of gratitude, etc., uttered by the user to the operator terminal 51 and display them. Furthermore, when the user's emotion is negative, the emotion deduction unit 28 outputs to the storage 13 the fact that the emotion was negative, along with the identification information of the operator and the category of the inquiry. Note that when a score is assigned to the user's emotion, the emotion deduction unit 28 may output the score. The information output to the storage 13 is checked by the operator's administrator at any time and used for training the operators, etc. In other words, categories that each operator is weak at are identified, and guidance on appropriate response methods is provided.

[0106] (Step S94: Operator emotion determination process) The emotion estimation unit 28 determines whether the emotion of the operator is in a positive state or a negative state from the conversation information received in step S91. Specifically, the emotion deduction unit 28 determines whether the emotion of the operator is positive or negative based on the content of the conversation information uttered by the operator and, in the case of a voice conversation, the tone and speed of the operator's voice. In this case, the emotion deduction unit 28 may assign a score to the emotion of the operator instead of determining whether the emotion of the operator is positive or negative. For example, the emotion deduction unit 28 assigns a score such that the more positive the emotion of the operator, the higher the score, and the more negative the emotion, the lower the score.

[0107] (Step S95: Stress output processing) The emotion estimation unit 28 quantifies the stress of the operator, the degree of attachment to the work, and the like from the emotion of the operator determined in step S94, and outputs the quantified values ​​together with the identification information of the operator to the storage 13. The information output to the storage 13 is confirmed by the operator's manager at any time and is used for the care of the operator, personnel allocation, and the like.

[0108] ***Effects of the eighth embodiment*** As described above, the response support device 10 according to the eighth embodiment notifies the operator of the user's emotion when the user's emotion is positive. When a user finishes their work for the day, the system notifies the operator of the user's positive emotions, allowing the operator to finish their work in a good mood and increase their motivation to work.

[0109] Furthermore, when the user's emotion is negative, the response support device 10 according to the eighth embodiment outputs the user's emotion together with the agent's identification information and the inquiry category to the storage 13. This allows categories that each agent is not good at handling to be identified, making it possible to provide guidance to the agent on appropriate response methods.

[0110] Furthermore, the response support device 10 according to the eighth embodiment quantifies the stress of the operator, the degree of attachment to the work, and the like from the emotions of the operator, and outputs the quantified values ​​to the storage 13. This makes it possible to appropriately care for the operators and allocate personnel.

[0111] ***Other Configurations***

[0112] <Variation 7> The emotion deduction unit 28 identifies an operator whose estimated user emotions are more positive than the model standard relative to the number of users who responded. Then, the emotion deduction unit 28 raises the evaluation of the identified operator. That is, the evaluation of the operator is increased. Furthermore, the emotion deduction unit 28 extracts, from the conversation information of the identified operator, conversation information when the user's emotion is estimated to be positive, as model response information. The emotion deduction unit 28 allows other operators to refer to the extracted model response information. For example, the emotion deduction unit 28 stores the model response information so that other operators can refer to it as needed.

[0113] The emotion estimation unit 28 identifies an operator whose estimated user emotions are negative at a rate higher than the evaluation standard, relative to the number of users who have responded. Then, the emotion estimation unit 28 lowers the evaluation of the identified operator. In other words, the evaluation of the operator is lowered.

[0114] <Variation 8> In some cases, the same user may make multiple inquiries, and the operator's feelings may be presumed to be negative. In such cases, the user may be registered, and inquiries from this user may be assigned to a specific operator, such as a veteran operator. Specifically, the emotion deduction unit 28 identifies a user whose estimated operator emotion is negative more times than a reference number of times as a specific user. Then, the emotion deduction unit 28 registers the specific user in the storage 13. For example, when a specific user makes an inquiry by holding an IC card or the like, the emotion deduction unit 28 registers the identification information included in the IC card as the identification information of the specific user. When an inquiry is made by a user whose identification information is registered in the storage 13, the inquiry allocation unit 26 determines that the inquiry is from a specific user. Then, the inquiry allocation unit 26 allocates the inquiry from the specific user to a specific operator. This makes it possible to respond appropriately even to users who are difficult to deal with. As a result, it is possible to avoid making both users and operators feel uncomfortable. As a result of avoiding users feeling uncomfortable, users' satisfaction with the railway operator increases, leading to an increase in the rate of repeat use. Furthermore, as a result of avoiding operators feeling uncomfortable, it is possible to increase operator satisfaction with their work, leading to a decrease in turnover.

[0115] <Variation 9> In the sixth embodiment, the response history 46 is used as learning data. In addition to this, the learning unit 25 may include the user's feelings toward the response from the operator in the learning data. The learning unit 25 can learn the content of the response as correct data if the emotion is positive, while the learning unit 25 can learn the content of the response as incorrect data if the emotion is negative.

[0116] The contents described in the fourth and fifth embodiments may be combined. Specifically, the contents described in the fourth embodiment may be combined to add the classification 47 to the training data. That is, the learning unit 25 may cause the response generation model 31 to learn the inquiry information 41, the classification 47, the content of the response, and the user's reaction as training data. Furthermore, the contents described in the fifth embodiment may be combined to cause the response generation model 31 to learn the inquiry information 41, the classification instruction 48, the content of the response, and the user's reaction as training data.

[0117] Embodiment 9 The ninth embodiment differs from the first to eighth embodiments in that it evaluates equipment such as the remote counter terminal 52 provided by the railway operator and used by users. In the ninth embodiment, this difference will be explained, and explanations of the same points will be omitted. In the ninth embodiment, a case where a modification is made to the first embodiment will be described. However, modifications can also be made to the second to eighth embodiments.

[0118] ***Configuration Description*** The configuration of a response support device 10 according to the ninth embodiment will be described with reference to FIG. 1 in that the response support device 10 includes a device evaluation unit 29 as a functional component. The function of the device evaluation unit 29 is realized by software, similar to the other functional components.

[0119] ***Explanation of Operation*** The device evaluation unit 29 records the usage status of devices such as the remote service terminal 52 that is the input source of the inquiry information 41. The usage status includes the user's attributes, the number of uses, and the usage time. Furthermore, the device evaluation unit 29 evaluates the user's satisfaction with the device used. The device evaluation unit 29 determines whether the user gave a positive or negative evaluation, and uses this as the user's satisfaction level. Specifically, the device evaluation unit 29 uses the user's voice or camera image to determine whether the user gave a positive or negative evaluation. At this time, the device evaluation unit 29 may record keywords included in the user's comments related to the evaluation. Furthermore, the device evaluation unit 29 may prompt the user to input an evaluation of the device after using it. In response to a request from the administrator of the response support device 10, the device evaluation unit 29 outputs and displays the record of the usage status and the satisfaction level with the device on the display device.

[0120] Here, the user's voice is, for example, the user's voice included in the conversation information between the user using the device to be evaluated and an operator responding to the user's inquiry, and the camera image is an image of the user using the device to be evaluated during the conversation between the user and the operator responding to the user's inquiry.

[0121] The equipment evaluation unit 29 may also output and display keywords contained in the user's comments related to the evaluation. The equipment evaluation unit 29 may display information indicating the equipment to be evaluated and keywords contained in the user's comments related to the evaluation so that other users can see them. For example, the equipment evaluation unit 29 may display the information indicating the equipment and the keywords on a display device installed in a place such as a station platform or passageway. When displaying information on a display device that can be seen while riding an escalator, the equipment evaluation unit 29 may display the information indicating the equipment and the keywords in accordance with the movement of the escalator. The display may also be moved to make it easier for users to see. Informing other users of a user's positive feedback leads to advertising for the device. Informing other users of a user's negative feedback also informs other users that the device is likely to be improved. The device evaluation unit 29 may identify the attributes of the person viewing the display device from an image obtained by a camera installed near the display device. The device evaluation unit 29 may then switch the display depending on the identified attributes. For example, if the attribute is a child, the device evaluation unit 29 may switch to using more hiragana characters so that children can easily read it. Also, if the attribute is an elderly person, the device evaluation unit 29 may also display an explanation of the device and how to use it so that the device can be easily understood. The device evaluation unit 29 may output in voice or sign language for people with visual or hearing impairments.

[0122] ***Effects of the 9th embodiment*** As described above, the response support device 10 according to the ninth embodiment records usage status and evaluates satisfaction levels, and displays them as requested. By comparing and displaying usage status records and satisfaction levels for different terminals in different locations, the response support device 10 can assist railway companies in determining which equipment to introduce and what improvements to make. For example, the response support device 10 can assist in determining whether there are enough measures in place for vulnerable road users, whether equipment is installed in a bad location, or whether there is a lack of information, such as equipment manuals.

[0123] ***Other Configurations*** <Modification 10> In the ninth embodiment, the evaluation was performed on devices provided by railway operators and used by users. However, the evaluation may also be performed on services provided by railway operators and used by users. Examples of such services include a children's free ride service and a book lending service. When a service is to be evaluated, it is possible to identify which service the conversation is about from the content of the conversation between the user and the operator. For example, the device evaluation unit 29 inputs the conversation information between the user and the operator into an evaluation model, which is a learning model, and causes the evaluation model to identify which service the conversation is about.

[0124] <Variation 11> The device evaluation unit 29 may input the user's comments and camera images into an evaluation model, which is a learning model, and obtain the user's satisfaction level output by the evaluation model. In this case, the device evaluation unit 29 may instruct the evaluation model to promptly present an improvement plan to improve the satisfaction level in addition to evaluating the user's satisfaction level, and obtain the user's satisfaction level and the improvement plan output by the evaluation model. The improvement plan is a way to make the evaluation better if the evaluation is positive, and a way to resolve the issues raised if the evaluation is negative. Furthermore, the device evaluation unit 29 may instruct the evaluation model to extract keywords related to the evaluation of satisfaction for comments with a higher satisfaction level than a first standard. Similarly, the device evaluation unit 29 may instruct the evaluation model to extract keywords related to the evaluation of satisfaction for comments with a lower satisfaction level than a second standard set lower than the first standard. Then, the device evaluation unit 29 may acquire keywords for comments with a higher satisfaction level than the first standard and keywords for comments with a lower satisfaction level than the second standard. Furthermore, the device evaluation unit 29 may use, as an evaluation model, instructions such as outputting keywords for high evaluations in itemized form, outputting keywords for low evaluations in itemized form, etc. Also, the device evaluation unit 29 may use, as an evaluation model, instructions such as outputting the ratio of high evaluations to low evaluations and the reasons for each evaluation together.

[0125] <Modification 12> In the ninth embodiment, the device evaluation unit 29 evaluates the device, but the device evaluation unit 29 may evaluate the user's satisfaction with the content of an automatically output response, such as the automatic response using the fixed answer 32 described in the third modification. For example, the device evaluation unit 29 may determine the user's satisfaction with the automatic response from the user's voice or camera image after the response. In other words, when a response is made by a chatbot, the device evaluation unit 29 may determine the user's satisfaction with this response.

[0126] <Variation 13> In the above embodiment, each functional component is realized by software. However, as a sixth modification, each functional component may be realized by hardware. The following describes the differences between this sixth modification and the abnormality embodiment.

[0127] When each functional component is realized by hardware, the response support device 10 includes an electronic circuit instead of the processor 11, the memory 12, and the storage 13. The electronic circuit is a dedicated circuit for realizing the functions of each functional component, the memory 12, and the storage 13.

[0128] Possible electronic circuits include single circuits, composite circuits, programmed processors, parallel programmed processors, logic ICs, GAs, ASICs, and FPGAs. GA stands for Gate Array. ASIC stands for Application Specific Integrated Circuit. FPGA stands for Field-Programmable Gate Array. Each functional component may be realized by one electronic circuit, or each functional component may be realized by distributing it among a plurality of electronic circuits.

[0129] <Variation 14> As a seventh modification, some of the functional components may be realized by hardware, and other functional components may be realized by software.

[0130] The processor 11, memory 12, storage 13, and electronic circuitry are collectively referred to as a processing circuit. In other words, the functions of the functional components are realized by the processing circuit.

[0131] Furthermore, the term "unit" in the above description may be read as a "circuit," "step," "procedure," "process," or "processing circuit."

[0132] Various aspects of the present disclosure are summarized below as appendices. (Appendix 1) a model control unit that inputs inquiry information from users at station counters into a response generation model, which is a learning model given a station work manual, and acquires a proposed response output by the response generation model; an output unit that outputs the response plan acquired by the model control unit; A response support device comprising: (Appendix 2) The model control unit further acquires a description of the basis for the response plan in the station work manual, The output unit further outputs the grounds description. 2. The response assistance device of claim 1. (Appendix 3) The model control unit further inputs counter information indicating the location of the counter into the response generation model. 3. A response assistance device according to claim 1 or 2. (Appendix 4) The model control unit further inputs train operation information into the response generation model. 4. A response support device according to any one of claims 1 to 3. (Appendix 5) The model control unit further inputs train fare information into the response generation model. 5. A response assistance device according to any one of claims 1 to 4. (Appendix 6) The model control unit further inputs device information managed by devices installed in the station and available to the user into the response generation model. 6. A response assistance device according to any one of appendices 1 to 5. (Appendix 7) 7. The response support device according to claim 1, wherein the model control unit further inputs personal information of the user into the response generation model. (Appendix 8) The model control unit further inputs ticket information about a ticket held by the user into the response generation model. 8. A response assistance device according to any one of appendices 1 to 7. (Appendix 9) The model control unit further inputs lost item information about the lost item lost by the user into the response generation model. 9. A response assistance device according to any one of appendices 1 to 8. (Appendix 10) The model control unit further inputs map information of the area around the station into the response generation model. 10. A response assistance device according to any one of appendices 1 to 9. (Appendix 11) The model control unit further inputs event information about events taking place around the station into the response generation model. 11. A response assistance device according to any one of appendices 1 to 10. (Appendix 12) The response generation model is given a response history, which is a set of past inquiry information and response contents to the inquiry information. 12. A response assistance device according to any one of claims 1 to 11. (Appendix 13) The model control unit inputs a classification based on at least one attribute of the station, the user, and the inquiry information into the response generation model, and acquires the proposed response generated by the response generation model on the basis of the classification. 13. A response assistance device according to any one of appendices 1 to 12. (Appendix 14) The model control unit inputs, to the response generation model, an instruction to classify the response plans according to the difficulty of automatically generating the response plans, and acquires the response plans generated by the response generation model, the response plans classified according to the difficulty. 14. A response assistance device according to any one of claims 1 to 13. (Appendix 15) The response assistance device further comprises: a classification control unit that classifies the response proposals acquired by the model control unit according to the difficulty of automatically generating the response proposals; Equipped with The output unit outputs the response plan classified by the classification control unit. 14. A response assistance device according to any one of claims 1 to 13. (Appendix 16) The response assistance device further comprises: a learning unit that causes the response generation model to learn, as learning data, a response history that is a set of inquiry information, the proposed response output by the output unit, and the response content given by an operator to the user in response to the description of the basis; 16. A response support device according to any one of appendices 1 to 15, comprising: (Appendix 17) The response assistance device further comprises: an inquiry distribution unit that determines the priority of an inquiry based on the category to which the content of the inquiry information belongs and the response waiting time of the user, and that gives priority to inquiries with high priority and has the operator handle the inquiries; 17. A response support device according to any one of claims 1 to 16, comprising: (Appendix 18) The response assistance device further comprises: an emotion estimation unit that estimates the emotion of the user from a reply from the user to a response from an operator corresponding to the inquiry information; 18. A response support device according to any one of appendices 1 to 17, comprising: (Appendix 19) The response assistance device further comprises: an input unit that receives at least a statement from a user who uses at least one of equipment and services provided by a railway operator as an evaluation target and an operator who responds to an inquiry from the user; a device evaluation unit that determines the user's level of satisfaction with the evaluation target used by the user based on the user's comments received by the input unit; 19. A response support device according to any one of appendices 1 to 18, comprising: (Appendix 20) A computer inputs inquiry information from a user at a station ticket counter into a response generation model, which is a learning model given a station work manual, and obtains a proposed response output by the response generation model. A response support method in which a computer outputs the response proposal. (Appendix 21) A model control process inputs inquiry information from users at station counters into a response generation model, which is a learning model given a station work manual, and acquires a proposed response output by the response generation model; an output process for outputting the response plan acquired by the model control process; A response support program that causes a computer to function as a response support device that performs the following.

[0133] The embodiments and modifications of the present disclosure have been described above. Some of these embodiments and modifications may be combined and implemented. Also, one or more of them may be implemented partially. Note that the present disclosure is not limited to the above embodiments and modifications, and various modifications are possible as needed. [Explanation of symbols]

[0134] 10 Response support device, 11 Processor, 12 Memory, 13 Storage, 14 Communication interface, 21 Input unit, 22 Model control unit, 23 Output unit, 24 Classification control unit, 25 Learning unit, 26 Inquiry distribution unit, 27 Response determination unit, 28 Emotion estimation unit, 29 Equipment evaluation unit, 31 Response generation model, 41 Inquiry information, 42 Station work manual, 43 Response proposal, 44 Reason description, 45 Additional information, 46 Response history, 47 Classification, 48 Classification instruction, 51 Operator terminal, 52 Remote counter terminal.

Claims

1. an input unit that receives conversation information between a station user and an operator who responds to an inquiry from the user; an emotion estimation unit that estimates the emotion of the user at the time of the operator's response from the conversation information received by the input unit, and outputs the estimated emotion of the user to an operator terminal used by the operator if the emotion of the user is positive; A response support device comprising:

2. When the user expresses gratitude, the feeling estimation unit outputs the gratitude to the operator terminal. The response assistance device according to claim 1 .

3. When the user's emotion is a negative emotion, the emotion estimation unit records the user's emotion, the identification information of the operator, and the category of the content of the inquiry. The response assistance device according to claim 1 .

4. the input unit receives the conversation information as voice data, The emotion estimation unit estimates the emotion of the user from the words uttered by the user included in the conversation information and the tone and speed of the voice when the words are uttered. The response assistance device according to claim 1 .

5. The emotion estimation unit causes other agents to refer to the conversation information of an agent in which the estimated user's emotion is more likely to be positive than a model standard as model response information. The response assistance device according to claim 1 .

6. The emotion estimation unit lowers the evaluation of an agent whose estimated user emotion is negative at a rate higher than an evaluation standard. The response assistance device according to claim 1 .

7. The emotion estimation unit estimates the emotion of the operator from the conversation information and records the estimated emotion of the operator and identification information of the operator. The response assistance device according to claim 1 .

8. the emotion estimation unit registers, as a specific user, a user whose estimated emotion is negative more than a reference number of times; The response assistance device further comprises: An inquiry distribution unit that distributes inquiries from the specific users to specific operators The response assistance device according to claim 7 , comprising:

9. The computer receives conversation information between a station user and an operator who responds to an inquiry from the user, A response support method in which a computer estimates the user's emotions at the time the operator responds from the conversation information, and if the user's emotions are positive, outputs the estimated user's emotions to an operator terminal used by the operator.

10. an input process for receiving conversation information between station users and operators who respond to inquiries from the users; an emotion estimation process of estimating the emotion of the user at the time of the operator's response from the conversation information received by the input process, and outputting the estimated emotion of the user to an operator terminal used by the operator if the emotion of the user is positive; A response support program that causes a computer to function as a response support device that performs the following.

Images