Information processing device, information processing system and program

Abstract

To provide an information processing device capable of obtaining a plurality of opinions for an interaction content.SOLUTION: An information processing server 1 includes: a page output unit 11 that outputs an interactive screen including an input region; an interactive content reception unit 13 that receives an interactive content to the input region included in the output interactive screen; a model processing unit 16 that sequentially performs input based on the interactive content received by the interaction content reception unit 13 to a plurality of learning models each having different expert knowledge to acquire a reply for each virtual expert corresponding to each expert knowledge from each learning model; and a page update unit 17 that outputs the reply acquired by the model processing unit 16 on the interactive screen. The model processing unit 16 performs input to the learning model including replies already output by the other learning models to acquire the reply from the learning model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

[0002] Conventionally, technologies have been disclosed that display answers from a learning model when a question is entered into an input field. Among these, a technology has been disclosed that uses a technology in which an AI answers a human question in an interactive format, providing an answer tailored to the user's situation (for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7304666 Summary of the Invention [Problem to be solved by the invention]

[0004] The system described in Patent Document 1 outputs answers to questions, but it is limited to one-to-one dialogue and is not capable of providing various insights. In particular, there are cases where people want to hear the opinions of multiple experts as answers to various consultations, but this system was not able to accommodate such situations.

[0005] SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide an information processing device or the like that makes it possible to obtain a plurality of opinions on the contents of a dialogue. [Means for solving the problem]

[0006] The present invention relates to an information processing device comprising a first screen output means for outputting an interactive screen including an input area, a dialogue content receiving means for receiving dialogue content for the input area in the interactive screen output by the first screen output means, a model processing means for sequentially inputting based on the dialogue content received by the dialogue content receiving means to a plurality of learning models each having different specialized knowledge, and obtaining responses from each learning model of a virtual expert corresponding to each specialized knowledge, and a second screen output means for outputting the responses obtained by the model processing means to the interactive screen, wherein the model processing means inputs the responses to the learning model, including the responses already output by other learning models, and obtains the responses from the learning models.

[0007] The information processing device may further include a specialized knowledge identification means for identifying a plurality of pieces of specialized knowledge based on the dialogue content received by the dialogue content receiving means, and the model processing means may sequentially input based on the dialogue content to a plurality of learning models each having the specialized knowledge identified by the specialized knowledge identification means, and obtain the response of a virtual expert corresponding to each piece of specialized knowledge from each learning model.

[0008] In addition, the information processing device may be provided with a learning means that learns each learning model having each specialized knowledge identified by the specialized knowledge identification means based on the content of the dialogue, and the model processing means may use the learning model learned by the learning means.

[0009] The information processing device may further include an information organization means for generating organized information that organizes the dialogue content received by the dialogue content receiving means, wherein the specialized knowledge identification means identifies a plurality of pieces of specialized knowledge based on the organized information generated by the information organization means, the model processing means inputs the organized information generated by the information organization means to a plurality of learning models each having the specialized knowledge identified by the specialized knowledge identification means, and acquires the responses of virtual experts corresponding to each piece of specialized knowledge from each learning model, and the second screen output means further outputs the organized information generated by the information organization means to the dialogue screen.

[0010] In addition, the information processing device may be provided with an interaction setting output means for outputting a setting screen regarding an interaction method, and an interaction setting acceptance means for accepting the setting of the interaction method using the setting screen, and the information organizing means may generate the organizing information by further adding information based on the setting of the interaction method accepted by the interaction setting acceptance means.

[0011] The information processing device may also include an additional dialogue receiving means for receiving additional dialogue content for the input area on the dialogue screen, wherein the specialized knowledge identification means identifies one or more of the specialized knowledge based on the additional dialogue content received by the additional dialogue receiving means, and the model processing means performs input based on the additional dialogue content received by the additional dialogue receiving means to the learning model having the specialized knowledge identified by the specialized knowledge identification means, and obtains the response of a virtual expert corresponding to the specialized knowledge from the learning model.

[0012] In the information processing device, the second screen output means may output the reply in association with specific information that can identify the specialized knowledge.

[0013] The present invention also relates to an information processing system comprising a first screen output means for outputting an interactive screen including an input area to a display device, a dialogue content receiving means for receiving from an input device dialogue content for the input area in the interactive screen output by the first screen output means, a model processing means for sequentially inputting based on the dialogue content received by the dialogue content receiving means to a plurality of learning models each having different specialized knowledge, and obtaining responses from each learning model of a virtual expert corresponding to each specialized knowledge, and a second screen output means for outputting the responses obtained by the model processing means to the interactive screen of the display device, wherein the model processing means inputs the responses to the learning model, including the responses already output by other learning models, and obtains the responses from the learning models.

[0014] The present invention also relates to a program for causing a computer to function as the information processing device. [Effects of the Invention]

[0015] The present invention can provide an information processing device and the like that makes it possible to obtain a plurality of opinions on the content of a dialogue. [Brief explanation of the drawings]

[0016] [Figure 1] 1 is a diagram illustrating an overall configuration of an interactive system according to an embodiment of the present invention and a functional block diagram of an information processing server. [Figure 2] 10 is a flowchart showing a dialogue process of the dialogue system according to the present embodiment. [Figure 3] FIG. 10 is a diagram showing an example of an interactive screen output to the terminal according to the embodiment. [Figure 4] FIG. 10 is a diagram showing an example of an interactive screen output to the terminal according to the embodiment. [Figure 5] FIG. 10 is a diagram showing an example of a setting screen output to the terminal according to the embodiment. [Figure 6] 10 is a flowchart showing a dialogue acquisition process of the information processing server according to the present embodiment. [Figure 7] 10 is a flowchart showing model processing of the information processing server according to the present embodiment. [Figure 8] 10A and 10B are diagrams showing examples of dialogue content and responses output to the terminal according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0017] Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, this is merely an example, and the technical scope of the present invention is not limited to this example. (Embodiment) [Overall configuration of the dialogue system 100] FIG. 1 is a diagram showing the overall configuration of a dialogue system 100 according to this embodiment and a functional block diagram of an information processing server 1. The dialogue system 100 (information processing system) is a system that, when a user inputs dialogue content, outputs a reply to the dialogue content for each expert (virtual expert). For example, in response to an inquiry such as a question about taxes, the dialogue system 100 can answer the user's question by outputting a reply using a learning model that has the knowledge (specialized knowledge) of an expert such as a tax accountant.

[0018] Hereinafter, an expert refers to a person with specialized knowledge in one or more specific fields. An expert may be a real person or a fictitious person. Here, the expert who responds to the dialogue content using a learning model is a computer that responds based on the content learned by the learning model, and is a virtual expert. Therefore, even if the learning model is based on a real person, the response is not provided by a real person.

[0019] The interactive system 100 shown in Fig. 1 includes an information processing server 1 (information processing device) and a terminal 4, which are communicatively connected via a communication network N. The communication network N is, for example, a communication line network such as the Internet. The communication network N may include a local area network (LAN) or a wide area network (WAN), and may be wired or wireless.

[0020] [Information Processing Server 1] The information processing server 1 is, for example, a server, but may also be a terminal such as a personal computer (PC). The information processing server 1 may be configured with one computer or multiple computers. When multiple computers are used, these computers are connected via, for example, a communication network N. The information processing server 1 may also be configured as a virtual server (virtual machine) provided on, for example, a cloud.

[0021] The information processing server 1 includes a control unit 10, a storage unit 20, and a communication IF (interface) unit 29. The control unit 10 is an information processing unit (CPU) that calculates and processes information, and performs overall control of the information processing server 1. The control unit 10 works in cooperation with the above-mentioned hardware to execute various functions by appropriately reading and executing an operating system (OS) and various application programs stored in the storage unit 20.

[0022] Specifically, in this embodiment, an example will be described in which the information processing server 1 is realized by causing a computer to execute a program. The program can be recorded on a computer-readable non-transitory information recording medium such as a compact disc, a flexible disk, a hard disk, a magneto-optical disk, a digital video disk, a magnetic tape, a ROM (Read Only Memory), an EEOPROM (Electrically Erasable Programmable ROM), a flash memory, or a semiconductor memory. This information recording medium can be distributed and sold independently of the computer.

[0023] Generally, a computer reads a program recorded on a non-transitory information recording medium into RAM (Random Access Memory), which is a temporary storage device included in the storage unit 20, and then the CPU serving as the control unit 10 executes the instructions included in the read program. In addition, the program can be distributed and sold to computers, etc. from a program distribution server, etc. (not shown) via a temporary transmission medium such as a communication network N, independent of the computer on which the program is executed.

[0024] The program can also be written in a programming language for describing the behavior of electronic circuits. In this case, various design drawings, such as wiring diagrams and timing charts for the electronic circuit, are generated from the program written in the programming language for describing the behavior of electronic circuits, and the electronic circuit that constitutes the information processing server 1 can be created based on the design drawings. For example, the information processing server 1 can be configured on reprogrammable hardware using FPGA (Field Programmable Gate Array) technology from a program written in a programming language for describing the behavior of electronic circuits. It is also possible to configure an electronic circuit dedicated to a specific application using ASIC (Application Specific Integrated Circuit) technology.

[0025] The information processing server 1 is configured so that the control unit 10 controls the following components to execute the processes described in this embodiment. The control unit 10 includes a page output unit 11 (first screen output means), a dialogue setting processing unit 12 (dialogue setting output means, dialogue setting reception means), a dialogue content reception unit 13 (dialogue content reception means), an organizing information generation unit 14 (information organizing means), an expert knowledge identification unit 15 (expert knowledge identification means), a model processing unit 16 (model processing means), a page update unit 17 (second screen output means), an additional dialogue reception unit 18 (additional dialogue reception means), and a learning unit 19 (learning means).

[0026] The page output unit 11 outputs a web page (interactive screen) (hereinafter simply referred to as a "page") including an input area to the terminal 4. The input area is a field where the user of the terminal 4 enters dialogue content such as a question, and allows the dialogue content to be input as text, for example. The dialogue setting processing unit 12 outputs a setting screen for the dialogue method. The dialogue setting processing unit 12 also accepts the setting of the dialogue method via the setting screen. Dialogue methods include, for example, discussion and debate. In a discussion, multiple experts express their ideas and opinions about the dialogue content, and answers (knowledge) are obtained from a broad perspective on the dialogue content, working toward a common understanding and solution for the dialogue content. On the other hand, in a debate, multiple opposing experts assert their respective positions and clash with each other about the dialogue content, allowing for the identification of loopholes in things and the insight to be dug deeper by observing the process.

[0027] The dialogue content receiving unit 13 receives dialogue content in an input field on the page. When a user inputs dialogue content into the input field on the page from the terminal 4 and performs an operation for transmission, the dialogue content receiving unit 13 receives the dialogue content. The organized information generation unit 14 generates organized information by organizing the dialogue content received by the dialogue content receiving unit 13. The organized information is, for example, information spoken by a person playing the role of a leader. The leader is a virtual person who plays the role of facilitating the dialogue between the user and the expert, and is actually a computer. Furthermore, the organizing information generating unit 14 generates organizing information that further adds information based on the setting of the dialogue method accepted by the dialogue setting processing unit 12. For example, the information based on the setting of the dialogue method is, for example, "Let's discuss about XX" in the case of a discussion setting, or "Let's discuss whether A or B is correct about XX" in the case of a debate setting.

[0028] The specialized knowledge identification unit 15 identifies a plurality of pieces of specialized knowledge based on the content of the dialogue. The specialized knowledge identification unit 15 may identify a plurality of pieces of specialized knowledge based on the organized information generated by the organized information generation unit 14. The specialized knowledge identification unit 15 can, for example, extract words from the content of the dialogue or the text of the organized information, and identify a plurality of pieces of specialized knowledge based on the extracted words. Furthermore, the specialized knowledge identifying unit 15 identifies one or more pieces of specialized knowledge based on the content of the additional dialogue accepted by the additional dialogue accepting unit 18.

[0029] The model processing unit 16 sequentially inputs based on the dialogue content to a plurality of learning models stored in the learning model storage unit 23, and acquires responses from the virtual experts corresponding to each piece of specialized knowledge from each learning model. The model processing unit 16 also inputs responses already output by other learning models to the learning models, and acquires responses from the learning models. The model processing unit 16 may input the organizing information generated by the organizing information generation unit 14 to the learning models stored in the learning model storage unit 23, and acquire responses from the virtual experts corresponding to each piece of specialized knowledge from each learning model. Furthermore, the model processing unit 16 inputs additional dialogue content to the learning model and acquires a response from the virtual expert corresponding to the specialized knowledge from the learning model.

[0030] The page update unit 17 outputs a page including the response acquired by the model processing unit 16 to the terminal 4. The page update unit 17 outputs the response in association with identification information that identifies the expert. The page update unit 17 also outputs a page that further includes the organization information generated by the organization information generation unit 14 to the terminal 4. The identification information may be the name of an actual expert or the name of an item of specialized knowledge. When using the name of an actual expert, it is desirable to state that the response was not actually made by that expert. The additional dialogue receiving unit 18 receives, from the terminal 4, additional dialogue content for an input area on a page. The learning unit 19 learns each learning model stored in the learning model storage unit 23, which has each piece of specialized knowledge identified by the specialized knowledge identification unit 15, based on the organized information.

[0031] The storage unit 20 is a storage area such as a hard disk or semiconductor memory element for storing programs, data, etc. required for the control unit 10 to execute various processes. The storage unit 20 includes a program storage unit 21 and a learning model storage unit 23. The program storage unit 21 is a storage area that stores various programs. The program storage unit 21 stores an information processing program 22, which is a program for executing each function of the control unit 10 described above.

[0032] The learning model storage unit 23 is a storage area that stores a plurality of learning models having specialized knowledge. The learning model storage unit 23 has, for example, learning models each having different specialized knowledge. If the virtual expert corresponding to the specialized knowledge is a real person, the learning model may be one that has learned, for example, publications and books published by that person, data posted on that person's social networking service (SNS) account, etc. Also, if the virtual expert corresponding to the specialized knowledge is a fictitious person, the learning model may be one that has learned, for example, publications, books, web pages, etc. in the specialized field. The learning model is learned by the learning unit 19. The learning model may be learned continuously at appropriate timings such as periodic timings. The communication IF unit 29 is an interface for communicating with the terminal 4.

[0033] [Terminal 4] The terminal 4 is used by a user and is, for example, a PC, but may also be a mobile terminal such as a smartphone or a tablet. Although not shown, the terminal 4 includes a control unit, a storage unit, an input unit, a display unit, a communication IF unit, etc. The terminal 4 may also include a touch panel display that combines the functions of the input unit and the display unit.

[0034] Note that a computer refers to an information processing device equipped with a control unit, a storage device, etc., and the information processing server 1 and the terminal 4 are both information processing devices equipped with a control unit, a storage device, etc., and are included in the concept of a computer.

[0035] [Processing Description] Next, the processing in the information processing server 1 will be described. FIG. 2 is a flowchart showing the dialogue processing of the dialogue system 100 according to this embodiment. 3 and 4 are diagrams showing examples of interactive screens output to the terminal 4 according to this embodiment. FIG. 5 is a diagram showing an example of a setting screen output to the terminal according to this embodiment. FIG. 6 is a flowchart showing the dialogue acquisition process of the information processing server 1 according to this embodiment. FIG. 7 is a flowchart showing the model processing of the information processing server 1 according to this embodiment. FIG. 8 is a diagram showing an example of the dialogue content and reply output to the terminal 4 according to this embodiment.

[0036] Terminal 4, for example, displays the website (home page) of a tax accountant, and when the user selects a link on the website that takes the user to a page where questions can be asked, in step S (hereinafter, "step S" will be simply referred to as "S") 11 of Figure 2, the control unit of terminal 4 sends a page request to information processing server 1. In the information processing server 1, upon receiving the page request, the page output unit 11 outputs the page including the input area to the terminal 4 in S12. In S13, the control unit of the terminal 4 causes the page received from the information processing server 1 to be displayed.

[0037] FIG. 3 is an example of a page 51 displayed on the terminal 4. Page 51 states, "Ask us anything about XX," and includes an input area 51a and a setting area 51b (setting screen). In this example, since the page was transitioned from a tax accountant's website, "XX" is, for example, "tax accountant," "tax," "tax return," etc. The input area 51a is a field for inputting the content of the dialogue as text. The user can input the content of the dialogue as text in the input area 51a. The user can also select and operate the send button 51c. The setting area 51b is an area for setting the dialogue method. For example, as shown on page 51, the information processing server 1 may output to the terminal 4 as a default setting that the discussion mode is always ON.

[0038] It should be noted that the page displayed on the terminal 4 does not have to include a setting area, and for example, the setting area may be set in advance by the operations manager. FIG. 4 is an example of a page 52 displayed on the terminal 4, which is an example of a page that does not include a setting area. The page 52 includes an input area 52a. The user can input the content of the conversation as text in the input area 52a. The user can also select a send button 52c.

[0039] Next, an example of a setting screen that is set in advance by the operations manager will be described. FIG. 5(A) is an example of a setting screen 61 displayed on a terminal (not shown) used by an operations manager. The setting screen 61 includes a first mode area 61a, a details area 61b, and a second mode area 61c. The first mode area 61a is an area related to setting the discussion mode. The details area 61b is an area for specifying details in the discussion mode. When designated summoning is selected here, the setting screen 62 shown in FIG. 5(B) is displayed. The administrator operates the setting area 62a of the setting screen 62 to specify the name of the expert and the learning data tag. The name of the expert may be the name of a real person or the name of a field of expertise. The tag is a word related to the expert's field of expertise. The second mode area 61c is an area related to the setting of the debate mode.

[0040] In this way, the interaction method may be set by the user himself or by the administrator of the page. Also, the screen example is just an example, and other page configurations may be used.

[0041] In S14 of FIG. 2, the control unit of the terminal 4 accepts the setting of the interaction method. In S15, the control unit of the terminal 4 accepts input of the dialogue content. In S16, the control unit of the terminal 4 transmits the dialogue content and the setting content to the information processing server 1. These processes on the terminal 4 can be performed, for example, by the user setting up the setting area 51b on the page 51 shown in Figure 3 above, entering the dialogue content as text in the input area 51a, and then selecting the send button 51c. In the case of page 52 shown in FIG. 4, the process of S15 in FIG. 2 is not performed.

[0042] In S17, the control unit 10 of the information processing server 1 performs a response acquisition process. The reply acquisition process will now be described with reference to FIG. In S31 of FIG. 6, the interaction setting processing unit 12 accepts the setting of the interaction method. In S32, the dialogue content receiving unit 13 receives the dialogue content. In S33, the organization information generating unit 14 generates organization information based on the settings of the dialogue content and dialogue method. In S34, the page update unit 17 updates the page by adding the organizing information to the page, and outputs the updated page to the terminal 4. In S35, the specialized knowledge identification unit 15 identifies a plurality of specialized knowledge based on the generated organized information. In S36, the model processing unit 16 identifies one piece of specialized knowledge from the identified plurality of pieces of specialized knowledge. In S37, the control unit 10 performs model processing.

[0043] Here, the model processing will be explained with reference to FIG. In S41 of FIG. 7, the model processing unit 16 inputs the organizing information and the responses of the other learning models to the learning model corresponding to the specified specialized knowledge. Here, when the process for identifying specialized knowledge is performed for the second time or later, there are responses from the learning model corresponding to other specialized knowledge. Therefore, the model processing unit 16 also inputs responses from the learning model corresponding to other specialized knowledge into the learning model. However, when the process for identifying specialized knowledge is performed for the first time, there are no responses from other learning models, so only the organized information is input.

[0044] In S42, the model processing unit 16 obtains a response from the learning model. In S43, the page update unit 17 updates the page by adding the response, and outputs the updated page to the terminal 4. In S44, the learning unit 19 performs a learning process to learn the learning model used in the process of S41 based on the sorting information. After that, the control unit 10 moves the process to S38 in FIG.

[0045] In S38 of Fig. 6, the control unit 10 determines whether or not all of the plurality of pieces of specialized knowledge identified in the processing of S35 have been processed. If all of the plurality of pieces of specialized knowledge have been processed (S38: YES), the control unit 10 shifts the processing to S18 of Fig. 2. On the other hand, if all of the plurality of pieces of specialized knowledge have not been processed (S38: NO), the control unit 10 shifts the processing to S36, identifies the pieces of specialized knowledge that have not been processed, and then performs model processing.

[0046] 2, the control unit of the terminal 4 displays the page transmitted by the information processing server 1. The control unit of the terminal 4 may perform the process of S18 every time the information processing server 1 updates the page and transmits it to the terminal 4. FIG. 8 shows an example of a page 71 including an input example into the input area 51a and a reply to the input dialogue content. After inputting the dialogue content into the input area 51a, the user can select and operate the send button 51c, thereby erasing the dialogue content that had been input into the input area 51a and allowing the next dialogue content to be input. Furthermore, when the user selects the send button 51c, a reply area 73 is output together with a sorted information area 72, as shown on page 71.

[0047] The summary information area 72 is an area for outputting information that summarizes the dialogue content input in the input area 51a. The summary information is presented as if the leader is speaking. The reply area 73 is an area for outputting a reply to the dialogue content, which is output together with specific information that identifies each expert. In the reply area 73, specific information for each expert is output, and a reply to the dialogue content is also output. The page 71 shown in FIG. 8 may output responses in order while the input area 51a remains in a fixed position.

[0048] In S19 of FIG. 2, the control unit of the terminal 4 determines whether or not an additional dialogue has been accepted. The control unit of the terminal 4 can determine whether or not an additional dialogue has been accepted by checking whether dialogue content has been added to the input area 51a. If an additional dialogue has been accepted (S19: YES), the control unit of the terminal 4 moves the process to S16. On the other hand, if an additional dialogue has not been accepted (S19: NO), the control unit of the terminal 4 ends this process.

[0049] As described above, the information processing server 1 of this embodiment has the following advantages. (1) A page including an input area is output, dialogue content for the input area on the output page is accepted, input based on the accepted dialogue content is sequentially made to a plurality of learning models each having different specialized knowledge, responses from virtual experts corresponding to each specialized knowledge are obtained from each learning model, and the acquired responses are output as updates to the page. In this case, responses already output by other learning models are also input to the learning models, and responses are obtained from the learning models. Therefore, the user can check the replies of the virtual expert corresponding to a plurality of specialized knowledge in response to the dialogue content, and the user can refer to the replies backed by various specialized knowledge.

[0050] (2) Multiple pieces of specialized knowledge are identified based on the received dialogue content, and input based on the dialogue content is sequentially made to learning models having the identified specialized knowledge, and responses from virtual experts corresponding to each piece of specialized knowledge are obtained from each learning model. Therefore, since multiple pieces of specialized knowledge are identified based on the content of the dialogue, it is possible to confirm a response from a learning model having multiple pieces of specialized knowledge that are in line with the content of the dialogue.

[0051] (3) Each learning model with the identified specialized knowledge is trained based on the content of the dialogue, and the trained learning model is used to generate responses to the content of the dialogue. Therefore, the learning model is one that has learned the content of the dialogue, and can always provide specialized responses.

[0052] (4) The system generates organized information by organizing the received dialogue content, identifies multiple pieces of specialized knowledge based on the organized organized information, inputs the organized information into multiple learning models each having the identified specialized knowledge, obtains responses from each learning model with a virtual expert corresponding to each piece of specialized knowledge, and further outputs the organized information to a page. Therefore, by identifying specialized knowledge from organized information that organizes the dialogue content and inputting the organized information into the learning model, it is possible to input the information into the learning model in a format that is easy to respond to.In addition, by outputting the organized information, it is possible to allow the user to confirm the organized information.

[0053] (5) A setting screen for the interaction method is output, the setting of the interaction method on the setting screen is accepted, and organized information is generated by further adding information based on the accepted setting of the interaction method. Therefore, by setting the dialogue method, it is possible to obtain a response to the dialogue content that matches the dialogue method.

[0054] (6) Accept additional dialogue content for the input field on the page on which the response is output, identify specialized knowledge based on the accepted additional dialogue content, provide input based on the accepted additional dialogue content to a learning model having the identified specialized knowledge, and obtain a response from a virtual expert corresponding to the specialized knowledge from the learning model. Therefore, it is possible to continue a dialogue with the user using the learning model, and the content of the user's dialogue can be deepened.

[0055] (7) A response is output in association with specific information that can identify specialized knowledge. Therefore, the user can confirm what kind of specialized knowledge the virtual expert has when he or she replies.

[0056] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments. Furthermore, the effects described in the embodiments are merely a list of the most preferable effects resulting from the present invention, and the effects of the present invention are not limited to those described in the embodiments. Note that the above-described embodiments and the modified embodiments described below can be used in appropriate combinations, but detailed description thereof will be omitted.

[0057] (Variations) (1) In this embodiment, the discussion has been mainly described, but the present invention is not limited to this. In the case of a debate, each expert can respond with their position to the question, "Regarding XX, is it A or B?" When the dialogue method described above is a discussion, responses (knowledge) can be obtained from a broad perspective on the content of the dialogue through responses output by multiple learning models with specialized knowledge.

[0058] On the other hand, when the dialogue method is debate, it has the following two characteristics. - Multiple virtual experts (for example, up to five) with the appropriate expertise for the user's conversation content are identified, and each one responds using their own characteristics and skills. Each virtual expert uses its own characteristics and skills to refute the responses (opinions) of other virtual experts from different angles, resulting in a debate between the learning models.

[0059] When the dialogue method is a debate, multiple virtual experts with specialized knowledge will argue their opinions, allowing you to see if there are any loopholes, and by observing the process, you can delve deeper into your knowledge. Furthermore, by selecting both discussion and debate as dialogue methods for a given dialogue, users can prevent hallucination, in which the learning model gives incorrect responses.

[0060] (2) In this embodiment, the example of a tax accountant has been described, but this is not limiting. It is also possible to call in various virtual experts to respond. If the learning model storage unit 23 does not have a learning model for the virtual expert, for example, a generation AI that has learned all the data may be used, and a response may be obtained from the generation AI by inputting keywords including specialized terms (e.g., patent attorney) in addition to the organized information, which may be regarded as a response from one virtual expert.

[0061] (3) In the present embodiment, multiple learning models each having multiple pieces of specialized knowledge are used as an example, but this is not limiting. A single learning model having multiple pieces of specialized knowledge may also be used.

[0062] (4) In the present embodiment, the response is output as text, but this is not limiting. The system may have a function for outputting the response as voice, and may output the response as text and as voice at the same time. In this case, a voice quality pattern may be stored in a storage unit, and the voice quality of the response output voice may be different for each virtual expert.

[0063] (5) In this embodiment, the number of virtual experts corresponding to the specialized knowledge is not specifically mentioned, but the number may be an item that can be set on a setting screen. In this case, the number may be selectable, for example, between two and five.

[0064] (6) In each embodiment, a web page is used as an example, but the present invention is not limited to this. An app may also be used. [Explanation of symbols]

[0065] 1. Information processing server 4. Terminal 10 Control Unit 11 Page Output Section 12 Dialogue setting processing section 13. Dialogue Content Reception Department 14 Organizing information generation section 15 Specialized Knowledge Identification Department 16 Model processing section 17 Page Update Section 18 Additional Dialogue Reception Department 19 Learning Department 20 Memory section 22 Information Processing Program 23 Learning model memory unit Pages 51, 52, and 71 51a, 52a Input area 61, 62 Setting screen 100 Dialogue Systems

Claims

1. a first screen output means for outputting an interactive screen including an input area; a dialogue content receiving means for receiving dialogue content to be entered into the input area of ​​the dialogue screen output by the first screen output means; a model processing means for sequentially inputting, to a plurality of learning models each having different specialized knowledge, an input based on the dialogue content received by the dialogue content receiving means, and acquiring, from each learning model, a response from a virtual expert corresponding to each specialized knowledge; a second screen output means for outputting the response acquired by the model processing means to the interactive screen; Equipped with The model processing means inputs the responses already output by other learning models into the learning model, and acquires the responses from the learning model.

2. 2. The information processing device according to claim 1, a specialized knowledge identification unit that identifies a plurality of pieces of specialized knowledge based on the dialogue content received by the dialogue content reception unit, The model processing means sequentially inputs based on the dialogue content to a plurality of learning models each having the specialized knowledge identified by the specialized knowledge identification means, and obtains the response of a virtual expert corresponding to each specialized knowledge from each learning model.

3. 3. The information processing device according to claim 2, a learning means for learning each learning model having each piece of specialized knowledge identified by the specialized knowledge identification means based on the content of the dialogue; The model processing means uses the learning model learned by the learning means.

4. 3. The information processing device according to claim 2, an information organizing means for generating organized information that organizes the dialogue content received by the dialogue content receiving means, the specialized knowledge identification means identifies a plurality of pieces of specialized knowledge based on the organized information generated by the information organization means; The model processing means inputs the organized information generated by the information organizing means into a plurality of learning models each having the specialized knowledge identified by the specialized knowledge identifying means, and acquires the responses of the virtual experts corresponding to each specialized knowledge from each learning model; The second screen output means further outputs the organized information generated by the information organizing means to the interactive screen.

5. 5. The information processing device according to claim 4, an interaction setting output means for outputting a setting screen relating to an interaction method; an interaction setting receiving means for receiving a setting of the interaction method on the setting screen; Equipped with The information organizing means generates the organized information by further adding information based on the setting of the interaction method accepted by the interaction setting accepting means.

6. 3. The information processing device according to claim 2, an additional dialogue accepting means for accepting additional dialogue content to be added to the input area on the interactive screen; the specialized knowledge identifying means identifies one or more pieces of specialized knowledge based on the additional dialogue content accepted by the additional dialogue accepting means; The model processing means inputs based on the additional dialogue content accepted by the additional dialogue accepting means to the learning model having the specialized knowledge identified by the specialized knowledge identifying means, and obtains the response of a virtual expert corresponding to the specialized knowledge from the learning model.

7. 2. The information processing device according to claim 1, The second screen output means outputs the response in association with specific information that can identify the specialized knowledge.

8. a first screen output means for outputting an interactive screen including an input area to a display device; a dialogue content receiving means for receiving, from an input device, dialogue content to be entered into the input area of ​​the dialogue screen output by the first screen output means; a model processing means for sequentially inputting, to a plurality of learning models each having different specialized knowledge, an input based on the dialogue content received by the dialogue content receiving means, and acquiring, from each learning model, a response from a virtual expert corresponding to each specialized knowledge; a second screen output means for outputting the response acquired by the model processing means onto the interactive screen of the display device; Equipped with An information processing system in which the model processing means inputs the responses, including those already output by other learning models, into the learning model and obtains the responses from the learning model.

9. A program for causing a computer to function as the information processing device according to any one of claims 1 to 7.

Images