Information processing device, information processing system, information processing method, and program
The chat system addresses the issue of inappropriate responses in conventional chatbots by using a large-scale language model to generate follow-up questions, ensuring sufficient user input is gathered, thus enhancing response accuracy.
Patent Information
- Application Number
- JP2024156535
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2024-03-21
- Filing Date
- 2024-09-10
- Publication Date
- 2026-01-27
- Estimated Expiration
- 2044-09-10
AI Technical Summary
Conventional chatbots struggle to provide appropriate responses when the amount of user input is small, leading to multiple possible outputs and difficulty in delivering the desired response.
A chat system that includes an input receiving unit, an output unit, and a request unit to utilize a large-scale language model to generate follow-up questions, searching user-specific information to obtain additional input data when necessary, ensuring sufficient information is gathered before generating a response.
The system effectively prompts users for additional information, reducing the likelihood of inappropriate responses by ensuring sufficient input is provided, thereby improving the accuracy of the chatbot's output.
Smart Images

Figure 0007806856000001 
Figure 0007806856000002 
Figure 0007806856000003
Abstract
Description
[Technical Field]
[0001] The present invention relates to an information processing device, an information processing system, an information processing method, and a program. [Background technology]
[0002] Chatbots that output responses to questions entered by a user have been known for some time. One example of such chatbots is a technique that obtains responses based on user input information entered by a user from a large-scale language model. Summary of the Invention [Problem to be solved by the invention]
[0003] In the conventional chatbots described above, if the amount of information input by the user is small, a large number of possible responses to the input information will be output, making it difficult to output the appropriate response that the user desires.
[0004] The disclosed technology has been made in consideration of the above circumstances, and aims to provide a chat system that outputs a follow-up question to output an appropriate response that the user desires. [Means for solving the problem]
[0005] The disclosed technology includes an input receiving unit that receives input question data; searching a first storage device, in which question and answer information in which registered question data and registered answer data are associated with each other, for the input question data; an output unit that outputs to a terminal device registered answer data associated with the registered question data identified based on the question indicated by the input question data; and To, Request a large-scale language model to generate a question to obtain additional input data If the registered question data specified based on the question does not exist, a second storage device in which user-specific information is stored for each chunk is searched for the input question data, and if multiple chunks are obtained as a result of the search,an output unit that outputs the question sentence for obtaining the additional input question data, which is output from the large-scale language model, to a terminal device; and a request unit that requests the large-scale language model to generate a question sentence for obtaining additional input question data. [Effects of the Invention]
[0006] A chat system can be provided that outputs follow-up questions to prompt the user to output the appropriate response desired. [Brief explanation of the drawings]
[0007] [Figure 1] FIG. 1 illustrates an example of a system configuration of a chat system. [Figure 2] FIG. 2 illustrates an example of a hardware configuration of an information processing device. [Figure 3] FIG. 2 illustrates an example of a hardware configuration of a terminal device. [Figure 4] FIG. 2 illustrates an example of a question and answer storage unit. [Figure 5] FIG. 2 is a diagram illustrating the functions of each device in the chat system of the first embodiment. [Figure 6] FIG. 2 is a first sequence diagram illustrating the operation of the chat system according to the first embodiment. [Figure 7] FIG. 10 illustrates an example of a prompt according to the first embodiment. [Figure 8A] FIG. 10 is a first diagram illustrating a display example of the terminal device according to the first embodiment. [Figure 8B] FIG. 10 is a second diagram illustrating a display example of the terminal device according to the first embodiment. [Figure 8C] FIG. 10 is a third diagram illustrating a display example of the terminal device according to the first embodiment. [Figure 8D] FIG. 10 is a fourth diagram illustrating a display example of the terminal device according to the first embodiment. [Figure 9] FIG. 10 is a fifth diagram illustrating a display example of the terminal device according to the first embodiment. [Figure 10A]FIG. 6 is a sixth diagram illustrating a display example of the terminal device according to the first embodiment. [Figure 10B] FIG. 7 is a seventh diagram illustrating a display example of the terminal device according to the first embodiment. [Figure 10C] FIG. 11 is an eighth diagram illustrating a display example of the terminal device according to the first embodiment. [Figure 11] FIG. 10 illustrates an example of a system configuration of a chat system according to a second embodiment. [Figure 12] FIG. 2 illustrates an example of a document storage unit. [Figure 13] FIG. 10 is a diagram illustrating the functions of each device in the chat system of the second embodiment. [Figure 14] FIG. 10 is a sequence diagram illustrating the operation of a chat system according to a second embodiment. [Figure 15] FIG. 10 is a diagram illustrating a display example of a terminal device according to the second embodiment. [Figure 16] FIG. 10 is a first diagram illustrating an example of a prompt according to the second embodiment. [Figure 17] FIG. 10 is a second diagram illustrating an example of a prompt according to the second embodiment. [Figure 18] FIG. 10 illustrates an example of a system configuration of a chat system according to a third embodiment. [Figure 19] FIG. 2 is a diagram illustrating an example of a conversation history storage unit. [Figure 20] FIG. 2 illustrates an example of a user information storage unit. [Figure 21] FIG. 10 is a diagram illustrating the functions of each device in the chat system of the third embodiment. [Figure 22] FIG. 11 is a sequence diagram illustrating the operation of a chat system according to a third embodiment. [Figure 23] FIG. 11 is a first diagram illustrating a display example of output response data on a terminal device according to the third embodiment. [Figure 24] FIG. 13 is a second diagram illustrating an example of how output response data is displayed on the terminal device according to the third embodiment. [Figure 25] FIG. 13 is a third diagram illustrating an example of display of output response data on the terminal device according to the third embodiment. [Figure 26] FIG. 11 is a fourth diagram illustrating an example of display of output response data on the terminal device according to the third embodiment. DETAILED DESCRIPTION OF THE INVENTION
[0008] (First embodiment) The present embodiment will be described below with reference to the drawings: Fig. 1 is a diagram showing an example of the system configuration of a chat system according to the first embodiment.
[0009] The chat system 100 of this embodiment includes an information processing device 200, a first storage device 300, a server device 500, and a terminal device 600, which are connected to each other via a communication network 1. The chat system 100 of this embodiment is an example of an information processing system.
[0010] In the chat system 100, the information processing device 200, the first storage device 300, and the server device 500 provide a chatbot service to the user of the terminal device 600.
[0011] When a question is input from the terminal device 600, the information processing device 200 references the question and answer storage unit 310 stored in the first storage device 300 and outputs the answer to the question to the terminal device 600.
[0012] The first storage device 300 has a question and answer storage unit 310. The question and answer storage unit 310 stores question and answer information in which questions and answers are associated with each other. In other words, the question and answer storage unit 310 is an example of a question and answer information group in which question data indicating a question is associated with answer data indicating an answer. The question and answer information is information in which question data indicating a question is associated with answer data indicating an answer in a pair. The question and answer information is information that is stored in advance in the question and answer storage unit 310 and may be stored for each tenant that introduces the chat system 100. In the following description, the question data included in the question and answer information stored in the question and answer storage unit 310 will be referred to as registered question data, and the answer data included in the question and answer information will be referred to as registered answer data.
[0013] The server device 500 has a large-scale language model storage unit 510 and realizes a generative AI (Artificial Intelligence). The generative AI generates text data according to an input prompt, such as ChatGPT, and outputs the text data to the information processing device 200. Note that the prompt in this embodiment is text data for expressing instructions or questions to the generative AI realized by the server device 500. Details of the large-scale language model will be described later.
[0014] The terminal device 600 is mainly used by a user who uses the chat system 100. More specifically, the terminal device 600 transmits question data indicating a question to the chat system 100 to the information processing device 200. Furthermore, the terminal device 600 displays answer data output from the information processing device 200, indicating an answer to the question indicated by the input question data, in association with the input question data.
[0015] In the following description, question data transmitted from the terminal device 600 to the information processing device 200 and input to the information processing device 200 will be referred to as input question data. Also, in the following description, answer data output from the information processing device 200 to the terminal device 600 will be referred to as output answer data.
[0016] In this embodiment, the terminal device 600 may be, for example, a smartphone or a tablet terminal.
[0017] In the chat system 100 of this embodiment, when input question data is input, the information processing device 200 determines whether the amount of information obtained from the input question data is sufficient for generating output answer data.
[0018] If the amount of information obtained from the input question data is insufficient, the information processing device 200 outputs a prompt to the server device 500 requesting the generation of an additional question sentence to prompt the user to input further information.
[0019] Furthermore, the information processing device 200 acquires, from the server device 500, output data output from the large-scale language model included in the server device 500. Here, the output data output from the large-scale language model is text data indicating an additional question sentence for prompting the user to input further information.
[0020] In other words, the output data output from the large-scale language model is text data that prompts the user to enter additional information related to the content of the question indicated by the input question data.
[0021] In the following description, the output data output from the large-scale language model may be referred to as inverse query data.
[0022] When the information processing device 200 acquires the output data from the server device 500, it displays the acquired output data on the terminal device 600 and prompts the user of the terminal device 600 to input further information.
[0023] In the chat system 100 of this embodiment, this operation is repeated until a sufficient amount of information is obtained in generating output response data.
[0024] Therefore, according to this embodiment, it is possible to prompt the user to input more information in addition to the initially input question data, and to prevent answers that the user does not intend from being presented.
[0025] 1, the information processing device 200 and the first storage device 300 are separate devices, but the present invention is not limited to this. The first storage device 300 may be included in the information processing device 200.
[0026] In addition, in FIG. 1, the information processing device 200 may be realized by one or more information processing devices, and may be realized by a plurality of information processing devices.
[0027] Next, an example of the hardware configuration of each device included in the chat system 100 of this embodiment will be described.
[0028] In this embodiment, the information processing device 200, the first storage device 300, and the server device 500 are general-purpose computers, and therefore, in FIG. 2, the hardware configuration of the information processing device 200 will be described as an example.
[0029] In this embodiment, the terminal device 600 is a smartphone, and the hardware configuration of the terminal device 600 will be described in FIG. 3 as an example.
[0030] Fig. 2 is a diagram showing an example of the hardware configuration of an information processing device. The information processing device 200 shown in Fig. 2 is constructed by a computer (information processing device), and includes a CPU 201, a ROM 202, a RAM 203, an HD 204, an HDD (Hard Disk Drive) controller 205, a display 206, an external device connection I / F (Interface) 208, a network I / F 209, a bus line 210, a keyboard 211, a pointing device 212, a DVD-RW (Digital Versatile Disk Rewritable) drive 214, and a media I / F 216, as shown in Fig. 2.
[0031] Of these, the CPU 201 controls the overall operation of the information processing device 200. The ROM 202 stores programs used to drive the CPU 201, such as an IPL. The RAM 203 is used as a work area for the CPU 501. The HD 204 stores various data, such as programs. The HDD controller 205 controls the reading and writing of various data from and to the HD 204 under the control of the CPU 501. The display 206 displays various information, such as a cursor, menus, windows, characters, or images. The external device connection I / F 208 is an interface for connecting various external devices. In this case, the external devices are, for example, USB (Universal Serial Bus) memories, printers, etc. The network I / F 209 is an interface for data communication using a communication network. The bus line 210 is an address bus, a data bus, etc. for electrically connecting the components, such as the CPU 201, shown in FIG. 2.
[0032] The keyboard 211 is a type of input means having multiple keys for inputting characters, numbers, various instructions, etc. The pointing device 212 is a type of input means for selecting and executing various instructions, selecting a processing target, moving a cursor, etc. The DVD-RW drive 214 controls reading and writing of various data from and to a DVD-RW 213, which is an example of a removable recording medium. Note that this is not limited to a DVD-RW, and may be a DVD-R, etc. The media I / F 216 controls reading and writing (storing) of data from and to a recording medium 215, such as a flash memory.
[0033] 3 is a diagram showing an example of the hardware configuration of a terminal device 600 according to this embodiment. The terminal device 600 includes a CPU 601, a ROM 602, a RAM 603, an EEPROM 604, a CMOS sensor 605, an image sensor I / F 606, an acceleration / direction sensor 607, a media I / F 609, and a GPS receiver 611.
[0034] Of these, the CPU 601 is an arithmetic processing unit that controls the overall operation of the terminal device 600. The ROM 602 stores the CPU 601 and programs used to drive the CPU 601, such as the IPL. The RAM 603 is used as a work area for the CPU 601. The EEPROM 604 reads or writes various data, such as smartphone programs, under the control of the CPU 601. The ROM 602, RAM 603, and EEPROM 604 are examples of storage devices of the terminal device 600.
[0035] The CMOS (Complementary Metal Oxide Semiconductor) sensor 605 is a type of built-in imaging means that captures an image of a subject (mainly a self-portrait) and obtains image data under the control of the CPU 601. Note that instead of a CMOS sensor, an imaging means such as a CCD (Charge Coupled Device) sensor may also be used.
[0036] The imaging element I / F 606 is a circuit that controls the driving of the CMOS sensor 605. The acceleration / azimuth sensor 607 is a variety of sensors, such as an electronic magnetic compass that detects geomagnetism, a gyrocompass, and an acceleration sensor. The media I / F 609 controls the reading and writing (storage) of data from and to a recording medium 608, such as a flash memory. The GPS receiver 611 receives GPS signals from GPS satellites.
[0037] The terminal device 600 also includes a long-distance communication circuit 612, an antenna 612a of the long-distance communication circuit 612, a CMOS sensor 613, an image sensor I / F 614, a microphone 615, a speaker 616, an audio input / output I / F 617, a display 618, an external device connection I / F (Interface) 619, a short-distance communication circuit 620, an antenna 620a of the short-distance communication circuit 620, and a touch panel 621.
[0038] Of these, the long-distance communication circuit 612 is a circuit that communicates with other devices via a communication network. The CMOS sensor 613 is a type of built-in imaging means that captures an image of a subject and obtains image data under the control of the CPU 601. The image sensor I / F 614 is a circuit that controls the driving of the CMOS sensor 613. The microphone 615 is a built-in circuit that converts sound into an electrical signal. The speaker 616 is a built-in circuit that converts an electrical signal into physical vibrations to produce sound such as music or voice. The audio input / output I / F 617 is a circuit that processes the input and output of audio signals between the microphone 615 and the speaker 616 under the control of the CPU 601.
[0039] The display 618 is a type of display means such as a liquid crystal display or organic electroluminescence (EL) display that displays an image of a subject, various icons, etc. The external device connection I / F 619 is an interface for connecting various external devices. The short-range communication circuit 620 is a communication circuit such as NFC (Near Field Communication) or Bluetooth (registered trademark). The touch panel 621 is a type of input means that allows a user to operate the terminal device 600 by pressing the display 618. The display 618 is an example of a display unit that the terminal device 600 has.
[0040] 3 is an example of the terminal device 600, and the hardware configuration of the terminal device 600 is not limited to this. The hardware configuration of the terminal device 600 may be the configuration shown in FIG.
[0041] Next, the question and answer storage unit 310 will be described with reference to Fig. 4. Fig. 4 is a diagram showing an example of the question and answer storage unit.
[0042] The question and answer information stored in the question and answer storage unit 310 of this embodiment may be provided for each tenant and managed by each tenant. The example of Fig. 4 shows question and answer information managed by a tenant identified by tenant ID "11".
[0043] The question and answer information includes information items such as a tenant ID, a question and answer ID, an update date and time, a question, and an answer, and the items "tenant ID" and "question and answer ID" are associated with other items. The question and answer information of this embodiment is information in which the value of the item "question and answer ID" is associated with the values of the other items.
[0044] The value of the item "Question and Answer ID" indicates identification information that identifies the question and answer information. The value of the item "Update Date and Time" indicates the date and time when the question and answer information was updated. The value of the item "Question" is the registered question data (text data), and the value of the item "Answer" is the registered answer data (text data). The values of the items "Question" and "Answer" are stored in advance by the administrator of each tenant, etc.
[0045] In addition to the items shown in FIG. 4, the question and answer information may include, for example, items such as "Contact Information" and "Link." The value of the item "Contact Information" indicates the contact information for the user of the terminal device 600 when making an inquiry related to the registered question data. The contact information may be, for example, an email address or a telephone number. The value of the item "Link" indicates a link that shows information related to the registered question data or the registered answer data.
[0046] The items "contact information" and "links" do not have to be stored in the question and answer storage unit 310, as long as the question and answer storage unit 310 stores registered question data and registered answer data.
[0047] Furthermore, in the case where the first storage device 300 is provided in the information processing device 200, the question and answer information of this embodiment does not need to be assigned a tenant ID, and may be referenced only by users who can access the information processing device 200.
[0048] Next, the functional configuration of each device included in the chat system 100 of this embodiment will be described with reference to Fig. 5. Fig. 5 is a diagram illustrating the functions of each device included in the chat system of the first embodiment.
[0049] First, a description will be given of the functional configuration of the information processing device 200. The information processing device 200 of this embodiment includes an input receiving unit 251, a search request unit 252, a determination unit 253, an information acquisition unit 254, a prompt generation unit 255, a request unit 256, and an output unit 257.
[0050] The input receiving unit 251 receives various inputs to the information processing device 200. Specifically, the input receiving unit 251 receives inputs such as input question data from the terminal device 600.
[0051] The search request unit 252 makes a search request to the first storage device 300. Specifically, the search request unit 252 sends to the first storage device 300 a search request for the question and answer storage unit 310 using the input question data as a search key.
[0052] The determination unit 253 determines, from the search results in response to a search request made by the search request unit 252 to the first storage device 300, whether or not the amount of information obtained from the input question data is sufficient for generating output response data.
[0053] In other words, the determining unit 253 determines whether the amount of information acquired from the question and answer storage unit 310 of the first storage device 300 based on the input question data is equal to or greater than a predetermined amount of information.
[0054] Specifically, when the number of registered answer data obtained as a result of searching the question and answer memory unit 310 of the first memory device 300 is multiple, the judgment unit 253 may determine that the amount of information obtained as the search result is equal to or greater than a predetermined amount of information.
[0055] The case where the amount of information obtained as a search result is equal to or greater than a predetermined amount means that the information obtained from the input question data is insufficient and the registered answer data cannot be narrowed down sufficiently.
[0056] Therefore, when a plurality of registered answer data are acquired as a result of searching the question and answer storage unit 310, the determining unit 253 may determine that the information obtained from the input question data is insufficient.
[0057] Furthermore, when one piece of registered answer data is acquired as a result of searching the question and answer storage unit 310, the determining unit 253 may determine that the information obtained from the input question data is sufficient.
[0058] The information acquisition unit 254 acquires the results of a search in response to a search request from the search request unit 252 from the first storage device 300. In other words, the information acquisition unit 254 acquires question and answer information, which is the search result, from the first storage device 300.
[0059] The prompt generation unit 255 generates a prompt to be output to the server device 500. For example, when the determination unit 253 determines that information obtained from the input question data is insufficient, the prompt generation unit 255 generates a prompt to request the server device 500 to generate a question sentence to prompt the user to input further information.
[0060] The request unit 256 outputs the prompt generated by the prompt generation unit 255 to the server device 500. In other words, the request unit 256 makes a request based on the prompt to the large-scale language model 511. In this embodiment, making a request based on the prompt to the large-scale language model 511 may be expressed as sending the prompt generated by the prompt generation unit 255 to the server device 500. The output unit 257 outputs various types of information from the information processing device 200.
[0061] Next, we will explain the functions of the first storage device 300. The first storage device 300 includes a question and answer storage unit 310, a search unit 330, and a storage unit 340. The search unit 330 searches the question and answer storage unit 310 in response to a search request from the information processing device 200, and transmits the search results to the information processing device 200.
[0062] Next, we will explain the functions of the server device 500. The server device 500 has a large-scale language model storage unit 510, an input unit 520, and an output unit 530. The large-scale language model storage unit 510 stores a large-scale language model 511 that realizes a generative AI.
[0063] The large-scale language model 511 of this embodiment is a computer language model that is generated by executing a learning process using a huge amount of unlabeled text as learning data and is composed of an artificial neural network with a large number of parameters. The large-scale language model 511 can capture much of the syntax and meaning of human language by being sufficiently trained using a context learning technique such as next sentence prediction, which understands the context by determining whether sentence 1 and sentence 2 are consecutive, or a masked language model, which understands the context by masking words in a sentence and predicting the masked words from the words before and after them.
[0064] The input unit 520 inputs a prompt transmitted from the information processing device 200 to the large-scale language model 511. The output unit 530 outputs (transmits) output data output from the large-scale language model 511 to the information processing device 200. In other words, the output unit 530 transmits output response data output from the large-scale language model 511 to the information processing device 200.
[0065] Next, a description will be given of the functions of the terminal device 600. The terminal device 600 includes a communication control unit 630, an input receiving unit 640, and a display control unit 650.
[0066] The communication control unit 630 controls communication between the terminal device 600 and the information processing device 200. The input receiving unit 640 receives input to the terminal device 600. The display control unit 650 controls display on the terminal device 600.
[0067] Next, the operation of the chat system 100 of this embodiment will be described with reference to Fig. 6. Fig. 6 is a first sequence diagram illustrating the operation of the chat system of the first embodiment.
[0068] The operation shown in FIG. 6 is performed when the user of the terminal device 600 logs in to the chat system 100 using his / her own user ID and the tenant ID to which he / she belongs.
[0069] In the chat system 100 of this embodiment, when the terminal device 600 receives a question input via the input receiving unit 640 (step S601), the terminal device 600 transmits the input question data to the information processing device 200 (step S602). Note that the transmitted input question data may be assigned the tenant ID of the tenant to which the user of the terminal device 600 belongs.
[0070] When the information processing device 200 receives the input of the input question data via the input receiving unit 251, the search request unit 252 sends a search request including the tenant ID to the first storage device 300 (step S603).
[0071] When the first storage device 300 receives this search request, the search unit 330 searches the question and answer storage unit 310 (step S604) and transmits the search results to the information processing device 200 (step S605). Specifically, the first storage device 300 may search the question and answer storage unit 310 using the input question data as a search key, and transmit to the information processing device 200 a set of registered question data similar to the input question data and registered answer data associated with the registered question data as the search results.
[0072] When the information processing device 200 acquires the search results from the first storage device 300 via the information acquisition unit 254, the determination unit 253 determines whether or not a plurality of registered answer data items have been acquired as the search results (step S606). That is, the information processing device 200 determines via the determination unit 253 whether or not the amount of information obtained from the input question data input to the information processing device 200 in step S602 is sufficient for generating output answer data.
[0073] When one registered answer data is obtained as a search result, this means that one-to-one output answer data is obtained corresponding to the input question data, and this indicates that the amount of information obtained from the input question data is sufficient for generating the output answer data.
[0074] Furthermore, when multiple registered answer data are obtained as a search result, it means that it is not possible to obtain one-to-one output answer data corresponding to the input question data, and the amount of information obtained from the input question data is insufficient for generating the output answer data.
[0075] If one piece of registered response data is acquired as the search result in step S606, information processing device 200 proceeds to step S618, which will be described later.
[0076] If a plurality of registered answer data are acquired as the search result in step S606, the information processing device 200 causes the prompt generation unit 255 to generate a prompt requesting the generation of a question sentence for prompting the user of the terminal device 600 to input further information (step S607). In other words, the information processing device 200 causes the prompt generation unit 255 to generate a prompt requesting the generation of inverse question data indicating an inverse question.
[0077] The generated prompt includes a plurality of registered answer data items acquired by information processing device 200 as search results in step S605.
[0078] Subsequently, information processing device 200 transmits the prompt generated by prompt generating unit 255 to server device 500 (step S608). Details of the prompt will be described later.
[0079] The server device 500 inputs a prompt to the large-scale language model 511 via the input unit 520, and transmits output data output from the large-scale language model 511 to the information processing device 200 via the output unit 530 (step S609). Here, the output data transmitted to the information processing device 200 is a question sentence for prompting the user of the terminal device 600 to input further information.
[0080] When the information processing device 200 receives input of output data from the server device 500, the output unit 257 outputs the output data to the terminal device 600 (step S610). In other words, the information processing device 200 transmits a question to the terminal device 600 to prompt the user of the terminal device 600 to input further information.
[0081] The terminal device 600 causes the display control unit 650 to display the output data acquired from the information processing device 200 on the display 618 (step S611). That is, the terminal device 600 displays a question (reverse question) for prompting the user to input further information.
[0082] When the terminal device 600 receives additional input corresponding to the question sentence via the input receiving unit 640 (step S612), it transmits the additionally input data to the information processing device 200 (step S613). In the following description, the additionally input data by the user is referred to as additional input data. The additional input data is an example of additional input question data.
[0083] When the information processing device 200 receives the additional input data through the input receiving unit 251, the search request unit 252 outputs a search request based on the input question data and the additional input data to the first storage device 300 (step S614).
[0084] The operations from step S615 to step S617 in FIG. 6 are the same as the processes from step S604 to step S606 in FIG. 6, and therefore a description thereof will be omitted.
[0085] In this embodiment, the operations from step S607 to step S617 are repeated until only one registered answer data is acquired as a search result from the first storage device 300.
[0086] When one piece of registered response data is acquired as the search result, the information processing device 200 causes the output unit 257 to output the acquired registered response data to the terminal device 600 as output response data (step S618).
[0087] The terminal device 600 causes the display control unit 650 to display the output response data received from the information processing device 200 on the display 618 (step S619).
[0088] In this embodiment, a question sentence is generated that prompts the user to enter additional information until registered answer data that matches the input question data one-to-one is obtained. Therefore, this embodiment allows the user to enter more information, thereby improving the accuracy of the answer to the input question data.
[0089] Next, a prompt generated by the information processing device 200 will be described with reference to Fig. 7. Fig. 7 is a diagram showing an example of a prompt according to the first embodiment.
[0090] The prompt 10 shown in Fig. 7 is an example of the prompt generated in step S609 in Fig. 6. The prompt 10 includes text data 10a, 10b, and 10c.
[0091] The text data 10a includes text data indicating that there are multiple registered answer data for the input question data, and text data requesting the generation of a question sentence to prompt the user to enter further information to narrow down the registered answer data.
[0092] The text data 10b is text data including the input question data. The text data 10c may include registered question data. The registered question data included in the text data 10c may be registered question data similar to the input question data that has been extracted as a search result of the question and answer storage unit 310 based on the input question data.
[0093] In this embodiment, by including the input question data and registered question data similar to the input question data in the prompt, the large-scale language model 511 generates a question sentence that prompts the user to enter further information in order to reduce the number of candidates for output answer data.
[0094] Next, a description will be given of a display example of the terminal device 600 of this embodiment. Fig. 8A is a first diagram showing a display example of the terminal device of the first embodiment.
[0095] The screen 110 shown in FIG. 8A is an example of a screen displayed on the terminal device 600 in step S611 of FIG.
[0096] 8A includes display areas 111 to 116. Display area 111 displays input question data input by the user of terminal device 600 in step S601. Display area 112 displays text data indicating the question sentence generated by large-scale language model 511.
[0097] Specifically, the text data displayed in the display area 112 is output data that the large-scale language model 511 outputs in response to the input of the prompt 10 shown in Fig. 7. In other words, the text data displayed in the display area 112 indicates a question prompting the user for additional input.
[0098] The display area 113 displays text data further input by the user of the terminal device 600 as an answer to the question generated by the large-scale language model 511. In other words, the display area 113 displays additional input data additionally input by the user.
[0099] Display area 114 displays text data indicating a question sentence generated by large-scale language model 511 based on the input question data displayed in display area 111 and the additional input data (additional input question data) displayed in display area 113.
[0100] Display area 115 displays additional input data that the user has further input in response to the question displayed in display area 114.
[0101] Display area 116 displays output answer data including registered answer data obtained by searching question and answer storage unit 310 based on the input question data displayed in display area 111, the additional input data displayed in display area 113, and the additional input data displayed in display area 115.
[0102] In other words, in the example of Figure 8A, once the additional input data has been entered twice, the candidates for registered answer data obtained when the input question data was entered have been narrowed down to one registered answer data.
[0103] In the example of Figure 8A, when multiple registered answer data are obtained in a search of the question and answer storage unit 310 using input question data, the multiple registered answer data are not displayed on the terminal device 600, but this is not limited to this.
[0104] When a plurality of registered answer data are acquired as a search result of the question and answer storage unit 310, the information processing device 200 may cause the terminal device 600 to display a plurality of registered question data corresponding to the plurality of registered answer data.
[0105] In this embodiment, the user may be notified of the reason why text data indicating a question prompting additional input is displayed. In other words, the information processing device 200 outputs, via the output unit 257, information indicating the reason why a question prompting additional input is displayed to the terminal device 600. Fig. 8B is a second diagram showing a display example of the terminal device of the first embodiment.
[0106] 8B, display area 112-1 displays notification text data T1 saying, "Let me ask you a question so that we can provide you with a more appropriate answer," as the reason for displaying text data indicating a question prompting additional input. In the following description, text data indicating the reason for displaying a question prompting additional input will be referred to as notification text data. Notification text data T1 shown in FIG. 8B is displayed before the text data indicating a question prompting additional input that is displayed in display area 112-1.
[0107] In this embodiment, by displaying the notification text data on the screen 110-1 of the terminal device 600 in this manner, it is possible to reduce the user's confusion when asked a question prompting additional input.
[0108] The notification text data may be any content that indicates the reason for displaying a question prompting additional input, and the content is not limited to the example shown in Fig. 8B. In the example of Fig. 8B, the notification text data is displayed in display area 112-1 together with text data indicating a question prompting additional input, but this is not limiting. The notification text data may be displayed in a display area other than display area 112-1.
[0109] 8C is a third diagram showing a display example of the terminal device of the first embodiment. In screen 110-2 shown in FIG. 8C, text data "We may ask you a question to provide a more appropriate answer" is displayed outside display area 112 as notification text data T2. In this embodiment, notification text data may be displayed outside the display area shown as a speech bubble. The notification text data may be displayed anywhere on screen 110A.
[0110] In this embodiment, the user of the terminal device 600 may be notified that the chatbot service provided by the chat system 100 will answer the user's question. Fig. 8D is a fourth diagram showing a display example of the terminal device of the first embodiment.
[0111] In the screen 110-3 shown in Fig. 8D, in addition to the display areas shown on the screen 110-1 of Fig. 8B, a display area 117 is displayed. In the display area 117, text data T3 saying "The chatbot will answer your question" is displayed. Note that the text data T3 shown in Fig. 8D is an example. The text data T3 displayed in the display area 117 may be any content that notifies the user that the chatbot service will answer the question entered by the user.
[0112] In this embodiment, by notifying the user in this manner, the user can be made aware that the person with whom he or she is interacting is a chatbot service, which may trigger the user to change his or her mind if the response is not what the user intended.
[0113] Next, a display example in which a plurality of registered question data items are displayed will be described.
[0114] 9 is a fifth diagram showing a display example of the terminal device of the first embodiment. A screen 110A shown in FIG. 9 is an example of a screen displayed on the terminal device 600 in step S611 of FIG.
[0115] A screen 110A shown in FIG. 9 includes display areas 111, 112A, 113, 114A, 115, and 116.
[0116] Display area 112A displays text data 112a that prompts the user to enter additional input data, and a list 112b of registered question data that is similar to the input question data, which is obtained as a result of searching question and answer storage unit 310 using the input question data displayed in display area 111.
[0117] In addition, the list 112b may display registered question data in descending order of similarity to the input question data. Furthermore, if the number of registered question data obtained as the search result is greater than a predetermined number, the list 112b may display only the predetermined number of registered question data, and the remaining registered question data may be grouped under "other."
[0118] Also, in the example of Figure 9, if the registered question data is not selected in the list 112b displayed in the display area 112A and the additional input data shown in the display area 113 is entered, the additional input data is sent to the information processing device 200.
[0119] 9, registered question data may be selected from the list 112b. When the registered question data is selected, the information processing device 200 transmits the registered answer data corresponding to the selected registered question data to the terminal device 600 as output answer data.
[0120] Display area 114A displays text data 114a that prompts the user to enter additional input data, and a list 114b of registered question data obtained as a result of searching question and answer storage unit 310 using the input question data displayed in display area 111 and the additional input data displayed in display area 113.
[0121] In the example of Figure 9, if the registered question data is not selected in the list 114b displayed in the display area 114A and the additional input data shown in the display area 115 is entered, the additional input data is sent to the information processing device 200.
[0122] 9, registered question data may be selected from the list 114b. When the registered question data is selected, the information processing device 200 transmits the registered answer data corresponding to the selected registered question data to the terminal device 600 as output answer data.
[0123] 10A is a sixth diagram showing a display example of the terminal device of the first embodiment. A screen 110B shown in FIG. 10A is an example of a screen displayed on the terminal device 600 in step S611 of FIG.
[0124] A screen 110B shown in FIG. 10A includes display areas 111, 112B, 111A, 114B, and 113.
[0125] 10A displays a list 112b of registered question data similar to the input question data and an operation button 112c. The operation button 112c is an operation button that indicates that the question intended by the user does not exist in the list 112b.
[0126] Display area 111A displays the content selected in display area 112B. In the example of Fig. 10, text data indicating that operation button 112c has been operated is displayed in display area 112B. Text data prompting the input of additional input data is displayed in display area 114B.
[0127] In this embodiment, when multiple pieces of registered question data similar to the input question data are extracted, the multiple pieces of registered question data may be presented to the user of the terminal device 600. When the displayed registered question data is selected, the registered answer data associated with this registered question data in the question and answer storage unit 310 may be used as output answer data.
[0128] In this embodiment, in the example shown in Fig. 10A, notification text data may be displayed in the same manner as in the examples shown in Fig. 8B and Fig. 8C. Fig. 10B is a seventh diagram showing a display example of the terminal device of the first embodiment, and Fig. 10C is an eighth diagram showing a display example of the terminal device of the first embodiment.
[0129] In the screen 110B-1 shown in FIG. 10B, notification text data T1 is displayed in a display area 114B-1 that displays text data indicating a question prompting the user to make an additional input.
[0130] Furthermore, on screen 110B-2 shown in FIG. 10C, notification text data T2 is displayed in an area other than display areas 112B and 114B displayed as balloons on screen 110B-2.
[0131] In this embodiment, by displaying the notification text data in this manner, it is possible to prevent the user from feeling confused while using the chat system 100, and to improve the ease of use of the chat system 100 for the user. Also, in this embodiment, the text data T3 shown in Fig. 8D may be displayed on the screens displayed in Figs. 10A to 10C. Also, the text data shown in Fig. 8D may be displayed on each screen in a display example of the terminal device 600 described later.
[0132] In this way, according to this embodiment, it is possible to promptly present to the user an answer to the question that the user intended.
[0133] (Second embodiment) The second embodiment will be described below with reference to the drawings. The second embodiment differs from the first embodiment in that output response data to input question data is generated by a large-scale language model 511 provided in a server device 500. In the following description of the second embodiment, differences from the first embodiment will be described, and components having the same functional configuration as those in the first embodiment will be assigned the same reference numerals as those used in the description of the first embodiment, and descriptions thereof will be omitted.
[0134] FIG. 11 is a diagram illustrating an example of a system configuration of a chat system according to the second embodiment.
[0135] The chat system 100A of this embodiment is an example of an information processing system including an information processing device 200A, a second storage device 400, a server device 500, and a terminal device 600, all of which are connected via a communication network 1.
[0136] In the chat system 100A, the information processing device 200A, the second storage device 400, and the server device 500 provide a chatbot service to the user of the terminal device 600.
[0137] The second storage device 400 has a document storage unit 410 and a user information storage unit 420. The document storage unit 410 stores user-specific information that is not publicly available.
[0138] A user of chat system 100 may be, for example, a tenant, in which case the user-specific information is tenant-specific information. A user of chat system 100 may also be an individual user, in which case the user-specific information is user-specific information.
[0139] In the following description, user-specific information will be referred to as in-house information, and in-house information will be taken as an example of user-specific information.
[0140] In-house information for each tenant may be stored in the document storage unit 410. Details of the document storage unit 410 will be described later.
[0141] When the information processing device 200A receives input question data from the terminal device 600, it searches the document storage unit 410 stored in the second storage device 400 based on the input question data. If the information processing device 200A extracts multiple pieces of information that could be sources of information for the answer to the question as a search result, it transmits a prompt to the server device 500 requesting the user of the terminal device 600 to generate a question sentence to prompt the user to input additional input data.
[0142] 11, the information processing device 200A and the second storage device 400 are separate devices, but the present invention is not limited to this. The second storage device 400 may be included in the information processing device 200A.
[0143] Next, the document storage unit 410 of the second storage device 400 will be described with reference to Fig. 12. Fig. 12 is a diagram showing an example of the document storage unit.
[0144] Fig. 12 is a diagram illustrating an example of a document storage unit. In the document storage unit 410 of this embodiment, internal company information is stored for each tenant and may be managed for each tenant. Fig. 12 illustrates an example of internal company information managed by a tenant identified by tenant ID "11."
[0145] In the document storage unit 410 of this embodiment, the internal company information is divided into predetermined units and each divided piece of information is stored. In the following description, each divided piece of internal company information is called a chunk. In the document storage unit 410, the chunk may be stored in association with identification information that identifies the chunk.
[0146] 12, one page of internal information in an internal document is defined as one chunk, and the page number is used as identification information to identify the chunk. Note that a chunk is not limited to one page of internal information, and may be, for example, each paragraph or each chapter in an internal document.
[0147] Next, the functional configuration of each device included in the chat system 100A of this embodiment will be described with reference to Fig. 13. Fig. 13 is a diagram illustrating the functions of each device included in the chat system of the second embodiment.
[0148] First, a functional configuration of the information processing device 200 A will be described. The information processing device 200 of this embodiment includes an input receiving unit 251, a search request unit 252A, a determination unit 253A, an information acquisition unit 254A, a prompt generation unit 255A, a request unit 256, and an output unit 257.
[0149] The input receiving unit 251 receives various inputs to the information processing device 200. Specifically, the input receiving unit 251 receives inputs such as input question data from the terminal device 600.
[0150] The search request unit 252A issues a search request to the second storage device 400. Specifically, the search request unit 252A sends to the second storage device 400 a search request from the document storage unit 410 using the input question data as a search key.
[0151] Judgment unit 253A judges, from the search results in response to a search request made to second storage device 400 by search request unit 252A, whether the amount of in-house information obtained from the input question data is sufficient for generating output response data.
[0152] In other words, the determination unit 253A determines whether the amount of in-house information obtained from the second storage device 400 based on the input question data is equal to or greater than a predetermined amount.
[0153] The judgment unit 253A of this embodiment judges that the amount of internal company information obtained from the input question data is equal to or greater than a predetermined amount of information when multiple chunks are obtained as a result of searching the document memory unit 410 of the second memory device 400.
[0154] That is, the determination unit 253A may determine that the information obtainable from the input question data is insufficient when multiple chunks are obtained as a result of searching the document storage unit 410 of the second storage device 400. Furthermore, the determination unit 253A may determine that the information obtainable from the input question data is sufficient when one chunk is obtained as a result of searching the document storage unit 410 of the second storage device 400.
[0155] The information acquisition unit 254A acquires the results of a search in response to a search request from the search request unit 252A from the second storage device 400. In other words, the information acquisition unit 254A acquires chunks that are the search results from the second storage device 400.
[0156] The prompt generation unit 255A generates a prompt to be output to the server device 500. The prompt generation unit 255A generates a prompt that requests the generation of an answer to the question indicated by the input question data. Furthermore, when the determination unit 253A determines that information obtained from the input question data is insufficient, the prompt generation unit 255A generates a prompt that requests the generation of a question sentence for allowing the user to input additional input data.
[0157] Next, the functions of the second storage device 400 will be described. The second storage device 400 includes a document storage unit 410, a search unit 430, and a storage unit 440. The search unit 430 searches the document storage unit 410 in response to a search request from the information processing device 200A, and transmits the search results to the information processing device 200A. When an administrator of the chat system 100A performs an operation to upload in-house information, the storage unit 440 stores the new in-house information in the document storage unit 410.
[0158] Next, the operation of the chat system 100A of this embodiment will be described with reference to Fig. 14. Fig. 14 is a sequence diagram illustrating the operation of the chat system of the second embodiment.
[0159] In the chat system 100A of this embodiment, when the terminal device 600 receives a question input via the input receiving unit 640 (step S1401), the terminal device 600 transmits the input question data to the information processing device 200A (step S1402). Note that the transmitted input question data may be assigned the tenant ID of the tenant to which the user of the terminal device 600 belongs.
[0160] When the input receiving unit 251 of the information processing device 200A receives the input of the input question data, the search request unit 252A sends a search request including the tenant ID to the second storage device 400 (step S1403).
[0161] When the second storage device 400 receives this search request, the search unit 430 searches the document storage unit 410 (step S1404) and transmits the search results to the information processing device 200A (step S1405). Specifically, the second storage device 400 searches the document storage unit 410 using the input question data as a search key and obtains chunks that will be used as information sources for generating an answer.
[0162] When the information processing device 200A acquires the search results from the second storage device 400 using the information acquisition unit 254A, the determination unit 253A determines whether or not a plurality of chunks have been acquired as the search results (step S1406). In other words, the determination unit 253A determines whether or not the number of chunks acquired from the second storage device 400 is equal to or greater than a threshold value. Here, the threshold value is 1.
[0163] If the number of chunks acquired as the search result in step S1406 is one, the information processing device 200A proceeds to step S1418, which will be described later.
[0164] If, in step S1406, multiple chunks are obtained as search results, the information processing device 200A causes the prompt generation unit 255A to generate a prompt requesting the generation of a question sentence to prompt the user of the terminal device 600 to enter further information (step S1407).
[0165] The operations from step S1408 to step S1413 in FIG. 14 are the same as the operations from step S608 to step S613 in FIG. 6, and therefore a description thereof will be omitted.
[0166] When information processing device 200A receives input of additional input data from terminal device 600, search request unit 252A transmits a search request to second storage device 400 using the input question data and additional input data as search keys (step S1414).
[0167] The operations from step S1415 to step S1417 in FIG. 14 are the same as the operations from step S1403 to step S1406, and therefore a description thereof will be omitted.
[0168] In this embodiment, the processes from step S1407 to step S1417 are repeated until the number of chunks acquired as search results from the second storage device 400 becomes one.
[0169] When one chunk is acquired as a search result, the information processing device 200A causes the prompt generation unit 255A to generate a prompt requesting the generation of an answer if it is possible to generate an answer using the acquired chunk as an information source (step S1418). In other words, the prompt generated in step S1418 is a prompt requesting a determination of whether it is possible to generate an answer using the chunk as an information source, and the generation of an answer if it is determined that it is possible to generate an answer.
[0170] Subsequently, information processing device 200A causes request unit 256 to transmit the generated prompt to server device 500 (step S1419).
[0171] If the large-scale language model 511 determines that an answer can be generated and output data including an answer to the input question data is output, the server device 500 transmits the output data to the information processing device 200A (step S1420).
[0172] When the information processing device 200A acquires the output data including the answer from the server device 500, the output unit 257 transmits the output answer data including the answer to the terminal device 600 (step S1421).
[0173] Furthermore, if the large-scale language model 511 determines that it is impossible to generate an answer, the server device 500 transmits output data indicating that it is impossible to generate an answer to the information processing device 200A (step S1422).
[0174] When the information processing device 200A acquires the output data including the answer from the server device 500, the output unit 257 transmits output answer data indicating that it is not possible to generate an answer to the terminal device 600 (step S1423).
[0175] Terminal device 600 causes display control unit 650 to display the output answer data on display 618 (step S1424). That is, if server device 500 determines that an answer can be generated, terminal device 600 displays the answer to the question indicated by the input question data, and if server device 500 determines that an answer cannot be generated, it displays a message to that effect.
[0176] In this embodiment, the processes from step S1407 to step S1417 are repeated until only one chunk is acquired as a search result from the second storage device 400, but this is not limiting. In this embodiment, if the processes from step S1407 to step S1417 are repeated a predetermined number of times but still do not result in one chunk, text data prompting the user to make an inquiry to the person in charge may be displayed on the terminal device 600.
[0177] In other words, when the chunk does not become one even after repeating the processes from step S1407 to step S1417 a predetermined number of times, the output unit 257 of the information processing device 200A outputs information indicating that transition to a manned service is possible to the terminal device 600. In addition, at this time, the output unit 257 may output a display instruction to the terminal device 600 to cause the terminal device 600 to display an operating member for transitioning to a manned service.
[0178] In this case, information processing device 200A may determine, via determination unit 253A, whether the processes from step S1407 to step S1417 have been repeated a predetermined number of times. The predetermined number of times may be, for example, two times. The predetermined number of times may be set in advance by, for example, an administrator of chat system 100A.
[0179] A display example of the terminal device 600 when the chunk count is not one will be described below with reference to Fig. 15. Fig. 15 is a diagram showing a display example of the terminal device of the second embodiment.
[0180] The screen 110-4 shown in FIG. 15 may be displayed on the terminal device 600, for example, when the process from step S1407 to step S1417 in FIG. 14 is repeated twice but the chunk does not become one.
[0181] Screen 110-4 includes display areas 112 to 115 and 116A. Display area 116A displays text data T4 prompting the user to re-enter a revised question, text data T5 indicating a consultation point for the question, and an operation button 118 for switching from the chatbot service provided by chat system 100 to a manned chat service. Note that the content displayed in display area 116A is not limited to the example shown in FIG. 15 as long as it includes an operation member for switching to the chat service.
[0182] In this embodiment, by doing so, it is possible to prevent the chatbot service from repeatedly asking additional questions to the user.
[0183] Next, the prompts generated by the prompt generating unit 255A of this embodiment will be described with reference to FIGS.
[0184] 16 is a first diagram showing an example of a prompt according to the second embodiment. A prompt 20 shown in FIG. 16 is an example of a prompt generated by the prompt generating unit 255A in step S1407 of FIG.
[0185] Prompt 20 includes text data 10a, 10b, and 20a. Text data 20a includes multiple chunks. In other words, the chunks included in text data 20a are multiple chunks extracted as a search result of document storage unit 410 using input question data.
[0186] In this embodiment, by generating such a prompt 20, it is possible to generate a question sentence for the large-scale language model 511 by referring to the results of searching the document storage unit 410 based on the input question data.
[0187] Fig. 17 is a second diagram showing an example of a prompt according to the second embodiment. A prompt 30 shown in Fig. 17 includes text data 30a, 30b, and 30b.
[0188] The text data 30a is text data that requests a determination of whether an answer can be generated based on the chunk that is the information source, generation of an answer if generation is possible, and notification if generation of an answer is impossible.
[0189] The text data 30b indicates chunks that are used as information sources for generating answers. In other words, the chunks included in the text data 30b are single chunks extracted as search results from the document storage unit 410 using the input question data.
[0190] In this embodiment, by generating such a prompt 20, it is possible to have the large-scale language model 511 generate an answer by referring to the results of searching the document storage unit 410 based on the input question data.
[0191] In addition, in this embodiment, by dividing the internal company information included as an information source in each of the prompts 20 and 30 into chunks, the internal company information that serves as an information source can be narrowed down and sent to the server device 500, and highly accurate output data can be obtained from the large-scale language model 511 while satisfying the token restrictions in the large-scale language model.
[0192] (Third embodiment) The third embodiment will be described below with reference to the drawings. The third embodiment differs from the first embodiment in that, when registered question data corresponding to input question data does not exist in the question and answer storage unit 310, output answer data is generated by the large-scale language model 511. In the following description of the third embodiment, differences from the first embodiment will be described, and components having the same functional configuration as those in the first embodiment will be assigned the same reference numerals as those used in the description of the first embodiment, and their description will be omitted.
[0193] FIG. 18 is a diagram illustrating an example of a system configuration of a chat system according to the third embodiment.
[0194] The chat system 100B of this embodiment is an example of an information processing system that includes an information processing device 200B, a first storage device 300A, a second storage device 400A, a server device 500, a terminal device 600, and a terminal device 700, each of which is connected via a communication network 1.
[0195] The first storage device 300A includes a question and answer storage unit 310 and a conversation history storage unit 320. The conversation history storage unit 320 stores conversation history information that associates questions input from the terminal device 600 with answers output to the terminal device 600. The conversation history information may be managed for each tenant that introduces the chat system 100. The tenant in this embodiment may be an organization such as a company, or may be the organization to which a user who uses the chat system 100 belongs.
[0196] The second storage device 400A includes a document storage unit 410 and a user information storage unit 420. The user information storage unit 420 stores, for each tenant, user information relating to users who belong to the tenant.
[0197] The terminal device 700 is mainly used by an administrator who manages the chat system 100B. In this embodiment, the question and answer information stored in the question and answer storage unit 310 of the first storage device 300A, the document information stored in the document storage unit 410 of the second storage device 400A, and the user information stored in the user information storage unit 420 may be stored in each storage unit by the administrator, who is the user of the terminal device 700.
[0198] When the information processing device 200B receives input of input question data from the terminal device 600, it searches the question and answer storage unit 310 of the first storage device 300A. If there is no registered answer data corresponding to the input question data, the information processing device 200B searches the document storage unit 410 of the second storage device 400A to acquire a chunk that serves as an information source. The information processing device 200B then generates a prompt requesting the generation of an answer using the chunk that serves as the information source, and transmits the prompt to the server device 500.
[0199] The conversation history storage unit 320 and the user information storage unit 420 will be described below with reference to FIGS.
[0200] Fig. 19 is a diagram illustrating an example of a conversation history storage unit. The conversation history information stored in the conversation history storage unit 320 illustrated in Fig. 19 includes information items such as a conversation ID, date and time, a tenant ID, a question, and an answer, and is information in which values of each item are associated with each other.
[0201] The value of the item "Conversation ID" is identification information for identifying a conversation. The value of the item "Date and Time" indicates the date and time when input question data was input to the terminal device 600 or the date and time when output answer data was output to the terminal device 600. The value of the item "Tenant ID" is identification information for identifying the tenant to which the user of the terminal device 600 belongs.
[0202] The value of the item "Question" indicates input question data that is input as a question sentence to the terminal device 600. The value of the item "Answer" indicates output answer data that is output to the terminal device 600 as an answer sentence.
[0203] In this way, the conversation history information in this embodiment includes a conversation ID that identifies the conversation history information, the tenant ID of the tenant to which the user of the terminal device 600 belongs, the date and time when the input question data was input or the date and time when the output answer data was output, the input question data, and the output answer data.
[0204] Note that the conversation history information shown in FIG. 19 is an example, and the conversation history information of this embodiment may include items other than the items of information shown in FIG.
[0205] 20 is a diagram illustrating an example of the user information storage unit 420. The user information stored in the user information storage unit 420 of this embodiment is in-house information stored for each tenant, and may be managed for each tenant.
[0206] The user information of this embodiment includes information items such as a tenant ID, a user ID, a user name, and an email address. The value of the item "tenant ID" is identification information that identifies the tenant to which the user identified by the user ID belongs. The value of the item "user ID" is identification information that identifies the user of the terminal device 600. The item "user name" indicates the name of the user identified by the user ID, and the value of the item "email address" indicates the email address of the user identified by the user ID.
[0207] Next, the functions of each device included in chat system 100B of this embodiment will be described with reference to Fig. 21. Fig. 21 is a diagram illustrating the functions of each device included in the chat system of the third embodiment.
[0208] The functional configuration of the information processing device 200B of this embodiment will be described. The information processing device 200B of this embodiment includes an input receiving unit 251, a search request unit 252B, a determination unit 253B, an information acquisition unit 254B, a prompt generation unit 255B, a request unit 256, and an output unit 257.
[0209] The search request unit 252B issues a search request to the first storage device 300A and the second storage device 400A using the input question data as a search key.
[0210] The determination unit 253B determines whether the amount of information acquired from the first storage device 300A based on the input question data is equal to or greater than a predetermined amount, and whether the amount of information acquired from the second storage device 400A based on the input question data is equal to or greater than a predetermined amount. The determination unit 253B also determines whether information has been acquired from the first storage device 300A based on the input question data.
[0211] Specifically, the determination unit 253B determines whether or not registered answer data has been acquired from the first storage device 300A. The determination unit 253B also determines whether or not multiple pieces of registered question data have been acquired from the first storage device 300A. The determination unit 253B also determines whether or not multiple chunks have been acquired from the second storage device 400A.
[0212] The information acquisition unit 254B acquires the search results from the first storage device 300A. The information acquisition unit 254B also acquires the search results from the second storage device 400A.
[0213] The prompt generation unit 255B generates a prompt requesting the generation of a question sentence for obtaining additional input data when a plurality of registered answer data are obtained as search results from the first storage device 300A. The prompt generation unit 255B also generates a prompt requesting the generation of a question sentence for obtaining additional input data when a plurality of chunks are obtained as search results from the second storage device 400A. The prompt generation unit 255B also generates a prompt requesting the determination of whether an answer to the question indicated by the input question data can be generated, the generation of the answer if generation of the answer is possible, and a notification if generation of the answer is impossible.
[0214] In other words, the information processing device 200B of this embodiment searches the referenced storage device in response to receiving input question data, and if the information obtained as the search result is equal to or greater than a predetermined amount of information, outputs a prompt to the server device 500 requesting the generation of a question sentence to obtain additional input data.
[0215] The first storage device 300A and the second storage device 400A are examples of storage devices that are referenced in response to the reception of input question data. Furthermore, the question and answer information stored in the question and answer storage unit 310 and the user-specific information stored in the document storage unit 410 are examples of information that serves as the source of the answer to the question indicated by the input question data.
[0216] The first storage device 300A of this embodiment includes a search unit 330A and a storage unit 340A. The search unit 330A searches the conversation history storage unit 320 and the question and answer storage unit 310 in response to a search request from the information processing device 200B, and transmits the search results to the information processing device 200. When input question data is input and output answer data is output in the information processing device 200B, the storage unit 340A stores conversation history information including the input question data and the output answer data in the conversation history storage unit 320.
[0217] The second storage device 400A of this embodiment includes a search unit 430A and a storage unit 440A. The search unit 430A searches the document storage unit 410 and the user information storage unit 420 in response to a search request from the information processing device 200B, and transmits the search results to the information processing device 200B. The storage unit 440A stores the document information and user information input from the terminal device 700 in the document storage unit 410 and the user information storage unit 420 in the information processing device 200B.
[0218] The terminal device 700 includes a communication control unit 730 , an input receiving unit 740 , and a display control unit 750 .
[0219] The communication control unit 730 controls communication between the terminal device 700 and the information processing device 200B. The input receiving unit 740 receives input to the terminal device 700. The display control unit 750 controls display on the terminal device 700.
[0220] Next, the operation of the chat system 100B of this embodiment will be described with reference to Fig. 22. Fig. 22 is a sequence diagram illustrating the operation of the chat system of the third embodiment.
[0221] In the chat system 100B of this embodiment, when the information processing device 200B receives input question data from the terminal device 600, it first searches the question and answer storage unit 310 of the first storage device 300A. If the question and answer storage unit 310 does not contain registered answer data corresponding to the input question data, the information processing device 200B uses the large-scale language model 511 to obtain an answer to the question indicated by the input question data.
[0222] The processing from step S2101 to step S2105 in FIG. 22 is the same as the processing from step S601 to step S605 in FIG. 6, and therefore the description thereof will be omitted.
[0223] Following step S2105, the information processing device 200B determines, via the determination unit 253B, whether or not registered response data has been acquired in step S2104, and if registered response data has been acquired, whether or not multiple registered response data have been acquired (step S2106).
[0224] In step S2106, if registered answer data has not been acquired, that is, if there is no registered answer data corresponding to the input question data in question and answer storage unit 310, information processing device 200B proceeds to step S2120, which will be described later.
[0225] In step S2106, if registered response data is acquired and the acquired registered response data is one, information processing device 200B proceeds to step S2118, which will be described later.
[0226] In step S2016, if the registered response data is acquired and there are a plurality of pieces of registered response data, information processing device 200B proceeds to step S2107.
[0227] The operations from step S2107 to step S2117 in Fig. 22 are the same as the operations from step S607 to step S617 in Fig. 6, and therefore will not be described again. In this embodiment, the operations from step S2107 to step S2117 are repeated until only one registered response data is acquired as a search result from the first storage device 300A.
[0228] If one registered response data is acquired as the search result, information processing device 200B proceeds to step S2118. The operations of steps S2118 and S2119 in Fig. 22 are the same as the operations of steps S618 and S619 in Fig. 6, and therefore description thereof will be omitted.
[0229] If it is determined in step S2106 that the registered response data has not been acquired, the information processing device 200B causes the search request unit 252B to send a search request to the second storage device 400A (step S2120).
[0230] At this time, the search request unit 252B may, for example, acquire conversation history information immediately before the input question data is input from the conversation history storage unit 320, and send a search request including the input question data and the acquired conversation history information to the second storage device 400A.
[0231] The operations from step S2120 to step S2141 in FIG. 22 are the same as the operations from step S1403 to step S1424 in FIG. 14, and therefore the description thereof will be omitted.
[0232] In this embodiment, by including conversation history information in the search request, chunks serving as information sources for generating answers can be information corresponding to the conversation history. Therefore, according to this embodiment, when generating an answer to a question, the answer can be an answer corresponding to the conversation history. Specifically, for example, in this embodiment, an answer that is the same as the most recent answer may be given a lower priority, or an answer that satisfied the user last time may be given a higher priority based on trends such as the content and timing of the user's question. In this embodiment, by doing so, a more accurate answer to the question can be output.
[0233] In this embodiment, when the question and answer information stored in the question and answer storage unit 310 does not contain registered answer data corresponding to the input question data, the large-scale language model 511 is used.
[0234] Therefore, according to this embodiment, even if no answer-to-question information is stored in the question and answer storage unit 310, the large-scale language model 511 can be used to present the user with an answer to the question indicated by the input question data.
[0235] Next, a method for displaying output response data will be described with reference to Fig. 23 to Fig. 25. Fig. 23 is a first diagram showing an example of displaying output response data in a terminal device of the third embodiment.
[0236] A screen 121 shown in FIG. 23 is an example of a screen displayed on the terminal device 600 in step S2141 of FIG.
[0237] The screen 121 includes display areas 121a and 121b. The display area 121a displays input question data input by the user of the terminal device 600. The display area 121b displays output answer data from the chat system 100, as well as text data 121b-1, a link 121b-2, and text data 121b-3.
[0238] The text data 121b-1 includes information indicating contact information for the user of the terminal device 600 in case of further inquiries. The information indicating contact information may be, for example, an email address or a telephone number. This contact information may be stored in the user information storage unit 420, for example. Specifically, the user information storage unit 420 may store information indicating the work responsibilities of each user in association with the user ID.
[0239] When transmitting output response data to the terminal device 600, the information processing device 200B may obtain the email address of the user in charge of the business corresponding to the question indicated by the input question data, and transmit the email address to the terminal device 600 together with the output response data.
[0240] The link 121b-2 is a link for displaying information related to the content indicated by the input question data or the output answer data.
[0241] The text data 121b-3 is text data that reads, "Answer is generated using a generation AI, so it may be incorrect." In this embodiment, by displaying such text data 121b-3, the user can understand that the accuracy, recency, etc. of the content of the answer are not guaranteed.
[0242] Note that the text data 121b-3 shown in Fig. 23 is an example and is not limited to this. The text data 121b-3 may be text data including content indicating that the answer to the question was generated by a generation AI and that the answer may be incorrect, and the expression thereof is not limited to the example shown in Fig. 23.
[0243] In this embodiment, by notifying the user of the terminal device 600 of contact information for further inquiries and information related to the question and answer along with the output answer data, it is possible to provide the user with an answer that will satisfy them.
[0244] Furthermore, the information processing device 200B of this embodiment may send a notification to the contact information included in the text data 121b-1, indicating that an inquiry has been received from the user of the terminal device 600. By sending such a notification, it is possible to prompt the user who is the contact information when making an inquiry to make preparations for responding to further inquiries.
[0245] Fig. 24 is a second diagram showing an example of how output response data is displayed on the terminal device of the third embodiment. Screen 121A shown in Fig. 24 is another example of the screen displayed on terminal device 600 in step S2141 of Fig. 22.
[0246] The screen 121A includes display areas 121a, 121b, and 121d. In the display area 121d, a message is displayed notifying that the link 121b-2 displayed in the display area 121b has been sent to the email address of the user of the terminal device 600.
[0247] In this case, the user ID of the user of the terminal device 600 may be input to the information processing device 200B together with the input question data, and the information processing device 200B may refer to the user information storage unit 420 and acquire an email address identified from the tenant ID and the user ID. Then, when transmitting output response data to the terminal device 600, the information processing device 200B may transmit the link 121b-2 together with the output response data to the identified email address, and then cause the terminal device 600 to display the display area 121d.
[0248] In the example of Figure 24, by sending link 121b-2 to the user's email address to display information related to the user's inquiry, the user can refer to the information related to the inquiry even after finishing using chat system 100B.
[0249] 25 is a third diagram showing an example of how output response data is displayed on the terminal device of the third embodiment. Screen 121B shown in Fig. 25 is another example of the screen displayed on terminal device 600 in step S2141 of Fig. 22.
[0250] The screen 121B includes display areas 121a and 121e. The display area 121e includes text data 121b-1 indicating contact information for the user of the terminal device 600 to use in case of further inquiries, and a message 121f.
[0251] Message 121f is a message indicating that a link for displaying information related to the user's inquiry has been sent to the user's email address.
[0252] In this way, in the example of FIG. 25, a link for displaying information related to the user's inquiry can be notified to the user separately from the output response data.
[0253] Fig. 25 is a fourth diagram showing an example of display of output response data on a terminal device of the third embodiment. Screen 121C shown in Fig. 25 is another example of a screen displayed on terminal device 600 in step S2141 of Fig. 22. Screen 121C may be displayed when the user's email address is not stored in user information storage unit 420 of second storage device 400A.
[0254] Screen 121C includes display areas 121a, 121b, 121g, 121h, and 121i. Display area 121g displays a message inquiring about the destination of link 121b-2. Display area 121h displays the email address entered by the user. Display area 121i displays a message notifying the user that link 121b-2 has been sent to the email address entered by the user.
[0255] In the example of FIG. 25, even if the user's email address is not stored in the user information storage unit 420, the link 121b-2 can be sent to the user if the user so desires.
[0256] The functions of the above-described embodiments can be realized by one or more processing circuits. Here, the term "processing circuit" in this specification includes a processor programmed to execute each function by software, such as a processor implemented by an electronic circuit, as well as devices such as an ASIC (Application Specific Integrated Circuit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), and conventional circuit modules designed to execute each of the above-described functions.
[0257] Additionally, the devices described in each embodiment represent only one of multiple computing environments for implementing the embodiments disclosed herein.
[0258] In one embodiment, information processing apparatus 200 includes multiple computing devices, such as a server cluster, configured to communicate with each other via any type of communication link, including a network, shared memory, etc., to perform the processes disclosed herein. Similarly, server apparatus 500 and information processing apparatus 200 may include multiple computing devices configured to communicate with each other.
[0259] Furthermore, the information processing device 200 can be configured to share the disclosed processing steps in various combinations. For example, a process performed by a predetermined unit can be executed by the information processing device 200. Similarly, the function of a predetermined unit can be executed by the information processing device 200. Furthermore, in the information processing device 200, each element may be integrated into a single server device or may be separated into multiple devices.
[0260] The information processing device 200 may be any device that has a communication function. The information processing device 200 may be, for example, a PJ (Projector), an IWB (Interactive White Board: a white board with an electronic blackboard function that allows mutual communication), an output device such as digital signage, a HUD (Head Up Display) device, industrial machinery, an imaging device, a sound collection device, medical equipment, a network home appliance, an automobile (Connected Car), a notebook PC (Personal Computer), a mobile phone, a smartphone, a tablet terminal, a game console, a PDA (Personal Digital Assistant), a digital camera, a wearable PC, a desktop PC, or the like.
[0261] For example, aspects of the present invention are as follows. <1> an input receiving unit that receives input question data; an output unit that outputs, to a terminal device, registered answer data associated with the registered question data identified based on the question indicated by the input question data; a request unit that, when the number of registered question data identified based on the question is equal to or greater than a predetermined number, requests a large-scale language model to generate a question sentence for obtaining additional input question data; The output unit outputs the question sentence for obtaining the additional input question data, which is output from the large-scale language model, to a terminal device. <2> The registered answer data associated with the identified registered question data is stored in a first storage device. <1> The information processing device described. <3> the input receiving unit receives input of the additional input question data; the request unit requests the large-scale language model to generate a question sentence for obtaining the additional input question data when the number of registered question data based on the input question data and the additional input question data for the first storage device is equal to or greater than a predetermined number. <2> The information processing device described. <4> The request unit when a plurality of registered question data are acquired as a result of searching the first storage device with the input question data, a request is made to the large-scale language model to generate a question sentence for obtaining the additional input question data; <2> or <3> The information processing device described. <5> The request unit When a plurality of chunks are obtained as a result of searching a second storage device in which user-specific information is stored for each chunk using the input question data, a request is made to the large-scale language model to generate a question sentence for obtaining the additional input question data. <1> ~ <4> 10. The information processing device according to claim 9, wherein: <6> The output unit When a question sentence for obtaining the additional input question data is output to the terminal device the predetermined number of times or more, an instruction to display an operation member for transitioning to a manned chat service is output to the terminal device. <5> The information processing device described. <7> The request unit When one chunk is obtained as a result of searching the second storage device, a request is made to the large-scale language model to determine whether an answer can be generated based on the one chunk, to generate an answer if generation is possible, or to notify the large-scale language model if generation of an answer is impossible; The output unit outputting output data output from the large-scale language model to the terminal device; <6> The information processing device described. <8> The request unit If the registered question data is not acquired as a result of searching the first storage device, and if a plurality of chunks are acquired as a result of searching the second storage device in which user-specific information is stored for each chunk, a request is made to the large-scale language model to generate a question sentence for acquiring the additional input question data. <2> ~ <7> 10. The information processing device according to claim 9, wherein: <9> The output unit outputting to the terminal device information indicating the reason why the question sentence for obtaining the additional input question data is displayed; <1> ~ <8> An information processing device according to any one of the above. <10> An information processing system including an information processing device and a terminal device that communicates with the information processing device, The information processing device includes: an input receiving unit that receives input question data; an output unit that outputs, to a terminal device, registered answer data associated with the registered question data identified based on the question indicated by the input question data; a request unit that, when the number of registered question data identified based on the question is equal to or greater than a predetermined number, requests a large-scale language model to generate a question sentence for obtaining additional input question data; The output unit outputting a question sentence for obtaining the additional input question data, which is output from the large-scale language model, to a terminal device; The terminal device An information processing system having a display control unit that displays a question sentence for obtaining the additional input question data. <11> An information processing method by an information processing system including an information processing device and a terminal device that communicates with the information processing device, The information processing device, A procedure for accepting input question data; outputting, to a terminal device, registered answer data associated with the registered question data identified based on the question indicated by the input question data; and if the number of registered question data identified based on the question is equal to or greater than a predetermined number, requesting a large-scale language model to generate a question sentence for obtaining additional input question data; The terminal device, and a step of displaying a question sentence for obtaining the additional input question data, The output procedure includes: a query sentence for obtaining the additional input question data, the query sentence being output from the large-scale language model, to the terminal device; <12> A process of accepting input question data; outputting, to the terminal device, registered answer data associated with the registered question data identified based on the question indicated by the input question data; if the number of registered question data identified based on the question is equal to or greater than a predetermined number, requesting a large-scale language model to generate a question sentence for obtaining additional input question data; The output process includes: a program that causes the information processing device to execute a process of outputting, to a terminal device, a question sentence for obtaining the additional input question data, the question sentence being output from the large-scale language model;
[0262] Although the present invention has been described above based on the embodiments, the present invention is not limited to the requirements shown in the above embodiments. These requirements can be changed without departing from the spirit of the present invention, and can be appropriately determined depending on the application form. [Explanation of symbols]
[0263] 100, 100A, 100B Chat System 200, 200A, 200B Information processing equipment 251 Input Reception Unit 252 Search Request Section 253 Judgment section 254 Information Acquisition Department 255 Prompt Generation Unit 256 Request part 257 output section 300 First Storage Device 310 Question and answer storage section 320 Conversation History Storage Unit 400 Secondary Storage Device 410 Document Storage Unit 420 User information storage unit 500 Server device 511 Large-scale Language Models 600, 700 terminal equipment [Prior art documents] [Patent documents]
[0264] [Patent Document 1] Japanese Patent Publication No. 2023-076413
Claims
1. an input receiving unit that receives input question data; an output unit that searches a first storage device, which stores question and answer information in which registered question data and registered answer data are associated with each other, using the input question data, and outputs to a terminal device the registered answer data associated with the registered question data identified based on the question indicated by the input question data; a request unit that, when the number of registered question data identified based on the question is equal to or greater than a predetermined number, requests a large-scale language model to generate a question sentence for obtaining additional input question data; when the registered question data identified based on the question does not exist, searches a second storage device in which user-specific information is stored for each chunk using the input question data; and when a plurality of chunks are obtained as a result of the search, requests the large-scale language model to generate a question sentence for obtaining the additional input question data; The output unit outputs the question sentence for obtaining the additional input question data, which is output from the large-scale language model, to a terminal device.
2. the input receiving unit receives input of the additional input question data; The request unit 2. The information processing device according to claim 1, wherein when the number of registered question data identified from the question indicated by the input question data and the question indicated by the additional input question data is equal to or greater than a predetermined number, a request is made to a large-scale language model to generate a question sentence for obtaining the additional input question data.
3. The output unit 2. The information processing device according to claim 1, wherein when a question sentence for obtaining the additional input question data is output to the terminal device a predetermined number of times or more, an instruction to display an operating member for transitioning to a manned chat service is output to the terminal device.
4. The request unit When one chunk is obtained as a result of searching the second storage device, a request is made to the large-scale language model to determine whether an answer can be generated based on the one chunk, to generate an answer if generation is possible, or to notify the large-scale language model if generation of an answer is impossible; The output unit The information processing apparatus according to claim 1 , wherein output data output from the large-scale language model is output to the terminal device.
5. The output unit 2. The information processing apparatus according to claim 1, further comprising: outputting to said terminal device information indicating a reason why a question sentence for obtaining said additional input question data is displayed.
6. An information processing system including an information processing device and a terminal device that communicates with the information processing device, The information processing device includes: an input receiving unit that receives input question data; an output unit that searches a first storage device, which stores question and answer information in which registered question data and registered answer data are associated with each other, using the input question data, and outputs to a terminal device the registered answer data associated with the registered question data identified based on the question indicated by the input question data; a request unit that, when the number of registered question data identified based on the question is equal to or greater than a predetermined number, requests a large-scale language model to generate a question sentence for obtaining additional input question data; when the registered question data identified based on the question does not exist, searches a second storage device in which user-specific information is stored for each chunk using the input question data; and when a plurality of chunks are obtained as a result of the search, requests the large-scale language model to generate a question sentence for obtaining the additional input question data; The output unit outputting a question sentence for obtaining the additional input question data, which is output from the large-scale language model, to a terminal device; The terminal device An information processing system having a display control unit that displays a question sentence for obtaining the additional input question data.
7. An information processing method by an information processing system including an information processing device and a terminal device that communicates with the information processing device, The information processing device, A procedure for accepting input question data; a step of searching a first storage device, which stores question and answer information in which registered question data and registered answer data are associated with each other, using the input question data, and outputting to a terminal device the registered answer data associated with the registered question data identified based on the question indicated by the input question data; if the number of registered question data identified based on the question is equal to or greater than a predetermined number, requesting a large-scale language model to generate a question sentence for obtaining additional input question data; if the registered question data identified based on the question does not exist, searching a second storage device in which user-specific information is stored for each chunk using the input question data; and if a plurality of chunks are obtained as a result of the search, requesting the large-scale language model to generate a question sentence for obtaining the additional input question data; The terminal device, a step of displaying a question sentence for obtaining the additional input question data; The output procedure includes: a query sentence for obtaining the additional input question data, the query sentence being output from the large-scale language model, to the terminal device;
8. A process of accepting input question data; a process of searching a first storage device, which stores question and answer information in which registered question data and registered answer data are associated with each other, using the input question data, and outputting to a terminal device the registered answer data associated with the registered question data identified based on the question indicated by the input question data; if the number of registered question data identified based on the question is equal to or greater than a predetermined number, requesting a large-scale language model to generate a question sentence for obtaining additional input question data; if the registered question data identified based on the question does not exist, searching a second storage device in which user-specific information is stored for each chunk using the input question data; and if a plurality of chunks are obtained as a result of the search, requesting the large-scale language model to generate a question sentence for obtaining the additional input question data; The output process includes: a program that causes the information processing device to execute a process of outputting, to a terminal device, a question sentence for obtaining the additional input question data, the question sentence being output from the large-scale language model;
Citation Information
Patent Citations
Intelligent question and answer method and system combining retrieval and generation
CN112364150A
Question answering control program, question answering control server, and question answering control method
JP2015036945A
Dialog management server, dialog management method, and program
JP2019145102A
Method, computer device, and computer program for providing dialogue dedicated to domain by using language model
JP2023076413A
Interactive information providing method, interactive information providing system, computer device and program
JP2025124049A