Information processing system, information processing method, and program

Abstract

This invention provides an information processing system, information processing method, and program that can clarify the content of a question. [Solution] The system comprises a generation unit that generates content connectors corresponding to content selected or entered by the user, a transmission unit that transmits the connectors to a user terminal, and a display control unit that displays the content on the user terminal in response to the operation of the connectors on the user terminal, wherein the connectors are the connectors of the content identified from the related information obtained by quantifying the content in a language model, searching a vector database for related information associated with the analysis results in which the content has been quantified.

Description

Technical Field

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

Background Art

[0002] Services that respond to questions entered by users are known. For example, Patent Document 1 discloses a technique for transmitting a question to an appropriate response device among a plurality of response devices that output answers to the question so as to quickly obtain an answer to the question.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, the technique of Patent Document 1 had room for improvement in terms of providing an appropriate answer tailored to each individual user. In Patent Document 1, an answer can be obtained from an appropriate response device among a plurality of response devices, but it does not consider the case where the content of the question is not clear. For example, for the question "App A cannot be found", there can be cases such as "I installed App A but don't know where the icon is" and "I'm trying to install App A now but don't know where the file to be installed is located". In order to appropriately answer the user's question "App A cannot be found", it would be better to be able to clarify the content of the question.

[0005] The present invention has been made in view of such a situation, and provides an information processing system, an information processing method, and a program that can clarify the content of a question.

Means for Solving the Problems

[0006] To solve the above-mentioned problems of the present invention, the present invention is an information processing system comprising: a generation unit that generates content connectors corresponding to content selected or entered by a user; a transmission unit that transmits the connectors to a user terminal; and a display control unit that displays the content on the user terminal in response to the operation of the connectors on the user terminal, wherein the connector is the connector of the content identified from the analysis results in which the content is materialized in a language model, and related information related to the materialized content is retrieved from a vector database.

[0007] Furthermore, the present invention relates to an information processing method performed by a computer used in an information processing system, wherein a generation unit generates content connectors corresponding to content selected or input by a user, a transmission unit transmits the connectors to a user terminal, and a display control unit displays the content on the user terminal in response to the operation of the connectors on the user terminal, and the connector is the connector of the content identified from the analysis results in which the content is materialized into a language model, and related information related to the materialized content is retrieved from a vector database.

[0008] Furthermore, the present invention is a program for operating a computer as the information processing system described above, and is a program for causing the computer to function as each part of the information processing system. [Effects of the Invention]

[0009] According to the present invention, it is possible to provide accurate and unambiguous answers to questions, as well as answers that are easy to visually verify. [Brief explanation of the drawing]

[0010] [Figure 1] This is a block diagram showing an example configuration of the customer support system 1 according to the embodiment. [Figure 2] This is a block diagram showing an example configuration of the information processing device 10 according to the embodiment. [Figure 3] This diagram illustrates the user's experience in the customer support system 1 according to this embodiment. [Figure 4] This is a sequence diagram showing the processing flow performed by the customer support system 1 according to the embodiment. [Figure 5] This is a sequence diagram showing the processing flow performed by the customer support system 1 according to the embodiment. [Figure 6] This figure shows an example of the UI displayed on the user terminal 20 according to the embodiment. [Figure 7] This figure shows an example of the UI displayed on the user terminal 20 according to the embodiment. [Modes for carrying out the invention]

[0011] The embodiments of the invention will be described below with reference to the drawings. In the following description, an example will be given of a case where a company receives questions from its customers and provides video content that answers those questions. The embodiments can be applied to various tasks such as customer support, call center support, customer service, and sales where video content is used. In addition, the following description will be an example of a case where video content is provided, but the invention is not limited to this. In this embodiment, content may be provided consisting of any or a combination thereof of text content such as text, image content such as still images and moving images including illustrations and photographs, and sound content such as music, voice, and electronic sounds.

[0012] (Regarding Customer Support System 1) Figure 1 is a block diagram showing an example configuration of a customer support system 1 according to an embodiment. The customer support system 1 includes, for example, an information processing device 10, a user terminal 20, a corporate server 30, a large-scale language model 40, a vector DB 50, and a content DB 60. The customer support system 1 may be configured to include multiple user terminals 20. Furthermore, the functions of the information processing device 10 in the example of Figure 1 may be realized by multiple information processing devices.

[0013] The information processing device 10 is a computer managed by a business operator that has received a customer support request from a company. The information processing device 10 can be, for example, a server or a PC (Personal Computer). The information processing device 10 is connected to the user terminal 20, the company server 30, and the large-scale language model 40 via a communication network NW. The communication network NW includes, for example, some or all of the following: the internet, a WAN (Wide Area Network), a LAN (Local Area Network), provider equipment, a wireless base station, or a dedicated line.

[0014] Furthermore, the information processing device 10 is connected to the vector DB 50 and content DB 60 via wired or wireless communication, without using a communication network NW. By connecting the vector DB 50 and content DB 60 to the information processing device 10 without using a communication network NW, it is possible to suppress the leakage of information stored in the vector DB 50 and content DB 60 to the outside. Highly accurate answers can be generated by using confidential information within a limited scope, such as within the company's own system. However, if confidentiality of the information stored in the vector DB 50 and content DB 60 is not required, the information processing device 10 may be connected to the vector DB 50 and content DB 60 via a communication network NW.

[0015] The user terminal 20 is a computer operated by the user. As the user terminal 20, a smartphone, a tablet terminal, a PC, a mobile phone, a touch pad, a wearable terminal, etc. can be applied. The user terminal 20 communicates with the enterprise server 30 and the information processing device 10 via the communication network NW.

[0016] The enterprise server 30 is a computer managed by the enterprise. As the enterprise server 30, a server, a PC, etc. can be applied. The enterprise server 30 communicates with the user terminal 20 and the information processing device 10 via the communication network NW.

[0017] The large language model 40 is a language model specialized in natural language processing (NLP: Natural Language Processing), and is a kind of so-called generative AI (Artificial Intelligence). When an instruction text called a prompt is input to the large language model 40, it outputs a text generated based on that instruction. As the large language model 40, an LLM (Large Language Model) can be used. For example, Claude (registered trademark) 3.5 sonnet, GPT-4 (registered trademark), Llama2, PaLM2, etc. can be utilized. Note that the large language model 40 is not limited to just an LLM, as long as it can perform at least natural language processing tasks. Small-scale language models, multi-modal language models that can handle images and other formats, etc. can be applied. Also, in this embodiment, the large language model 40 is configured as a server that aggregates data from terminal devices such as the user terminal 20 and performs natural language processing on the aggregated data, but it is not limited to this. For example, an LLM may be provided in a terminal device such as the user terminal 20, and natural language processing may be performed using the LLM provided in the terminal device.

[0018] The content DB 60 is a DB (database) in which information related to video content prepared in advance as an answer to a question is stored. The video content stored in the content DB 60 includes, for example, video content related to answers to each of technical support, troubleshooting, manuals, frequently asked questions, or assumed questions that are assumed to be asked by users.

[0019] All or part of these video contents may be generated by so-called generative AI. For example, by notifying generative AI with an assumed question as a prompt, the generative AI can be made to generate an answer to the assumed question, and a solution video based on the answer generated by the generative AI can be created. The generative AI here may be the large language model 40, or may be different from the large language model 40. When the generative AI is made to generate an answer, so-called hallucination may occur, such that the generated answer contains errors, logical contradictions, or information different from facts. From the perspective of suppressing such hallucination, it is desirable to perform a quality check in which a product generated by generative AI is confirmed in advance by a person in charge (a human). In this case, only the video content whose content has been confirmed to contain no hallucination after passing the quality check by the person in charge (a human) is stored in the content DB 60 as video content to be notified to the user.

[0020] The video content stored in the content DB 60 may be personalized video. Personalized video is video content configured to be an original video for each viewer that displays and explains the name, contract status, etc. of the user who is the viewer. For example, when the information processing device 10 distributes content to the user terminal 20, personal information of users held by the company is notified from the company server 30 or the like to the user terminal 20. Thereby, even if the information processing device 10 does not hold the personal information of the user, a personalized video can be displayed on the user terminal 20.

[0021] Furthermore, the video content stored in the content DB60 may be interactive videos. An interactive video is a video that has input fields for selecting options, and is configured so that the video to be played is selected based on the options entered in those input fields.

[0022] The video content stored in Content DB60 may include narration that explains the content's content. Furthermore, the content stored in Content DB60 may also include subtitles or explanatory text.

[0023] Vector DB50 is a database that stores vector data referenced when searching for video content stored in Content DB60. For example, Vector DB50 can be generated by applying a technique called RAG (Retrieval-Augmented Generation), which references external information when generating responses in a large-scale language model 40. Vector data is data in which the information stored in Content DB60 has been vectorized. Vector data is multidimensional data in which the words and sentences contained in the content are converted into numerical values ​​based on the feature quantities of each dimension, corresponding to the meaning and relationships of each dimension.

[0024] For example, the information processing device 10 generates a summary text using a generation AI or the like, which is a summary of all or part of the text corresponding to the narration, etc., contained in the content stored in the content DB 60.

[0025] For example, the following summary can be generated. How to download app A on iPhone (registered trademark) • Steps to use the App Store (registered trademark) • Points to note when downloading the app

[0026] When such text is vectorized, it becomes high-dimensional vector data consisting of a list of numerical values ​​corresponding to feature quantities in each dimension, such as (1.0546875,0.318359375,0.2578125,-0.0534667969,…). Vector DB50 stores such high-dimensional vector data. When two texts are related, the vector data generated from them is mapped to be close to each other in the vector space. Therefore, related texts or texts with similar meanings can be clustered in the vector space. Vector DB50 can be used to search for texts that correspond to vector data mapped to be close to the vectorized data of the target text, as texts related to the search target.

[0027] (Regarding the information processing device 10) Here, the information processing device 10 will be described using Figure 2. Figure 2 is a block diagram showing an example configuration of the information processing device 10 according to this embodiment. The information processing device 10 comprises a communication unit 11, a storage unit 12, a display control unit 13, a response control unit 14, and a generation unit 15.

[0028] The communication unit 11 communicates with the enterprise server 30 and the large-scale language model 40 via the communication network NW. The communication unit 11 also communicates with the vector DB 50 and the content DB 60 via wired or wireless connections.

[0029] The storage unit 12 can be an HDD (Hard Disk Drive), flash memory, RAM (Random Access Memory), etc. The storage unit 12 stores programs that are executed by the functional units (display control unit 13, response control unit 14, and generation unit 15) of the information processing device 10 in order to perform their functions, as well as data used when the programs are executed.

[0030] The display control unit 13 controls the image to be displayed on the user terminal 20. For example, in response to an operation performed on the user terminal 20 to access customer support, the display control unit 13 displays a screen for providing customer support on the user terminal 20.

[0031] A screen for providing customer support is, for example, an interactive chat screen. By providing an interactive chat screen as a screen for providing customer support, if an unclear question is entered, it becomes easy to notify the user of additional questions to clarify the question and to obtain answers to those additional questions. Therefore, even if an unclear question is entered, the content of the question can be clarified. An interactive chat screen is an example of an "interactive send / receive form."

[0032] Furthermore, customer support is not limited to being conducted through an interactive chat screen. In customer support system 1, multiple methods (multi-channel) can be applied as ways for users to ask questions. For example, customer support may be provided through inquiries made by telephone to a customer support center established by the company. Alternatively, customer support may be provided by notifying the company of questions entered into an inquiry form on the company's website, or by sending question emails to an inquiry email address.

[0033] The response control unit 14 generates answers to user questions. The response control unit 14 generates answers by extracting content videos that answer user questions from the content DB 60. The specific method by which the response control unit 14 generates answers will be explained in detail later.

[0034] The generation unit 15 generates a connector corresponding to the answer. Here, a connector is information that allows the user to obtain the answer. The connector is, for example, a URL (Uniform Resource Locator) that indicates the location where the video content corresponding to the answer is stored. Alternatively, the connector may be identification information that identifies the video content corresponding to the answer.

[0035] The generation unit 15 generates connectors according to the form of customer support. For example, customer support may be provided via an interactive chat screen. In this case, the generation unit 15 generates a URL (a URL indicating the location where the video content corresponding to the answer is stored) as a connector. By generating a URL as a connector, the location of the video content corresponding to the answer can be notified via the customer support chat.

[0036] Alternatively, customer support may be provided through a webpage used by a company to manage customer contract status and change procedures, such as a My Page. Some such My Pages have a pre-embedded video player, and various services are provided when a video is played on that video player. In this case, the generation unit 15 generates identification information for video content that corresponds to the answer and is playable on the video player, as a connector. By generating identification information for video content playable on the video player as a connector, the video content corresponding to the answer can be played on the My Page's video player.

[0037] Here, we will explain the processing flow of the customer support system 1 from the user's perspective using Figure 3. Figure 3 is a diagram illustrating the flow experienced by the user in the customer support system 1 according to this embodiment.

[0038] Step S1: User inquiry is received. The user accesses the inquiry reception and enters their inquiry. The inquiry reception is a form on a chat screen set up for customer support, or a customer support inquiry form provided on a web page, etc. For example, the user enters the question "I can't find app A" into the form via voice or text. The user's inquiry is received when the question entered by the user into the form is notified to customer support system 1.

[0039] Step S2: Inquiries from users are handled via the UI (User Interface). The UI is an interface for providing users with answers to their questions. If a chat screen is provided for customer support, that chat screen corresponds to the UI. Alternatively, if a user's question is received through a customer support inquiry form on a web page, the email address or phone number entered by the user as contact information corresponds to the UI. For example, in customer support system 1, in response to a user's question, "I can't find app A," additional questions such as "Is your smartphone an iPhone (registered trademark) or an Android (registered trademark)?" and "Please tell us specifically how you 'can't find' app A. For example, you can't find the icon for app A, or app A doesn't respond when you try to open it, etc." are handled via the UI.

[0040] Step S3: A search of the DB (Content DB60) is performed. In Customer Support System 1, the DB is searched according to the user's issue identified through the interaction shown in Step S2 above. For example, through multiple additional questions (two in this case) in Step S2 above, the user's issue is identified, and the issue of the question "I can't find App A" notified by the user is identified as "I installed App A on my OS-A (e.g., Android®) smartphone, but I don't know where the icon is." The DB is searched according to the user's issue identified in this way.

[0041] Step S4: Question and answer matching takes place. In Customer Support System 1, video content (solution content) that resolves the question notified by the user is extracted by searching the database in Step S3 described above. If no video content corresponding to the answer is found after searching the database, the system may switch to human support by linking with a call center or inquiry form.

[0042] Step S5: Present the user with a solution video (video content that resolves the question notified by the user). The customer support system 1 sends the content extracted in step S4 above to the user terminal 20 as a solution video. The DB (content DB 60) only contains content videos whose quality has been checked in advance by a staff member (human) from the perspective of suppressing hallucination. Therefore, the user can be presented with solution content that has been produced and checked by a staff member (human). Interactive videos may also be provided as solution videos to be presented to the user. By providing interactive videos, only videos that are appropriate as answers to the user's questions can be played.

[0043] Here, we will explain the processing flow of the customer support system 1 using Figures 4 and 5. Figures 4 and 5 are sequence diagrams showing the processing flow of the customer support system 1 according to this embodiment. Here, we will mainly explain the specific method by which the response control unit 14 generates a response using Figures 4 and 5. In generating a response, the response control unit 14 performs a problem estimation process to clarify the question and a matching process to extract content videos that match the question. The processes shown in steps S105 to S112 in Figure 4 correspond to the problem estimation process. The processes shown in steps S113 to S116 in Figure 5 correspond to the matching process.

[0044] Step S101: The user terminal 20 accesses a website or a customer support page. In response to the user accessing the website or customer support page, the user terminal 20 sends a notification to the corporate server 30 indicating that the operation has been performed.

[0045] Step S102: The enterprise server 30 sends an interface display request to the information processing device 10 in response to the user terminal 20 accessing a website or a customer support page. An interface display request is a notification requesting that an interface for receiving questions in customer support be displayed on the user terminal 20. This interface could be, for example, an interactive chat screen or a customer support inquiry form. The enterprise server 30 notifies the information processing device 10 that the user has accessed a website or a customer support page and requests that an interface for receiving questions in customer support be displayed on the user terminal 20.

[0046] Step S103: The display control unit 13 of the information processing device 10 displays an interface, such as a conversational chat screen or a customer support inquiry form, on the user terminal 20 in response to a request from the corporate server 30.

[0047] Step S104: The user enters a question by voice or text into an interface displayed on the user terminal 20, such as a conversational chat screen or a customer support inquiry form. In response to the user's action to send a question, the user terminal 20 sends the question entered by the user to the information processing device 10.

[0048] Step S105: The answer control unit 14 of the information processing device 10 sends a text analysis request to the large-scale language model 40. A text analysis request is a notification that requests the large-scale language model 40 to generate a more specific question based on the natural language indicating the question notified by the user. The answer control unit 14 outputs the text of the question notified by the user to the large-scale language model 40 and requests the large-scale language model 40 to generate a more specific question from the question notified by the user.

[0049] Here, when the response control unit 14 receives an inquiry from a user via telephone or other means, it converts the received voice into text (natural language) by speech recognition and notifies the large-scale language model 40 of the textualized question. Even when a question is received via voice, by converting the voice into text, it becomes possible to request the large-scale language model 40, which handles natural language, to generate a more specific question.

[0050] First, the response control unit 14 notifies the large-scale language model 40 of the text of the question (first text) notified by the user, and causes the model 40 to generate a second text from the notified first text. The second text is a search text used to obtain similarity with the vector DB 50, and is a formatted text that summarizes the content of the first text, corrects inaccurate expressions and typographical errors in the first text as needed, and removes words irrelevant to the search. Next, the response control unit 14 converts the second text into vector data and extracts data from the vector DB 50 that has a similarity to the vector data of the second text that is equal to or greater than a threshold, as vector data of text similar to the second text (third text). The response control unit 14 then notifies the large-scale language model 40 of the text (third text) corresponding to the vector data extracted from the vector DB 50, causing it to generate more specific questions.

[0051] Step S106: The response control unit 14 obtains the analysis results from the large-scale language model 40. The analysis results are the results of the analysis corresponding to the text analysis request in step S105, and are more specific questions generated based on the questions notified by the user. The large-scale language model 40 generates more specific questions in response to prompts notified by the information processing device 10. For example, the large-scale language model 40 generates questions that make the question "I can't find app A" more specific, such as "The icon for app A is not displayed on the smartphone screen" and "The chat history for app A is not displayed." The large-scale language model 40 sends the generated products (more specific questions) to the information processing device 10. The response control unit 14 of the information processing device 10 obtains the generated products (more specific questions) generated by the large-scale language model 40 as analysis results.

[0052] Step S107: The response control unit 14 searches for relevant information from the vector DB 50. Relevant information is information related to the product (more specific question) generated by the large-scale language model 40. The response control unit 14 searches the vector DB 50 using the product (more specific question) obtained from the large-scale language model 40 and the conversation history as input. The conversation history here is the history of conversations exchanged between the user and customer support. For example, the conversation history is the history of chats conducted via an interactive chat screen related to customer support. The response control unit 14 searches all the data stored in the vector data stored in the vector DB 50 and extracts vector data as relevant information whose similarity to the vectorized data of the input string (string corresponding to the more specific question and conversation history) is above a threshold.

[0053] Step S108: The response control unit 14 retrieves vector data retrieved from the vector DB 50 as search results for related information. The response control unit 14 retrieves content corresponding to strings such as "How to use App A", "How to use App A on devices other than smartphones", and "How to download App A on iPhone (registered trademark)" as search results.

[0054] Step S109: The response control unit 14 sends an additional question generation request to the large-scale language model 40. An additional question generation request is a notification requesting the generation of additional questions. Additional questions are questions to clarify the content of questions asked by the user in the conversation history, and questions to identify which content video corresponding to the vector data obtained from the search is the appropriate answer for the user. The response control unit 14 makes an additional question generation request by sending the search results from the vector DB 50 to the large-scale language model 40 and requesting the large-scale language model 40 to generate additional questions based on the search results. Here, the response control unit 14 may also notify the large-scale language model 40 of the conversation history and a summary of the conversation history along with the search results from the vector DB 50 to request the generation of additional questions. Alternatively, instead of the conversation history and a summary of the conversation history, other attribute information or information selected by the user in the video may be notified to the large-scale language model 40 along with the search results from the vector DB 50. For example, the response control unit 14 retrieved answer A, answer B, and answer C as candidate answers to the question "I can't find app A". The large-scale language model 40 is notified with a prompt such as, "Generate additional questions that will allow the user to determine which of these answer candidates is the appropriate answer," and is requested to generate additional questions.

[0055] Step S110: The response control unit 14 receives additional questions generated by the large-scale language model 40. The large-scale language model 40 generates additional questions in response to prompts notified by the response control unit 14 of the information processing device 10 and sends the generated additional questions back to the information processing device 10. For example, the large-scale language model 40 generates additional questions to better understand the question "I can't find app A": "Is your smartphone an iPhone® or an Android®?" and "Please tell us specifically how you 'can't find' app A. For example, you can't find the icon for app A, or app A doesn't respond when you try to open it." The large-scale language model 40 sends the generated items (additional questions) to the information processing device 10. The response control unit 14 receives the generated items (additional questions) generated by the large-scale language model 40.

[0056] Step S111: The response control unit 14 asks additional questions to the user terminal 20. The response control unit 14 sends the generated product (additional questions) obtained from the large-scale language model 40 to the user terminal 20 and asks additional questions. The response control unit 14 asks additional questions in a format that continues the conversation that has been exchanged with the user so far in customer support, for example, by sending the additional questions to the user terminal 20 via an interactive chat related to customer support.

[0057] Step S112: The response control unit 14 obtains answers to additional questions from the user terminal 20. The user terminal 20 displays the additional questions notified by the information processing device 10, "Is your smartphone an iPhone (registered trademark) or an Android (registered trademark)?" and "Please tell us specifically how app A 'cannot be found.' For example, the icon for app A cannot be found, or app A does not respond when you try to open it, etc." on the chat screen related to customer support. The user views the additional questions displayed on the chat screen of the user terminal 20 and enters answers to the additional questions into the chat screen. For example, the user enters an answer such as "It's an Android (registered trademark). I can't find the icon for app A," into the chat screen and taps the send button. In response to the tap of the send button, the user terminal 20 sends the information entered by the user (answers to the additional questions) to the information processing device 10. When the information processing device 10 receives information (answers to additional questions) from the user terminal 20, the answer control unit 14 obtains the answers to the additional questions.

[0058] Step S113: As shown in Figure 5, the response control unit 14 searches for content that solves the problem (video content that solves the question notified by the user). Based on the answers to the additional questions, the response control unit 14 identifies the problem of the question notified by the user. For example, suppose the additional questions are "Is your smartphone an iPhone® or an Android®?" and "Please tell us specifically how you 'cannot find' app A. For example, you can't find the icon for app A, or app A doesn't respond when you try to open it." and the response is "It's an Android®. I can't find the icon for app A." In this case, the response control unit 14 identifies the problem of the question "I can't find app A" notified by the user as "I installed app A on my Android® smartphone, but I don't know where the icon is." Based on the identified problem, the response control unit 14 searches the vector DB 50 and searches for information related to the identified problem. The response control unit 14 searches all the data stored in the vector DB 50, and extracts vector data whose similarity to the vectorized string representing the problem is above a threshold as information related to the problem.

[0059] Step S114: The answer control unit 14 obtains a content ID corresponding to the vector data retrieved from the vector DB 50. The content ID is identification information that identifies the content. In steps S115 to S116, described later, the content corresponding to the solution video is retrieved from the content DB 60, and the retrieved solution video content is sent to the user terminal 20 in step S117. For this reason, in this step, the content ID corresponding to the vector data retrieved from the vector DB 50 in step S113 is obtained as identification information for information related to the problem. Furthermore, since providing the user with multiple content videos as solution videos makes it easier for the user to obtain an answer that matches their question, it is desirable for the answer control unit 14 to obtain multiple (one or more) content IDs in this step.

[0060] Step S115: The response control unit 14 refers to the content DB 60 and requests the content DB 60 to send the content corresponding to the content ID obtained in step S114.

[0061] Step S116: In response to the request from the response control unit 14 notified in step S114, the content DB 60 transmits the content corresponding to the content ID notified by the response control unit 14 to the information processing device 10.

[0062] Step S117: The information processing device 10 transmits content to the user terminal 20. The communication unit 11 of the information processing device 10 retrieves the content transmitted from the content DB 60 in step S116. The generation unit 15 of the information processing device 10 generates a connector for the content transmitted from the content DB 60. The display control unit 13 of the information processing device 10 displays the connector for the content corresponding to the solution video on the chat screen of the user terminal 20. As a result, the information processing device 10 transmits the content corresponding to the solution video to the user terminal 20.

[0063] Step S118: The user views the content. The user taps the connector (the connector for the content corresponding to the solution video) displayed on the chat screen of the user terminal 20. In response to the user's tap, the user terminal 20 sends a message to the information processing device 10 indicating that an operation to play the content corresponding to the solution video has been performed. The communication unit 11 of the information processing device 10 receives information from the user terminal 20 (information indicating that an operation to play the content corresponding to the solution video has been performed). The display control unit 13 of the information processing device 10 displays the solution video (an example of content) on the user terminal 20 based on the information received by the communication unit 11. As a result, the content corresponding to the solution video is played on the user terminal 20. The user views the content.

[0064] Step S119: The information processing device 10 requests feedback from the user terminal 20. Feedback here refers to areas for improvement and evaluations of customer support. For example, the display control unit 13 of the information processing device 10 displays phrases such as "Has the problem been resolved?" along with options such as "Resolved," "Resolved but not satisfactory," and "Not resolved" on the chat screen of the user terminal 20. This prompts the information processing device 10 to request feedback from the user terminal 20. Here, the information processing device 10 may also notify the user terminal 20 of an interactive video or an interactive image as the feedback it requests. An interactive video is an interactive video in which feedback options are displayed, and user feedback is obtained when the user selects one of the displayed options (e.g., by touching the screen). An interactive image is an image in which feedback options are displayed, and user feedback is obtained when the user selects one of the displayed options (e.g., by touching the screen).

[0065] Step S120: The user terminal 20 sends feedback to the information processing device 10. The user performs an operation to select an option displayed on the chat screen of the user terminal 20 and performs an operation to send feedback. In response to the selection of an option and the sending operation, the user terminal 20 sends the option selected by the user to the information processing device 10. In this way, the user terminal 20 sends feedback to the information processing device 10. If a touchable video or touchable image is notified as feedback, the user terminal 20 sends the option selected by the user to the information processing device 10 in response to the selection of an option in the touchable video or touchable image.

[0066] If the user is notified of options indicating dissatisfaction with customer support, such as "resolved but not satisfactory" or "not resolved," the information processing device 10 may prompt the user terminal 20 to request more specific feedback. More specific feedback would include, for example, displaying a message on the chat screen such as "Please describe specifically how the issue was not resolved." This allows the information processing device 10 to request more specific feedback from the user terminal 20. The information processing device 10 may also request more specific feedback using video or audio.

[0067] Step S121: The information processing device 10 refers to the vector DB 50 based on the information fed back from the user terminal 20. The communication unit 11 of the information processing device 10 receives information indicating the options sent from the user terminal 20 in step S120. If the options sent from the user terminal 20 indicate that the user is not satisfied with customer support, the display control unit 13 of the information processing device 10 causes the user terminal 20 to display wording requesting more specific feedback. The user views the wording displayed on the user terminal 20 (wording requesting specific feedback) and, in response to this wording, inputs wording that specifically explains the perspective from which the user is asking the question. In response to the input of the specific explanation and the transmission operation being performed, the user terminal 20 sends to the information processing device 10 wording that specifically explains what aspects the user is dissatisfied with. The communication unit 11 of the information processing device 10 receives information indicating the wording sent from the user terminal 20 (wording that specifically explains the perspective from which the user is asking the question). The information processing device 10 refers to the vector DB 50 based on the wording (wording that specifically describes the perspective from which the user is asking the question) that is fed back from the user terminal 20. The response control unit 14 searches all the vector data stored in the vector DB 50 and extracts vector data whose similarity to the vectorized data of the input wording (wording that specifically describes the perspective from which the question is asking) is above a threshold. Furthermore, in this step, the information processing device 10 may, after receiving feedback from the user terminal 20, generate search feedback via the large-scale language model 40, and then use the search feedback to search the vector DB 50. In this case, the response control unit 14 notifies the large-scale language model 40 of the feedback (first feedback) received from the user terminal 20, and causes the model 40 to generate second feedback from the notified first feedback. The second feedback is a text sentence for searching to obtain similarity with the vector DB 50, and is a formatted text sentence that corrects inaccurate expressions and typos in the first feedback and removes words irrelevant to the search. The response control unit 14 then converts the second feedback into vector data, and extracts data from the vector DB 50 that has a similarity of above a threshold to the vector data of the second feedback as vector data of text (third feedback) related to the second feedback.

[0068] Step S122: The information processing device 10 displays hints for creating new content. The information processing device 10 obtains a content ID corresponding to the vector data retrieved from the vector DB 50 in step S121. The information processing device 10 displays the wording fed back from the user terminal 20 (wording that specifically explains the perspective from which the question is being asked), the content ID corresponding to the vector data retrieved from the vector DB 50 based on that wording, and the summary text corresponding to the content ID. This makes it possible to display the user's question in association with the content ID and summary text of the content related to that question. This makes it easy to verify whether the user's problem was correctly identified through additional questions, and whether an appropriate solution video that solves the user's problem was selected. For example, if it is determined that the user's problem was not correctly identified through additional questions, measures can be taken to review each step of the problem estimation process, such as improving the prompt notified to the large-scale language model 40 in step S105. Alternatively, if it is determined that the user's problem has been correctly identified but the appropriate solution video has not been selected, the matching process can be reviewed, and measures can be taken, such as pre-producing video content that corresponds to the user's problem notified in this instance. Furthermore, in this step, the system may be configured to display not only the text provided as feedback from the user terminal 20, but also the content of the question entered by the user in step S104, the answer to the additional question entered by the user in step S112, and the content ID identified from the vector DB 50 in step S114. By displaying these as targets for analysis when creating new content, it becomes possible to produce solution videos with greater accuracy.

[0069] Here, an example of a chat screen (UI) displayed on the user terminal 20 will be explained using Figures 6 and 7. Figures 6 and 7 are diagrams showing examples of the UI displayed on the user terminal 20 according to the embodiment.

[0070] Figure 6 shows the interaction involved in the problem estimation process. First, the user notifies customer support with the question, "I can't find app A." In response to this question, customer support notifies the user with additional questions such as, "...we would like to ask you the following questions... Is your smartphone an OS-A device (for example, Android®) or an OS-B device (for example, iPhone®)?" In response to the additional questions, the user notifies customer support with the question, "It's an OS-A device (for example, Android®)..." In response to this answer, customer support notifies the user with a second additional question, "Could you please answer the following questions?"

[0071] In this way, multiple additional questions may be asked in customer support system 1. For example, if the number of generated products (specific questions) in step S106 exceeds a threshold, the system may be configured to divide the generated products (specific questions) into multiple groups and ask additional questions appropriate to each group.

[0072] Figure 7 shows the interactions involved in the matching process. Once the user's issue is identified through the interactions shown in Figure 6, customer support notifies the user with their opinion, "...I understand the situation well. ...We estimate that the user's initial concern is that the icon is not being displayed." Subsequently, customer support notifies the user with solutions in text, such as "We recommend that you try the following methods..." Furthermore, customer support states, "...We have prepared a solution video..." and the link (in this case, a URL) to the solution video content is displayed on the chat screen of the user's terminal 20.

[0073] As described above, the customer support system 1 of this embodiment (an example of an information processing system) comprises a generation unit 15 and a display control unit 13. The generation unit 15 generates a connector for a solution video (an example of content) corresponding to a question from the user (an example of content selected or entered by the user). The display control unit 13 displays the solution video (an example of content) on the user terminal 20 in response to the user operating the connector. The customer support system 1, upon receiving a question from the user, materializes the content of the question in a large-scale language model 40 (an example of a language model). The customer support system 1 searches the vector DB 50 (an example of a vector database) for related information associated with the analysis result in which the content of the question has been materialized. The customer support system 1 sends an additional question generated based on the related information retrieved from the vector DB 50 and the content of the question to the user terminal 20. As a result, the customer support system 1 of this embodiment can notify the user terminal 20 of the additional question generated using the analysis result in which the content of the question has been materialized and related information, and the content of the question can be clarified by obtaining an answer to the additional question from the user.

[0074] Furthermore, in the customer support system 1 of this embodiment, the content of the question is selected or entered into an interactive chat screen (an example of an interactive send / receive form). This allows for interactive additional questions to be asked in response to the question.

[0075] As described above, the customer support system 1 of this embodiment comprises a generation unit 15, a communication unit 11 (an example of a transmission unit), and a display control unit 13. The generation unit 15 generates a connector for a solution video (an example of content) corresponding to a question from the user (an example of content selected or entered by the user). The communication unit 11 transmits the connector to the user terminal 20. The display control unit 13 displays the solution video (an example of content) on the user terminal 20 in response to the connector being operated on the user terminal 20. The connector is a connector for content identified from the analysis results in which the content of the question is materialized in a large-scale language model 40 (language model), and related information related to the analysis results in which the content of the question is materialized is retrieved from the vector DB 50 (vector database). As a result, the customer support system 1 of this embodiment can identify a solution video using the analysis results in which the content of the question is materialized and related information, and can identify a solution video that resolves the question after clarifying the content of the question.

[0076] Furthermore, in the customer support system 1 of this embodiment, the content of the question is selected or entered into an interactive chat screen (an example of an interactive sending and receiving form). This allows for the interactive presentation of a video solution to the question.

[0077] As described above, the customer support system 1 of this embodiment comprises a generation unit 15 and a display control unit 13. The generation unit 15 generates a connector for a solution video (an example of an image or video content that serves as the answer) that corresponds to the content selected or entered by the user on an interactive chat screen (an example of an interactive sending and receiving form). The display control unit 13 displays the solution video (an example of an image or video content that serves as the answer) on the user terminal 20 in response to the user's operation of the connector. As a result, the customer support system 1 of this embodiment can use an interactive chat screen (an example of an interactive sending and receiving form), so if the content of the user's question is unclear, the system can interactively confirm with the user and clarify the content of the question. Because the question can be answered after it has been clarified, an accurate and error-free answer can be provided. Furthermore, since a solution video (an example of an image or video content that serves as the answer) is provided as the answer, the answer can be made visually easy to confirm. Therefore, it is possible to suppress the giving of ambiguous answers to questions.

[0078] The customer support system 1 or information processing device 10 in the above-described embodiment may be implemented in whole or in part using a computer. In that case, the program for implementing this function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be loaded into the computer system and executed. Here, "computer system" includes hardware such as the OS and peripheral devices. Furthermore, "computer-readable recording medium" refers to portable media such as flexible disks, magneto-optical disks, ROMs, CD-ROMs, and storage devices such as hard disks built into the computer system. In addition, "computer-readable recording medium" may also include those that dynamically hold programs for a short period of time, such as communication lines used when transmitting programs via networks such as the Internet or communication lines such as telephone lines, and those that hold programs for a certain period of time, such as volatile memory inside the computer system that acts as a server or client in that case. Furthermore, the above-mentioned program may be for implementing a part of the above-mentioned function, or it may be a program that can implement the above-mentioned function in combination with a program already recorded in the computer system, or it may be implemented using a programmable logic device such as an FPGA.

[0079] While embodiments of this invention have been described in detail above with reference to the drawings, the specific configuration is not limited to these embodiments and includes designs and the like that do not depart from the spirit of this invention. [Explanation of Symbols]

[0080] 1…Customer support system (information processing system) 10…Information Processing Devices 11… Communications Department 12...Storage section 13…Display Control Unit 14…Response Control Unit 15…Generation section 20…User terminal 30… Enterprise Servers 40...Large-scale language models 50…Vector DB 60…Content DB

Claims

1. A generation unit that generates content connectors corresponding to the content selected or entered by the user, A transmitting unit that transmits the aforementioned connector to the user terminal, A display control unit that displays the content on the user terminal in response to the operation of the connector on the user terminal, Prepare, The aforementioned connector is the connector of the content identified from the aforementioned content, which embodies the content in a language model, retrieves related information from a vector database that is associated with the analysis result in which the content has been embodied, and embodies the content identified from the aforementioned related information. Information processing system.

2. The above content is selected or entered into the interactive send / receive form. The information processing system according to claim 1.

3. Information processing performed by a computer used in an information processing system, The generation unit generates content connectors corresponding to the content selected or entered by the user. The transmitting unit transmits the connector to the user terminal. The display control unit causes the content to be displayed on the user terminal in response to the operation of the connector on the user terminal. The aforementioned connector is the connector of the content identified from the aforementioned content, which embodies the content in a language model, retrieves related information from a vector database that is associated with the analysis result in which the content has been embodied, and embodies the content identified from the aforementioned related information. Information processing methods.

4. A program for operating a computer as an information processing system according to claim 1, wherein the program causes the computer to function as each part of the information processing system.

Images