Information processing apparatus, information processing method, and program

Abstract

Providing an appropriate answer to the user's question in a short time. 【Solution means】An acquisition unit that acquires content targeted by a question from a user, a division unit that divides the content into a plurality of regions according to the context, a first generation unit that generates a summary for each of the plurality of regions, a second generation unit that generates index information associating the summary, the category of the content that is the source of the summary, and a label indicating the content that is the source of the summary, a reception unit that receives a question associated with the category and the label, a narrowing-down unit that narrows down the index information based on the category and the label associated with the question, an extraction unit that extracts the index information relevant to the question from the narrowed-down index information, and a third generation unit that generates an answer to the question based on the extracted index information. An information processing apparatus comprising:

Description

Technical Field

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

Background Art

[0002] Conventionally, techniques for making it easier to obtain the answers that users want are known.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the conventional technology, there is room for further improvement in providing appropriate answers to user questions in a short time.

[0005] The present application has been made in view of the above, and an object thereof is to provide an appropriate answer to a user's question in a short time.

Means for Solving the Problems

[0006] The information processing apparatus according to the present application includes an acquisition unit that acquires content to be the object of a question from a user, a division unit that divides the content into a plurality of regions according to the context, a first generation unit that generates a summary for each of the plurality of regions, a second generation unit that generates index information associating the summary, the category of the content that is the source of the summary, and a label indicating the content that is the source of the summary, a reception unit that receives a question associated with the category and the label from the user, a narrowing-down unit that narrows down the index information based on the category and the label associated with the question, an extraction unit that extracts the index information relevant to the question from the narrowed-down index information, and a third generation unit that generates an answer to the question based on the extracted index information.

Effect of the Invention

[0007] According to one aspect of the embodiment, there is an effect that an appropriate answer to a user's question can be provided in a short time.

Brief Description of the Drawings

[0008]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Mode for Carrying Out the Invention

[0009] Hereinafter, embodiments for implementing an information processing apparatus, an information processing method, and a program according to the present application (hereinafter referred to as "embodiments") will be described in detail with reference to the drawings. Note that the information processing apparatus, information processing method, and program according to the present application are not limited by this embodiment. In addition, the same parts in the following embodiments are denoted by the same reference numerals, and redundant descriptions are omitted.

[0010] (Embodiment) [1. Outline of Information Processing] Using FIG. 1, the outline of the information processing according to the embodiment will be described. FIG. 1 is an explanatory diagram for explaining the outline of the information processing according to the embodiment. As shown in FIG. 1, the outline R includes outlines R1 to R6. The outline R1 is the acquisition of contents such as rules, web contents, and store information (such as sales history). Answers to the questions of the user U are generated using the sentences in this content. The outline R2 is an index builder, and the content acquired in the outline R1 is indexed. The outline R3 is the acquisition of the indexed information (index information) indexed in the outline R2. The outline R4 is a dispatcher, and tasks based on the operations of the user U are executed. For example, a task for answering the questions received from the user U is executed. The outline R5 is a prompt builder, and a prompt is generated using the index information acquired in the outline R3 and the questions received in the outline R4. For example, the index information corresponding to the questions received in the outline R4 is extracted, and a prompt is generated from the index information. The outline R6 is a generative AI model such as GPT, and an answer to the question of the user U is generated based on the prompt generated in the outline R5. The answer generated in the outline R6 is provided to the user U via the dispatcher.

[0011] 〔2. Configuration of Information Processing System〕 The information processing system 1 shown in FIG. 2 will be described. As shown in FIG. 2, the information processing system 1 includes a terminal device 10 and an information processing device 100. The terminal device 10 and the information processing device 100 are communicably connected by wire or wirelessly via a predetermined communication network (network N). FIG. 2 is a diagram showing a configuration example of the information processing system 1 according to the embodiment.

[0012] The terminal device 10 is an information processing device used by a user who asks questions. The user is, for example, a user who asks questions regarding the content of agreements, laws, rules, manuals, etc. For example, a user who asks a question such as "What is the upper limit of the accommodation fee in the company's accommodation regulations for business trips?" The user inputs a question using, for example, a chat tool that enables text-based communication. The terminal device 10 can be any device as long as it can implement the processing in the embodiment. Also, the terminal device 10 can be a device such as a smartphone, a tablet terminal, a notebook PC, a desktop PC, a mobile phone, a PDA, etc. FIG. 4 described later shows the case where the terminal device 10 is a smartphone.

[0013] The terminal device 10 is, for example, a smart device such as a smartphone or a tablet, and is a portable terminal device that can communicate with an arbitrary server device via a wireless communication network such as 4G - 5G (Generation) or LTE (Long Term Evolution). Also, the terminal device 10 has a screen such as a liquid crystal display, which has the function of a touch panel, and may receive various operations on display data such as content, such as tap operations, slide operations, and scroll operations, from the user using a finger, a stylus, etc. FIG. 4 described later shows the case where the terminal device 10 is used by the user U1.

[0014] The information processing apparatus 100 is an information processing apparatus aimed at providing an appropriate answer to a user's question, and it can be any apparatus as long as it can realize the processing in the embodiment. The information processing apparatus 100 performs two major types of processing. The first is the processing before receiving a question from the user, which is processing to enable an appropriate answer to be provided when a question is received from the user. The second is the processing when a question is received from the user, which is processing to generate an answer to the user's question using the information of the result of the first processing. Specifically, in the first processing, the information processing apparatus 100 generates information (hereinafter, appropriately referred to as "index information") obtained by summarizing content (such as regulations, laws, rules, and manuals) for each chapter, each article (clause), each sentence, each gist (gist with a predetermined meaning attached in advance), etc., in a predetermined unit (hereinafter referred to as "area" in the embodiment). Then, the generated index information is used in the second processing. Also, in the second processing, when the information processing apparatus 100 receives a question from the user, it inputs the index information generated in the first processing into the generation AI to generate an answer to the question.

[0015] Here, the index information according to the embodiment will be described. For generating the index information according to the embodiment, a LAMA index or the like is used. The LAMA index is a tool for generating context from a large amount of user data, and is a tool for performing data search by inputting a question received from the user and personal user data. That is, the LAMA index accesses an information source such as a database and assists in generating a prompt. The index information is structured by organizing the data imported from such an information source or the like so as to be easy to search. By using such index information, the information processing apparatus 100 can efficiently access private information when generating a prompt and generate a context including the private information. When the information processing apparatus 100 receives a question in natural language from the user, it searches for index information that matches the question from the user using a LAMA index or the like and inputs it into the generation AI.

[0016] The generative AI is, for example, a text generation model that has learned from data published on the Internet, etc. For example, it is a GPT (Generative Pre-trained Transformer) model that generates an answer to a question. Such a model generates an answer to a question from the data published on the Internet. Since new data is published on the Internet every day, the answers by the model will be updated every day. That is, even if the same question is asked, the data referred to by the model when generating an answer will be different depending on the timing of the question, etc., so different answers will be generated. In addition, since the model can provide detailed answers to questions in a wide range of fields, a wide variety of questions are asked by users.

[0017] Note that in FIG. 2, the case where the terminal device 10 and the information processing device 100 are separate devices is shown, but the terminal device 10 and the information processing device 100 may be integrated.

[0018] [3. An Example of Information Processing] Conventionally, techniques for making it easier for a user to obtain a desired answer are known. For example, a technique for providing an in-house chatbot system by utilizing in-house data is known (the above Patent Document 1). Also, for example, a technique for dividing regulations, manuals, etc. into predetermined units such as chapters and paragraphs and tagging them is known (the above Patent Document 2). However, in the conventional techniques, for example, since an answer is not generated after appropriately narrowing down the information used for generating the answer, there is room for further improvement for providing an appropriate answer to a user's question in a short time. The present application has been made in view of the above, and an object thereof is to provide an appropriate answer to a user's question in a short time by generating an answer after narrowing down the information (index information) used for generating the answer based on categories and labels.

[0019] Next, two types of information processing of the information processing system 1 according to the embodiment will be described. As described above, the first information processing (Embodiment 1) is information processing for generating index information, and the second information processing (Embodiment 2) is information processing for generating an answer to a question from a user. By combining the two types of information processing, it becomes possible to provide an appropriate answer to the user's question. In the first information processing (Indexing Stage), the information processing apparatus 100 sorts the data acquired from the data source and performs indexing. In the second information processing (Querying Stage), the information processing apparatus 100 analyzes the question input by the user, extracts relevant information from the index, and generates a final response. By using this mechanism, the LAM index can efficiently manage a large amount of data, and the user can quickly obtain the necessary information. In particular, it is useful when a lot of information is scattered or when specific information needs to be quickly retrieved, greatly improving the efficiency of generating prompts for generative AI and enabling better results to be obtained.

[0020] (Embodiment 1: Information Processing for Generating Index Information) FIG. 3 is a diagram showing an example of the information processing of the information processing system 1 according to Embodiment 1. The information processing apparatus 100 acquires content (Content A) that is the target of a question from the user (step S101). The content according to the embodiment is an example of regulations, laws, rules, manuals, etc., but it may be content of any kind and does not need to be particularly limited. In the embodiment described later, an example where the content is the "Business Trip Expense Regulations" will be described, but the content may be a "Service Use Regulations" that defines the rights and obligations of the service provider and the matters to be observed by the user between the service provider and the user. Note that the content according to the embodiment may be converted into a document object such as PDF. That is, information acquired from different databases may be converted into a document object. Then, the document object thus converted may be converted into a more manageable node and indexed for efficient data access.

[0021] The information processing apparatus 100 divides the acquired content into a plurality of regions (region A1, region A2, region A3, ···) (step S102). For example, the information processing apparatus 100 divides the content by regions such as by chapter, by article, by sentence, by main idea, etc. Note that the size of the region to be divided (the range of region division) may be determined from the context of the content. That is, the information processing apparatus 100 may divide the content into regions according to, for example, the context of the content.

[0022] The information processing apparatus 100 generates summaries (summary A1, summary A2, ···) for each of the divided regions (step S103). Here, the generated summary is, for example, "The business trip expenses have a ceiling of 5000 yen per night". The summary generated in this way is included in the index information.

[0023] The information processing apparatus 100 generates index information (index P1, index P2, ···) by associating the generated summary with the category of the content that is the source of the summary and a label (which may be a label indicating the entire text) indicating the content that is the source of the summary (step S104). The category (category B1, B2, ···) is information indicating, for example, in which department the content that is the source of the summary was created. For example, the category may be "Legal Affairs Department", "Personnel Department", "General Affairs Department", etc. The label (label C1, label C2, ···) is a label indicating, for example, what kind of regulations, laws, rules, or manuals it is. For example, in the case of content related to regulations, the label may be "Business Trip Expense Regulations", etc. Also, at this time, the information processing apparatus 100 may generate index information by associating, for example, the summary, the category, the label, and a label (hereinafter, appropriately referred to as a "range label") indicating the range within the content that is the source of the summary. The range label (range label D1, range label D2, ···) is a label indicating, for example, which chapter, article, sentence, etc. in the content the summary is. For example, the range label may be "Article 5", etc. Therefore, taking a specific example, the information processing apparatus 100 generates index information by associating a summary such as "The upper limit of business trip expenses per night is 5,000 yen" with a category such as "Legal Affairs Department", a label such as "Business Trip Expense Regulations", and a range label such as "Article 5". The label may be information indicating the parent-child relationship with the content, and the range label may be information indicating the parent-child relationship with the content or the label.

[0024] As described above, in the first embodiment, the information processing apparatus 100 divides the content into regions, generates a summary for each divided region, assigns a category and a label to the generated summary, and uses it as index information. For example, the information processing apparatus 100 assigns a category related to a department such as "Legal Affairs Department", a label related to regulations such as "Business Trip Expense Regulations", and a label related to an article such as "Article 5" to a summary such as "The upper limit of business trip expenses per night is 5,000 yen".

[0025] (Embodiment 2: Information Processing for Generating Answers to Questions from Users) FIG. 4 is a diagram showing an example of information processing of the information processing system 1 according to Embodiment 2. The information processing apparatus 100 receives a question Q1 from the user U1 (step S201). In the following embodiments, an example will be described in which the information processing apparatus 100 obtains information regarding a question Q1 having the content such as "What is the upper limit of the accommodation fee in the company's accommodation regulations?" from the user U1. Note that categories and labels are associated with the question Q1 based on the input from the user U1. In Embodiment 2, when a user's question is received, appropriate information is retrieved from the index information according to the question, relevant nodes are extracted, necessary information is selected, and it becomes part of an inquiry to the generation AI. As a result, it is possible to efficiently extract highly relevant information from a large amount of data, and it is possible to generate a natural language response corresponding to the user's question.

[0026] The information processing apparatus 100 narrows down the index information based on the categories and labels associated with the question Q1 (step S202). For example, when the question Q1 received from the user U1 is associated with the category of "Legal Affairs Department" and the label of "Business Trip Expense Regulations", the information processing apparatus 100 narrows down the index information on the condition that the category is "Legal Affairs Department" and the label is "Business Trip Expense Regulations".

[0027] The information processing device 100 extracts index information relevant to the question Q1 of user U1 from the filtered index information (i.e., index information obtained by summarizing predetermined content for each area) (step S203). For example, the information processing device 100 may generate a first vector obtained by vectorizing the question Q1 of user U1 and a second vector obtained by vectorizing the summary associated with the index information, and determine whether the index information is relevant to the question Q1 of user U1 by comparing the first vector and the second vector. For example, the information processing device 100 may vectorize the question Q1 and the summary using a natural language processing model of BERT (Bidirectional Encoder Representations from Transformers). Further, the information processing device 100 may calculate the Euclidean distance based on the first vector and the second vector, and calculate a score indicating the degree of relevance to the question based on the calculated Euclidean distance. Further, the information processing device 100 may determine that the index information for which the calculated score is equal to or greater than a predetermined threshold is index information relevant to the question of user U1. The information processing device 100 may extract the index information determined to be relevant to the question of user U1.

[0028] The information processing device 100 generates an answer to the question Q1 of user U1 from the extracted index information. Specifically, the information processing device 100 generates a prompt for generating an answer to the question Q1 from the extracted index information, and inputs it to a text generation AI such as a GPT model to generate an answer to the question Q1 (step S204). To give a specific example, based on the extracted index information, the information processing device 100 may generate the following prompt from the question Q1 such as "What is the upper limit of the accommodation fee in the company's accommodation regulations?" and the summary such as "The upper limit of the business trip allowance is 5,000 yen per night".

[0029] #Instruction: You are a professional editor. Please create an answer to the question based on the following context information. Please create the answer in the language of the question. #Context information: The upper limit of accommodation expenses for business trips is 5,000 yen per night. #Question: What is the upper limit of accommodation expenses in the company's accommodation regulations?

[0030] By generating a prompt as described above and inputting it into the text generation AI, an answer such as "In the business trip expense regulations, the upper limit of accommodation expenses per night is 5,000 yen" is output. Then, the information processing device 100 provides the answer to the question Q1 to the user U1 (step S205). Specifically, the information processing device 100 transmits information regarding the answer to the question Q1 (information for displaying the answer) to the terminal device 10. Then, the terminal device 10 displays the answer provided by the information processing device 100.

[0031] In this way, in the second embodiment, when the information processing device 100 receives the question Q1 from the user U1, it narrows down the index information by categories and labels, extracts relevant index information, and generates a prompt for the text generation AI using the summary of the extracted index information. If an answer is generated without narrowing down the information (index information) used for answer generation, the answer may be generated based on information not related to the question Q1, and thus an appropriate answer to the user's question may not be generated. Also, if the information (index information) used for answer generation is not narrowed down, a large amount of index information needs to be referred to when generating the answer, so it takes a long time to generate the answer. Therefore, according to the information processing device hundred of this embodiment, by narrowing down the information (index information) used for answer generation based on the categories and labels associated with the question Q1, an appropriate answer to the user's question can be provided in a short time.

[0032] User U1 inputs question Q1, for example, to the bulletin board of a specific tool. For example, User U1 inputs question Q1 after specifying a predetermined category and label. When information processing device 100 receives question Q1 from User U1, it generates a prompt using a text generation AI. At this time, the content is divided into regions, and a library in which index information is generated for each region is used to generate the prompt. At this time, for example, information processing device 100 calculates a score indicating the degree of relevance for each index information and extracts index information with a high relevance score. Information processing device 100 may calculate the Euclidean distance based on the first vector and the second vector, and calculate the relevance score based on the calculated Euclidean distance. Then, for example, information processing device 100 generates a prompt for generating an answer using the summary text data of the extracted index information. Also, for example, when there is no index information with a high relevance score, information processing device 100 generates a prompt for generating an answer without using the index information. In this case, for example, an answer to question Q1 is generated without index information (such as a summary). In this case, for example, information processing device 100 generates an answer to question Q1 by inputting a prompt for generating an answer without using the index information to the text generation AI. Note that, for example, when there is no index information with a high relevance score, information processing device 100 may end the information processing without generating a prompt for generating an answer.

[0033] (Content) FIG. 5 is a diagram showing an example of content managed by a company according to an embodiment. A company has a plurality of departments, and a plurality of contents are created for each of the plurality of departments and managed on a server within the company. In the example shown in FIG. 5, a plurality of departments such as the "Legal Affairs Department", "Personnel Department", and "General Affairs Department" exist in the "XX Corporation". Also, the Legal Affairs Department creates a plurality of contents such as business trip expense regulations and employment regulations. In the present embodiment, each department (Legal Affairs Department, Personnel Department, General Affairs Department, etc.) within the company corresponds to a "category", the title of the content (regulations, rules, manuals, etc.) created by each department corresponds to a "label", and each area (Article 1, Article 2, Article 3, etc.) within the content corresponds to a range label.

[0034] (Chatbot screen) FIG. 6 is a diagram showing an example of the home screen 200 of a chatbot according to an embodiment. When the user starts the chatbot application using the terminal device 10, the home screen 200 shown in FIG. 6 is displayed on the terminal device 10. On the home screen 200 of the chatbot, an explanation about the chatbot and a filter button B1 are displayed. The filter button B1 is a button for displaying a window for inputting a question to the chatbot and categories and labels.

[0035] FIG. 7 is a diagram showing an example of the window 210 of a chatbot according to an embodiment. The window 210 is displayed on the terminal device 10 when the user selects (taps) the filter button B1. In the window 210, a category input area 211, a label input area 212, a message input area 213, a cancel button B2, and a send button B3 are displayed.

[0036] The category input area 211 is an area for the user to input the category of the question. For example, when the user selects (taps) the category input area 211, options for the category (such as the legal department, the personnel department, the general affairs department, etc.) may be displayed in a pull-down manner. When one of the multiple categories is selected by the user, the selected category may be displayed in the category input area 211. Note that the user may directly input the category into the category input area 211 using a character input application.

[0037] The label input area 212 is an area for the user to input the label of the question. For example, when the user selects (taps) the label input area 212, options for the label corresponding to the selected category (such as the business trip expense regulations, the employment regulations, etc.) may be displayed in a pull-down manner. When one of the multiple labels is selected by the user, the selected label may be displayed in the label input area 212. Note that the user may directly input the label into the label input area 212 using a character input application.

[0038] The message input area 213 is an area for the user to input the content of the question. For example, the user may select (tap) the message input area 213 and input the content of the question using a character input application. For example, the content of the question may be "What is the upper limit of the accommodation fee in the company's accommodation regulations?"

[0039] The cancel button B2 is a button for canceling the input to window 210. When the user selects (taps) the cancel button B2, window 210 will be closed. The send button B3 is a button for instructing to send the question from the user. When the user selects (taps) the send button B3, the terminal device 10 will associate the category input in the category input area 211, the label input in the label input area 212, and the content of the question input in the message input area 213, and send them to the information processing device 100.

[0040] In the window 210, a range label input area for inputting range labels (such as Article 1, Article 2, Article 3, etc.) may be provided. In this case, when the user selects (taps) the send button B3, the terminal device 10 may associate the category input in the category input area 211, the label input in the label input area 212, the range label input in the range label input area, and the content of the question input in the message input area 213, and send them to the information processing device 100.

[0041] FIG. 8 is a diagram showing an example of the internal inquiry form 220 according to the embodiment. The user may input the inquiry content from the internal inquiry form 220 shown in FIG. 8 in addition to the window 210 shown in FIG. 7. When the user accesses the information processing device 100 using the terminal device 10, the internal inquiry form 220 shown in FIG. 8 is displayed on the terminal device 10. The internal inquiry form 220 displays a question text input area 221, an affiliated company selection area 222, a department name input area 223, a location name input area 224, a category input area 225, a label input area 226, a contact information input area 227, a cancel button B4, and a send button B5.

[0042] The question text input area 221 is an area for the user to input the content of the question. For example, the user may select (tap) the question text input area 221 and use a character input application to input the content of the question. For example, the content of the question may be "What is the upper limit of the accommodation fee in the company's accommodation regulations?"

[0043] The affiliated company input area 222 is an area for the user to input the name of the company to which the user belongs. In the example shown in FIG. 8, options such as "XX Co., Ltd.", "YY Co., Ltd.", and "ZZ Co., Ltd." are displayed. These companies may be group companies or related companies. Also, in the example shown in FIG. 8, it is shown that "XX Co., Ltd." is selected.

[0044] The department name input area 223 is an area for inputting the department name to which the user belongs. For example, the user may select (tap) the department name input area 223 and input the department name using a character input application.

[0045] The site name input area 224 is an area for inputting the site name to which the user belongs. For example, when the user selects (taps) the site name input area 224, a plurality of site name options (such as ○○ Office, △△ Office, □□ Office, etc.) may be displayed in a pull-down menu. When one of the plurality of site names is selected by the user, the selected site name may be displayed in the site name input area 224. Note that the user may directly input the site name into the site name input area 224 using a character input application.

[0046] The category input area 225 is an area for the user to input the category of the question. For example, when the user selects (taps) the category input area 225, category options (such as Legal Affairs Department, Personnel Department, General Affairs Department, etc.) may be displayed in a pull-down menu. When one of the plurality of categories is selected by the user, the selected category may be displayed in the category input area 225. Note that the user may directly input the category into the category input area 225 using a character input application.

[0047] The label input area 226 is an area for the user to input the label of the question. For example, when the user selects (taps) the label input area 226, label options (such as Business Trip Expense Regulations, Employment Regulations, etc.) corresponding to the selected category may be displayed in a pull-down menu. When one of the plurality of labels is selected by the user, the selected label may be displayed in the label input area 226. Note that the user may directly input the label into the label input area 226 using a character input application.

[0048] The contact information input area 227 is an area for entering the user's email address as contact information. For example, the user may select (tap) the contact information input area 227 and use a character input app to enter the contact information (email address). Note that the reply from the chatbot may be sent to the email address entered in the contact information input area 227.

[0049] The cancel button B4 is a button for canceling the input to the internal inquiry form 220. When the user selects (taps) the cancel button B4, the internal inquiry form 220 will be closed. The send button B5 is a button for instructing to send the question from the user. When the user selects (taps) the send button B5, the terminal device 10 associates the question text (inquiry content) entered in the question text input area 221, the affiliated company entered in the affiliated company selection area 222, the department name entered in the department name input area 223, the location name entered in the location name input area 224, the category entered in the category input area 225, and the label entered in the label input area 226, and sends them to the information processing device 100.

[0050] Note that in the internal inquiry form 220, a range label input area for entering range labels (Article 1, Article 2, Article 3, etc.) may be provided. In this case, when the user selects (taps) the send button B5, the terminal device 10 associates the question text (inquiry content) entered in the question text input area 221, the affiliated company entered in the affiliated company selection area 222, the department name entered in the department name input area 223, the location name entered in the location name input area 224, the category entered in the category input area 225, the label entered in the label input area 226, and the range label entered in the range label input area, and may send them to the information processing device 100.

[0051] Also, in the user's inquiry scenario, the input of categories and labels may be optional, and the category input area 225 and the label input area 226 may be omitted from the internal inquiry form 220. For example, an internal inquiry form 220 may be provided for each inquiry content, and categories and labels may be associated with the internal inquiry form 220. Accordingly, categories and labels can be automatically determined according to the used internal inquiry form 220, and the input load on the user for the internal inquiry form 220 can be reduced.

[0052] Conventionally, as a result of each department in the company freely creating various documents, it has been difficult to know where and what kind of information exists, and inquiries for inquiries (for example, inquiries such as what to inquire about, where to send the inquiry application, and whether an application is necessary in the first place) have occurred, resulting in inefficient work. On the other hand, according to the present embodiment, by indexing internal documents in association with categories and labels and causing the generation AI to generate an answer to a question from a user based on the index information, an appropriate answer to the question from the user can be generated, and work efficiency can be improved.

[0053] (Variations of processing) In the above embodiment, even when it is determined that there is index information relevant to a question, the index information accessible by the user may be restricted. For example, the information processing apparatus 100 may extract relevant index information from the index information accessible by the user and generate a prompt.

[0054] In the above embodiment, the information processing apparatus 100 may perform processing for optimizing the range (size of the area) of area division according to the answer accuracy. In the example of the above embodiment, when the user evaluates an answer that is an output result such as "In the business trip expense regulations, the upper limit of the accommodation expense per night is 5,000 yen" for a question such as "What is the upper limit of the accommodation expense in the company's accommodation regulations?", the user's evaluation may be received as the evaluation of the answer. In this way, the information processing apparatus 100 receives evaluations from (a plurality of) users. When the evaluation is low, a plurality of index information with the range of area division changed is set, and when changing to any one of the plurality of index information results in a higher evaluation, it may be changed to the index information when the evaluation is the highest.

[0055] In the above embodiment, when there is a predetermined exemplary answer (model answer) such as Q&A (or FAQ), the information processing apparatus 100 may generate a prompt using the model answer. For example, the information processing apparatus 100 may generate a prompt instructing to output the model answer together with the answer according to the above embodiment. For example, the information processing apparatus 100 may generate a prompt for outputting a model answer such as "The model answer is as follows: 〇×〇×···" together with an answer such as "In the business trip expense regulations, the upper limit of the accommodation expense per night is 5,000 yen".

[0056] In the above embodiment, the information processing apparatus 100 may perform linking for the generated answer. For example, the information processing apparatus 100 may perform linking to the content corresponding to the label associated with the index information used for generating the answer.

[0057] In the above-described embodiment, the case where the label is "Business Trip Expense Regulations", "Employment Regulations", "Childcare / Nursing Care Leave Regulations", etc. has been described as an example, but it is not particularly limited to these examples. The label according to the embodiment may be a label indicating a specific regulation, law, specific rule, manual, etc., but it is not particularly limited. Also, in the above-described embodiment, the case where the range label is "Article 5" etc. has been described as an example, but it is not particularly limited to this example. The range label according to the embodiment may be a label indicating a specific chapter, article, sentence, etc., but it is not particularly limited.

[0058] [4. Configuration of Terminal Device] Next, with reference to FIG. 9, the configuration of the terminal device 10 according to the embodiment will be described. FIG. 9 is a diagram showing a configuration example of the terminal device 10 according to the embodiment. As shown in FIG. 9, the terminal device 10 includes a communication unit 11, an input unit 12, an output unit 13, and a control unit 14.

[0059] (Communication Unit 11) The communication unit 11 is realized by, for example, a NIC (Network Interface Card) or the like. Then, the communication unit 11 is connected to a predetermined network N by wire or wirelessly, and information is transmitted and received between the communication unit 11 and an information processing device 100 or the like via the predetermined network N.

[0060] (Input Unit 12) The input unit 12 receives various operations from the user. In FIG. 4 described above, various operations from the user U1 are received. For example, the input unit 12 may receive various operations from the user via the display surface by the touch panel function. Also, the input unit 12 may receive various operations from buttons provided on the terminal device 10 or from a keyboard or mouse connected to the terminal device 10. For example, the input unit 12 receives an operation for inputting a question.

[0061] (Output Unit 13) The output unit 13 is a display screen of a tablet terminal or the like realized by, for example, a liquid crystal display or an organic EL (Electro-Luminescence) display, and is a display device for displaying various information. For example, the output unit 13 displays the information transmitted from the information processing apparatus 100. For example, the output unit 13 displays an answer to a user's question generated by the information processing apparatus 100.

[0062] (Control unit 14) The control unit 14 is, for example, a controller, and is realized by various programs stored in the storage device inside the terminal device 10 being executed with the RAM (Random Access Memory) as a work area by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like. For example, these various programs include the programs of applications installed in the terminal device 10. For example, these various programs include the programs of applications for displaying the information (such as an answer to a user's question) transmitted from the information processing apparatus 100. Further, the control unit 14 is realized by, for example, an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0063] As shown in FIG. 9, the control unit 14 has a receiving unit 141 and a transmitting unit 142, and realizes or executes the operations of information processing described below.

[0064] (Receiving unit 141) The receiving unit 141 receives, for example, the information transmitted from the information processing apparatus 100. For example, the receiving unit 141 receives the information regarding the answer to the user's question. For example, the receiving unit 141 receives the information for displaying the answer to the user's question.

[0065] (Transmitting unit 142) The transmission unit 142 transmits, for example, operation information performed by the user. For example, the transmission unit 142 transmits information regarding a question received from the user (for example, input by the user, specified by the user, selected by the user).

[0066] [5. Configuration of Information Processing Apparatus] Next, with reference to FIG. 10, the configuration of the information processing apparatus 100 according to the embodiment will be described. FIG. 10 is a diagram showing a configuration example of the information processing apparatus 100 according to the embodiment. As shown in FIG. 10, the information processing apparatus 100 includes a communication unit 110, a storage unit 120, and a control unit 130. Note that the information processing apparatus 100 may include an input unit (for example, a keyboard or a mouse) that receives various operations from the administrator of the information processing apparatus 100, and a display unit (for example, a liquid crystal display) that displays various information.

[0067] (Communication Unit 110) The communication unit 110 is realized by, for example, a NIC or the like. Then, the communication unit 110 is connected to the network N by wire or wirelessly, and performs information transmission and reception with the terminal device 10 or the like via the network N.

[0068] (Storage Unit 120) The storage unit 120 is realized by, for example, a semiconductor memory element such as a RAM or a flash memory, or a storage device such as a hard disk or an optical disk. As shown in FIG. 10, the storage unit 120 includes an index information storage unit 121 and an answer evaluation information storage unit 122.

[0069] The index information storage unit 121 stores index information. Here, FIG. 11 shows an example of the index information storage unit 121 according to the embodiment. The information stored in the index information storage unit 121 is used, for example, for generating an answer. As shown in FIG. 11, the index information storage unit 121 has items such as "index information ID", "content ID", "category", "label", "range label", and "summary".

[0070] "Index information ID" indicates identification information for identifying index information. "Content ID" indicates identification information for identifying content. "Category" indicates a category (a category indicating the type of content). "Label" indicates a label (a label indicating the content of the content). "Range label" indicates a range label (a label indicating a specific range within the content). "Summary" indicates a summary (a summary corresponding to a specific range within the content).

[0071] The answer evaluation information storage unit 122 stores evaluation information for an answer. Here, FIG. 12 shows an example of the answer evaluation information storage unit 122 according to the embodiment. The information stored in the answer evaluation information storage unit 122 is used, for example, for optimizing the range of area division. As shown in FIG. 12, the answer evaluation information storage unit 122 has items such as "Index information ID" and "Answer evaluation information".

[0072] "Index information ID" indicates identification information for identifying index information. "Answer evaluation information" indicates evaluation information for an answer. In the example shown in FIG. 12, an example is shown in which conceptual information such as "Answer evaluation information #1" and "Answer evaluation information #2" is stored in "Answer evaluation information", but actually, a score indicating a comprehensive evaluation by (multiple) users, etc., is stored.

[0073] (Control unit 130) The control unit 130 is a controller and is realized, for example, by various programs stored in a storage device inside the information processing apparatus 100 being executed with the RAM as a work area by a CPU, MPU, etc. Also, the control unit 130 is realized by an integrated circuit such as an ASIC or FPGA.

[0074] As shown in FIG. 10, the control unit 130 includes an acquisition unit 131, a division unit 132, a first generation unit 133, a second generation unit 134, a reception unit 135, a narrowing-down unit 136, an extraction unit 137, a third generation unit 138, a provision unit 139, and a vectorization unit 140, and realizes or executes the operations of information processing described below. Note that the internal configuration of the control unit 130 is not limited to the configuration shown in FIG. 10, and any other configuration may be used as long as it can perform the information processing described later.

[0075] (Acquisition Unit 131) The acquisition unit 131 acquires various types of information from the storage unit 120. Further, the acquisition unit 131 stores the acquired various types of information in the storage unit 120.

[0076] The acquisition unit 131 acquires various types of information from an external information processing device. The acquisition unit 131 acquires various types of information from another information processing device such as the terminal device 10.

[0077] The acquisition unit 131 acquires, for example, content that is the subject of a question from the user. For example, the acquisition unit 131 acquires text data of the content (for example, the entire text).

[0078] (Division Unit 132) The division unit 132 divides, for example, the content acquired by the acquisition unit 131. For example, the division unit 132 divides the content acquired by the acquisition unit 131 into regions according to the context (for example, regions for each chapter, each article, each sentence, etc.). The division unit 132 divides the content into one or more (plural) regions in this way, for example.

[0079] (First Generation Unit 133) The first generation unit 133 generates a summary for each region divided by the division unit 132, for example. For example, when the content is divided into regions for each chapter, the first generation unit 133 generates a summary for each chapter.

[0080] (Second Generation Unit 134) The second generation unit 134 generates, for example, index information that associates the summary generated by the first generation unit 133 with the category of the content that is the source of the summary and the label corresponding to the content that is the source of the summary. Further, the second generation unit 134 generates, for example, index information that associates the summary generated by the first generation unit 133 with the category of the content that is the source of the summary, the label corresponding to the content that is the source of the summary, and the range label corresponding to the range (the range corresponding to the summary generated by the first generation unit 133) within the content that is the source of the summary.

[0081] The index information generated in this way is used to generate a prompt for generating an answer when receiving a question from a user. Further, the index information generated in this way is used to extract index information relevant to the question when receiving a question from a user.

[0082] (Reception unit 135) The reception unit 135 receives, for example, a question associated with a category and a label from the user. Further, the reception unit 135 receives, for example, a question associated with a category, a label, and a range label from the user.

[0083] (Filtering unit 136) The narrowing-down unit 136 narrows down index information based on, for example, the categories and labels associated with the questions received by the reception unit 135. Also, the extraction unit 137 narrows down index information based on, for example, the categories, labels, and range labels associated with the questions received by the reception unit 135. For example, when the question received by the reception unit 135 is associated with the category of "Legal Affairs Department" and the label of "Business Trip Expense Regulations", the narrowing-down unit 136 narrows down the index information on the condition that the category is "Legal Affairs Department" and the label is "Business Trip Expense Regulations". Also, for example, when the question received by the reception unit 135 is associated with the category of "Legal Affairs Department", the label of "Business Trip Expense Regulations", and the range label of "Article 5", the narrowing-down unit 136 narrows down the index information on the condition that the category is "Legal Affairs Department", the label is "Business Trip Expense Regulations", and the range label is "Article 5".

[0084] (Extraction unit 137) The extraction unit 137 extracts index information relevant to the question from the index information narrowed down by the narrowing-down unit 136, for example. At this time, the extraction unit 137 calculates, for example, a score indicating the degree of relevance to the question for each index information, and extracts index information whose score is equal to or higher than a predetermined threshold value. For example, the extraction unit 137 may generate a first vector obtained by vectorizing the user's question and a second vector obtained by vectorizing the summary associated with the index information, and determine whether the index information is relevant to the user's question by comparing the first vector and the second vector. For example, the extraction unit 137 may vectorize the question and the summary using the natural language processing model of BERT. Also, the extraction unit 137 calculates the Euclidean distance based on the first vector and the second vector, calculates a score indicating the degree of relevance to the question based on the calculated Euclidean distance, and extracts index information whose score is equal to or higher than a predetermined threshold value. Note that the extraction unit 137 does not extract index information, for example, when there is no index information whose score is equal to or higher than a predetermined threshold value (that is, when there is no index information with high relevance).

[0085] (Third generation unit 138) The third generation unit 138 generates an answer to the question received from the user from the index information extracted by the extraction unit 137, for example. For example, the third generation unit 138 generates a prompt for generating an answer to the question received from the user, and sets the answer output by inputting the prompt to the text generation AI as the answer to the question received from the user.

[0086] If the index information is not extracted by the extraction unit 137, for example, the third generation unit 138 generates an answer to the question received from the user without using the index information. For example, the third generation unit 138 generates a prompt for generating an answer to the question received from the user without using the index information, and sets the answer output by inputting the prompt to the text generation AI as the answer to the question received from the user.

[0087] (Provision unit 139) The provision unit 139 provides the information regarding the answer generated by the third generation unit 138 to the user who asked the question, for example. For example, the provision unit 139 transmits the content indicating the answer generated by the third generation unit 138 to the terminal device 10. Also, for example, the provision unit 139 transmits the information for displaying the content indicating the answer generated by the third generation unit 138 to the terminal device 10.

[0088] [6. Information processing flow] Next, with reference to FIGS. 13 and 14, the information processing procedure by the information processing system 1 according to the embodiment will be described. FIGS. 13 and 14 are flowcharts showing the information processing procedure by the information processing system 1 according to the embodiment (Embodiment 1 and Embodiment 2).

[0089] As shown in FIG. 13 (Embodiment 1), the information processing apparatus 100 acquires the content to be learned (step S301).

[0090] The information processing apparatus 100 divides the acquired content into areas according to the context (step S302).

[0091] The information processing apparatus 100 generates a summary for each divided area (step S303).

[0092] The information processing apparatus 100 generates index information corresponding to the generated summary (step S304).

[0093] As shown in FIG. 14 (Embodiment 2), the information processing apparatus 100 receives a question from the user (step S401).

[0094] The information processing apparatus 100 narrows down the index information based on the category and label associated with the received question (step S402).

[0095] The information processing apparatus 100 determines whether there is index information relevant to the received question (step S403).

[0096] When the information processing apparatus 100 determines that there is index information relevant to the received question (step S403; YES), it extracts the index information highly relevant to the received question (step S404). Then, the information processing apparatus 100 generates an answer to the question based on the extracted index information (step S405).

[0097] On the other hand, when the information processing apparatus 100 determines that there is no index information relevant to the received question (step S403; NO), it generates an answer to the question without extracting the index information (step S406).

[0098] [7. Effects] As described above, the information processing apparatus 100 according to the embodiment includes an acquisition unit 131, a division unit 132, a first generation unit 133, a second generation unit 134, a reception unit 135, a narrowing-down unit 136, an extraction unit 137, and a third generation unit 138. The acquisition unit 131 acquires content that is the target of a question from the user. The division unit 132 divides the content acquired by the acquisition unit 131 into a plurality of regions according to the context. The first generation unit 133 generates a summary for each of the plurality of regions divided by the division unit 132. The second generation unit 134 generates index information in which the summary generated by the first generation unit 133 is associated with the category of the content that is the source of the summary and the label indicating the content that is the source of the summary. The reception unit 135 receives a question associated with the category and the label from the user. The narrowing-down unit 136 narrows down the index information based on the category and the label associated with the question received by the reception unit 135. The extraction unit 137 extracts index information relevant to the question from the index information narrowed down by the narrowing-down unit 136. The third generation unit 138 generates an answer to the question based on the index information extracted by the extraction unit 137.

[0099] Thereby, the information processing apparatus 100 according to the embodiment can appropriately acquire information (index information) used for generating an answer, and thus can provide an appropriate answer to the user's question in a short time.

[0100] Also, the vectorization unit 140 generates a first vector obtained by vectorizing the question and a second vector obtained by vectorizing the summary. The extraction unit 137 extracts index information relevant to the question by comparing the first vector and the second vector.

[0101] Thereby, the information processing apparatus 100 according to the embodiment can extract, with high accuracy, index information relevant to, for example, a question from the user.

[0102] In addition, the third generation unit 138 generates a prompt for generating an answer to the question based on the extracted index information, and generates the answer output by inputting the prompt into the sentence generation AI (trained model) as the answer to the question from the user.

[0103] Thereby, the information processing apparatus 100 according to the embodiment can generate, for example, an appropriate answer to a question from the user.

[0104] In addition, the second generation unit 134 generates index information associating a summary, a category, a label, and a range label indicating a range in the content serving as the summary source.

[0105] Thereby, the information processing apparatus 100 according to the embodiment can, for example, organize data and structure it so as to be easily searchable.

[0106] In addition, the second generation unit 134 generates index information in which summaries, categories, and labels for each region determined based on the evaluation of the answer are associated.

[0107] Thereby, the information processing apparatus 100 according to the embodiment can, for example, optimize the range of region division according to the answer accuracy.

[0108] In addition, the content is a regulation.

[0109] Thereby, the information processing apparatus 100 according to the embodiment can generate, for example, an appropriate answer regarding the regulation.

[0110] 〔8. Hardware Configuration〕 Further, the information processing apparatus 100 according to the above-described embodiments is realized by, for example, a computer 1000 configured as shown in FIG. 15. FIG. 15 is a hardware configuration diagram showing an example of a computer that realizes the functions of the information processing apparatus 100. The computer 1000 includes a CPU 1100, a RAM 1200, a ROM 1300, an HDD 1400, a communication interface (I / F) 1500, an input / output interface (I / F) 1600, and a media interface (I / F) 1700.

[0111] The CPU 1100 operates based on programs stored in the ROM 1300 or the HDD 1400 and controls each part. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 is started up, programs dependent on the hardware of the computer 1000, and the like.

[0112] The HDD 1400 stores programs executed by the CPU 1100, data used by such programs, and the like. The communication interface 1500 acquires data from other devices via a predetermined communication network and sends it to the CPU 1100, and transmits data generated by the CPU 1100 to other devices via the predetermined communication network.

[0113] The CPU 1100 controls output devices such as a display and a printer, and input devices such as a keyboard and a mouse via the input / output interface 1600. The CPU 1100 acquires data from the input devices via the input / output interface 1600. Further, the CPU 1100 outputs the generated data to the output devices via the input / output interface 1600.

[0114] The media interface 1700 reads a program or data stored in the recording medium 1800 and provides it to the CPU 1100 via the RAM 1200. The CPU 1100 loads such a program from the recording medium 1800 onto the RAM 1200 via the media interface 1700 and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc), a PD (Phase change rewritable Disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory.

[0115] For example, when the computer 1000 functions as the information processing apparatus 100 according to the embodiment, the CPU 1100 of the computer 1000 realizes the functions of the control unit 130 by executing the program loaded onto the RAM 1200. The CPU 1100 of the computer 1000 reads and executes these programs from the recording medium 1800. As another example, these programs may be acquired from another device via a predetermined communication network.

[0116] [9. Others] Also, among the various processes described in the above embodiment, all or part of the processes described as being automatically performed can be manually performed, or all or part of the processes described as being manually performed can be automatically performed by a known method. In addition, the processing procedures, specific names, and information including various data and parameters shown in the above document and drawings can be arbitrarily changed unless otherwise specified. For example, the various information shown in each figure is not limited to the illustrated information.

[0117] In addition, each component of each illustrated device is functionally conceptual and does not necessarily have to be physically configured as shown in the drawings. That is, the specific form of the distribution and integration of each device is not limited to that shown in the drawings, and all or part of it can be functionally or physically distributed and integrated in any unit according to various loads, usage situations, etc.

[0118] Also, the above-described embodiments can be appropriately combined within a range that does not conflict with the processing content.

[0119] As described above, some of the embodiments of the present application have been described in detail with reference to the drawings. However, these are examples, and the present invention can be implemented in other forms with various modifications and improvements based on the knowledge of those skilled in the art, including the aspects described in the column of the disclosure of the invention.

[0120] Also, the above-described "section, module, unit" can be read as "means", "circuit", etc. For example, the acquisition section can be read as an acquisition means or an acquisition circuit.

Explanation of Reference Numerals

[0121] 1 Information processing system 10 Terminal device 11 Communication section 12 Input section 13 Output section 14 Control section 100 Information processing device 110 Communication section 120 Storage section 121 Index information storage section 122 Answer evaluation information storage section 130 Control section 131 Acquisition section 132 Division section 133 First generation section 134 Second generation section 135 Reception section 136 Narrowing-down section 137 Extraction section 138 Third generation section 139 Providing unit 140 Vectorization unit 141 Receiving unit 142 Transmitting unit N Network

Claims

1. An acquisition unit that acquires content that is the subject of a question from a user; A division unit that divides the content into a plurality of regions according to the context; A first generation unit that generates a summary for each of the plurality of regions; A second generation unit that generates index information associating the summary, the category of the content that is the source of the summary, and a label indicating the content that is the source of the summary; A reception unit that receives a question associated with the category and the label from the user; A narrowing-down unit that narrows down the index information based on the category and the label associated with the question; A vectorization unit that generates a first vector obtained by vectorizing the question and a second vector obtained by vectorizing the summary associated with the narrowed-down index information; An extraction unit that extracts index information relevant to the question by comparing the first vector and the second vector; A third generation unit that generates a prompt for generating an answer to the question based on the summary associated with the extracted index information, and inputs the prompt into a learned model to generate an answer to the question; Comprising: The second generation unit receives an evaluation of the answer from the user, and when the evaluation is low, changes the range of region division in the index information; An information processing apparatus.

2. The second generation unit: Generates the index information associating the summary, the category, the label, and a range label indicating a range in the content that is the source of the summary; The information processing apparatus according to Claim 1.

3. The content is a regulation; The information processing apparatus according to Claim 1.

4. When the evaluation is low, the second generation unit sets a plurality of pieces of index information with the range of region division changed, and when changing to any one of the plurality of pieces of index information results in a higher evaluation, changes to the index information when the evaluation is the highest; The information processing apparatus according to Claim 1.

5. An information processing apparatus: Acquires content that is the subject of a question from a user; Divides the content into a plurality of regions according to the context; Generates a summary for each of the plurality of regions; Generate index information that associates the summary, the category of the content from which the summary is derived, and the label indicating the content from which the summary is derived. Accept questions associated with the category and the label from the user. Narrow down the index information based on the category and the label associated with the question. Generate a first vector obtained by vectorizing the question and a second vector obtained by vectorizing the summary associated with the narrowed-down index information. Extract the index information relevant to the question by comparing the first vector and the second vector. Generate a prompt for generating an answer to the question based on the summary associated with the extracted index information, and generate an answer to the question by inputting the prompt into a pre-trained model. Accept an evaluation of the answer from the user, and if the evaluation is low, change the range of region division in the index information. Information processing method.

6. In an information processing apparatus, Cause the apparatus to obtain content that is the target of a question from a user. Cause the content to be divided into a plurality of regions according to the context. Cause a summary to be generated for each of the plurality of regions. Cause index information to be generated that associates the summary, the category of the content from which the summary is derived, and the label indicating the content from which the summary is derived. Cause the apparatus to accept questions associated with the category and the label from the user. Cause the index information to be narrowed down based on the category and the label associated with the question. Cause a first vector obtained by vectorizing the question and a second vector obtained by vectorizing the summary associated with the narrowed-down index information to be generated. Cause the index information relevant to the question to be extracted by comparing the first vector and the second vector. Cause a prompt for generating an answer to the question to be generated based on the summary associated with the extracted index information, and cause an answer to the question to be generated by inputting the prompt into a pre-trained model. Cause the apparatus to accept an evaluation of the answer from the user, and if the evaluation is low, cause the range of region division in the index information to be changed. Program.

Images