Information processing systems, information processing methods, and programs

Abstract

This system provides a mechanism for displaying related information near the relevant parts of a response text generated by a generation AI. [Solution] The information processing system comprises: a receiving means for receiving a question from a user; a first acquisition means for acquiring an answer generated by a generation AI based on the question; a second acquisition means for acquiring related information based on at least one of the question and the answer; and a first display means for displaying the related information near a portion of the answer associated with the related information.

Description

Technical Field

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

Background Art

[0002] As a document disclosing the background art of this technical field, there is Patent Document 1. In this Patent Document 1, it is described that "a mediation device that mediates a conversation between a user and an artificial intelligence transmits one or more topics to the artificial intelligence as topic information. Based on the topic information, a product to be targeted for appeal related to the designated conversation is selected from among one or more products on the condition that the artificial intelligence has notified that the conversation with the user is a designated conversation including any one of the one or more topics. An insertion process for inserting appeal information for appealing the selected product to be targeted for appeal to the user into the conversation is executed."

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the above Patent Document 1, it is described that appeal information for appealing the selected product to the user is inserted into the conversation. However, no particular ingenuity has been put into the method of inserting (or rather, the display method) the appeal information. The present invention has been made in view of such circumstances, and provides a mechanism for displaying related information in the vicinity of a part related to the related information when displaying a response sentence generated by a generative AI.

Means for Solving the Problems

[0005] In order to solve the above problems, for example, the configuration described in the claims is adopted. The present invention includes multiple means for solving the above-mentioned problems, but one example is an information processing system comprising: a receiving means for receiving a question from a user; a first acquisition means for acquiring an answer generated by a generation AI based on the question; a second acquisition means for acquiring related information based on at least one of the question and the answer; and a first display means for displaying the related information near a portion of the answer related to the related information. [Effects of the Invention]

[0006] According to the present invention, when displaying a response text generated by a generation AI, related information can be displayed near the part of the text that is related to that information. Other issues, configurations, and effects not mentioned above will be clarified by the following description of the embodiments. [Brief explanation of the drawing]

[0007] [Figure 1] Figure 1 shows an example of the configuration of the information processing system 100. [Figure 2] Figure 2 shows an example of the configuration of the question answering server 101. [Figure 3] Figure 3 shows an example of the configuration of the LLM server 102. [Figure 4] Figure 4 shows an example of the configuration of user terminal 103. [Figure 5] Figure 5 shows an example of advertising information 500. [Figure 6] Figure 6 shows an example of 600 pieces of advertising-related information. [Figure 7] Figure 7 shows an example of the question answering process 700. [Figure 8] Figure 8 shows an example of the advertisement identification process 800. [Figure 9] Figure 9 shows an example of the chat screen 900. [Figure 10] Figure 10 shows an example of the chat screen 1000. [Figure 11] Figure 11 shows an example of the chat screen 1100. [Figure 12] FIG. 12 shows an example of a chat screen 1200. [Figure 13] FIG. 13 shows an example of a chat screen 1300.

BEST MODE FOR CARRYING OUT THE INVENTION

[0008] 1. Embodiment Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0009] 1-1. Overview First, an overview of the embodiments of the present invention will be described. Conventionally, in information acquisition using generative AI, unlike conventional searches, search results for a query are not summarized and listed, but rather content (such as text) is generated and presented. Therefore, for the user, the opportunity to click on relevant information and navigate to other pages to make their own selection of information while being exposed to diverse information is reduced. In particular, it becomes difficult to establish a click-based advertising business.

[0010] In view of such circumstances, in this embodiment, relevant information (for example, advertisements) is acquired based on the response text generated by generative AI, and the acquired relevant information is displayed in the vicinity of the response text (for example, see FIG. 10).

[0011] Furthermore, when the user additionally inputs a query, it is estimated whether the additional query is related to the above-mentioned relevant information. And when it is estimated that the additional query is related to the above-mentioned relevant information, it is regarded as if the above-mentioned relevant information has been clicked.

[0012] <* Also, when displaying the response text generated based on the additional query, more detailed information than the above-mentioned relevant information is displayed as additional relevant information (for example, see FIG. 11).

[0013] 1-2. Configuration Next, the configuration of this embodiment will be described. FIG. 1 shows an example of the configuration of an information processing system 100 according to this embodiment. The information processing system 100 includes a question-and-answer server 101, an LLM server 102, and a plurality of user terminals 103. Each device constituting this system is connected via a wired or wireless network and can transmit and receive information to and from each other.

[0014] Among the devices constituting this system, the question-and-answer server 101 is a server that provides a question-and-answer service. The LLM server 102 is a server that receives a query from the question-and-answer server 101 and outputs an answer corresponding to the received query. The plurality of user terminals 103 are terminal devices used by users of the question-and-answer service, respectively.

[0015] Each device constituting the information processing system 100 includes a processor that executes an operating system, applications, programs, etc., a main memory device such as a RAM (Random Access Memory), an auxiliary storage device such as an IC card, a hard disk drive, an SSD (Solid State Drive), a flash memory, etc., a communication control unit such as a network card, a wireless communication module, a mobile communication module, etc., an input device such as a touch panel, a keyboard, a mouse, an input by voice input, motion detection by imaging of a camera unit, etc., and an output device such as a monitor or a display. Note that the output device may be a device or a terminal that transmits information to be output to an external monitor, display, printer, device, etc.

[0016] Various programs and applications (modules) are stored in the main memory device, and by the processor executing these programs and applications, each functional element of the entire system is realized. Note that each module may be implemented in hardware by integration or the like. Also, each module may be an independent program or application, or may be implemented in the form of some sub-programs or functions in one integrated program or application.

[0017] In this specification, each module is described as the entity (subject) that performs the processing; however, in reality, the processor that processes various programs and applications (modules) executes the processing.

[0018] The auxiliary storage device stores various databases (DBs). A "database" is a functional element (storage unit) that stores a data set so that it can handle any data manipulation (e.g., extraction, addition, deletion, overwriting, etc.) from the processor or an external computer. The implementation method of a database is not limited; for example, it may be a database management system, spreadsheet software, or text files such as XML or JSON.

[0019] 1-2-1. Question Answering Server 101 Figure 2 shows an example of the configuration of the question answering server 101. The question answering server 101 consists of, for example, one or more servers located on the cloud. The server's main memory 201 stores programs and applications such as the reception module 210, the first acquisition module 211, the splitting module 212, the second acquisition module 213, the first display module 214, the judgment module 215, the third acquisition module 216, and the second display module 217. The processor 203 executes these programs and applications to realize each functional element of the question answering server 101. Each module will be described below.

[0020] The reception module 210 receives input from the user in the form of a question (in other words, a query).

[0021] The first acquisition module 211 acquires the answer text generated by the generation AI based on the user's question text. Specifically, the generation AI referred to here is the generation AI module 310, which will be described later.

[0022] The splitting module 212 divides the response text generated by the generation AI into multiple parts (in other words, chunks). Specifically, this module divides the response text into multiple parts on a paragraph basis. Furthermore, chunking is not limited to paragraph units; as will be explained later, chunking may be performed at other units as well.

[0023] The second acquisition module 213 acquires relevant information based on the response text generated by the generation AI. In doing so, this module acquires relevant information based on each of the multiple parts divided by the division module 212. Specifically, the relevant information acquired here is advertisements. Furthermore, the related information obtained may include information other than advertisements, as will be explained later.

[0024] The first display module 214 displays the response text generated by the generation AI and related information acquired by the second acquisition module 213. In doing so, the module displays the related information near the part of the response text that is related to that information (see, for example, Figure 10). Therefore, the user can easily grasp the correspondence between the related information and the part of the response text.

[0025] In addition, the module displays related information separately from the answer text (see, for example, Figure 10). Therefore, users can easily distinguish between related information and the answer text.

[0026] Furthermore, the module can switch the displayed related information from among the related information obtained for each of two or more parts, in accordance with the scrolling of the displayed response text (see, for example, Figures 12 and 13). This display method is particularly effective on user terminals 103 with limited display space (e.g., smartphones).

[0027] Next, after displaying the answer and related information, the determination module 215 estimates whether another question (in other words, a query) received from the user is related to the relevant information. If the question is related to the relevant information, the module determines that the user has viewed that information. The module then increments the number of times the relevant information has been viewed.

[0028] In this embodiment, the advertising fee is determined according to the number of times the relevant information is viewed, and the advertiser is charged an advertising fee based on the number of times the relevant information is viewed.

[0029] The third acquisition module 216 retrieves detailed information about related information if, after displaying the answer and related information, the user submits another question (in other words, a query), and another answer is obtained based on that question. However, detailed information is only obtained if the other question submitted by the user is related to the related information. In that case, it is presumed that the user is interested in that related information, and therefore the detailed information about that related information is displayed.

[0030] The second display module 217 displays the detailed information obtained by the third acquisition module 216 near the other answer sentence mentioned above (see, for example, Figure 11). In doing so, this module does not display any related information other than the detailed information near the other answer sentence. This prevents the user from being bothered by related information that is not of interest to them.

[0031] Next, we will describe the auxiliary storage device 202. The auxiliary storage device 202 stores advertising information 500, advertising-related information 600, and other information. The following describes each type of information.

[0032] Figure 5 shows an example of advertising information 500. The advertisement information 500 includes items such as advertisement ID 501, logo image ID 502, company name 503, display URL 504, advertisement headline 505, description 506, and site links 507, with values ​​similar to those exemplified by the sample values ​​entered for each item. The following describes each item.

[0033] Ad ID 501 is the identifier for the advertisement. Logo image ID 502 is the identification information for the company's logo image. Company name 503 is the name of the company. The displayed URL 504 is the URL of the landing page. Advertisement headline 505 is the headline of an advertisement. Description 506 is the advertisement description. A sitelink 507 is a set of links designed to direct a user to a specific page.

[0034] Next, Figure 6 shows an example of advertising-related information 600. The ad-related information 600 includes items such as ad ID 601, keyword 602, match type 603, maximum bid 604, ad quality 605, ad rank 606, and prompt count 607, with values ​​like those exemplified by the sample values ​​entered for each item. The following describes each item.

[0035] Ad ID 601 is the identifier for the advertisement. Keyword 602 is a keyword related to advertising. Match type 603 specifies the range of words and phrases covered by a keyword. There are three types of match type 603: exact match, phrase match, and intent match. When you select exact match, your ad will be displayed when a word or phrase with the exact same meaning as your keyword is found. When you select phrase match, your ad will be displayed when a word or phrase that "contains" the exact same meaning as your keyword is found. When you select intent match, your ad will be displayed when a word or phrase that is "related" to your keyword is found.

[0036] The maximum bid price of 604 is the amount that can be paid per ad view. Ad Quality 605 is a measure of ad quality. This Ad Quality 605 is determined by the question answering service provider based on three aspects: estimated click-through rate, ad relevance, and landing page usability. Ad rank 606 is the rank of the ad. This ad rank 606 is calculated by multiplying the maximum bid price of 604 by the ad quality of 605. The ad's ranking is determined according to this ad rank 606. The prompt count of 607 represents the number of times the advertisement was viewed by the user. This prompt count of 607 is incremented by the judgment module 215 described above.

[0037] 1-2-2.LLM Server 102 Figure 3 shows an example of the configuration of the LLM server 102. The LLM server 102 consists of, for example, one or more servers located on the cloud. The main memory 301 of this server stores programs and applications such as the generation AI module 310. The processor 303 executes these programs and applications to realize each functional element of the LLM server 102.

[0038] Among the functional elements to be implemented, the generation AI module 310 is a deep learning model that has been pre-trained on a large dataset. In other words, this module is an LLM (i.e., a Large-Scale Language Model). This module outputs a response to an input query (in other words, a prompt or command). Because this module uses a deep learning model pre-trained on a large dataset, it can be used without training data or additional training.

[0039] 1-2-3. User terminal 103 Figure 4 shows an example of the configuration of user terminal 103. The user terminal 103 is, for example, a terminal device such as a smartphone, tablet, notebook PC, or desktop PC.

[0040] The main memory 401 of this terminal stores programs and applications such as the browser module 410. The processor 403 executes these programs and applications to realize the various functional elements of the user terminal 103.

[0041] The browser module 410 exchanges information with the question answering server 101.

[0042] 1-3.Operation Next, the question answering process 700, which is executed by the question answering server 101, will be explained with reference to Figure 7. Figure 7 is a flowchart showing an example of the question answering process 700.

[0043] First, the question answering server 101 receives a request from the user terminal 103 and displays the chat screen 900 on the user terminal 103 (step 701).

[0044] Figure 9 shows an example of a chat screen 900. The screen shown in the figure includes a query input field 901 and a send button 902. On this screen, the user enters a query in the query input field 901 and selects the send button 902. As a result, the user terminal 103 sends the entered query to the question answering server 101. The sent query is received by the reception module 210.

[0045] When the reception module 210 receives a query (YES in step 702), the first retrieval module 211 inputs the received query to the generation AI module 310 (step 703). The generation AI module 310 outputs a response corresponding to the input query. The first retrieval module 211 retrieves the outputted response (step 704).

[0046] Once the response text is obtained, the splitting module 212 divides the obtained response text into multiple chunks, paragraph by paragraph (step 705).

[0047] Next, the second acquisition module 213 identifies advertisements for each of the divided chunks (step 706). Specifically, this module performs an advertisement identification process 800.

[0048] Figure 8 is a flowchart showing an example of the advertisement identification process 800. In the flow shown in the figure, first, the second acquisition module 213 identifies the chunk to be processed from among multiple chunks (step 801).

[0049] Next, the module identifies ads that match the chunk to be processed (step 802). Specifically, the module refers to ad-related information 600 to identify ads that contain phrases determined by keywords and match types within the chunk. If multiple ads are identified, the module refers to ad-related information 600 to identify the ad with the highest ad rank among the identified ads. On the other hand, if the chunk does not contain phrases determined by keywords and match types, the module does not identify any ads.

[0050] Next, the module determines whether all chunks have been processed (step 803). If the result of this determination is that not all chunks have been processed (NO in step 803), the module returns to step 801 and processes another chunk. On the other hand, if all chunks have been processed (YES in step 803), the module terminates the advertisement identification process 800. As a result of the ad identification process 800 described above, a matching ad is identified for each of the multiple chunks.

[0051] After the ad identification process 800 is completed, the first display module 214 updates the chat screen 900 and displays the chat screen 1000 on the user terminal 103 (step 707). The chat screen 1000 contains the response text obtained in step 704 and the ad identified in step 706.

[0052] Figure 10 shows an example of a chat screen 1000. The screen shown in the figure includes a response text 1001, advertisements 1002 and 1003, a query input field 1004, a send button 1005, and a query 1022.

[0053] Of these, answer text 1001 is displayed in the answer text area 1006 on the left side of the screen. In contrast, advertisements 1002 and 1003 are displayed in the advertisement area 1007 on the right side of the screen. In other words, answer text 1001 and advertisements 1002 and 1003 are displayed separately.

[0054] Of the two ads, 1002 and 1003, ad 1002 is displayed to the right of chunk 1008, which makes up the answer text 1001. Ad 1002 is an ad that matches chunk 1008, and therefore is displayed in the vicinity of chunk 1008.

[0055] Advertisement 1002 consists of a sponsorship notice 1009, a logo image 1010, a company name 1011, a display URL 1012, an advertisement headline 1013, and a description 1014. Advertisement 1002 includes a sponsorship notice 1009 to clearly indicate that it is an advertisement, thereby making it possible to distinguish Advertisement 1002 from the response text 1001.

[0056] On the other hand, advertisement 1003 is displayed to the right of chunk 1015, which makes up the answer sentence 1001. This advertisement 1003 is an advertisement that matches chunk 1015, and therefore is displayed in the vicinity of chunk 1015.

[0057] Advertisement 1003 consists of a sponsorship notice 1016, a logo image 1017, a company name 1018, a display URL 1019, an advertisement headline 1020, and a description 1021. Advertisement 1003 includes a sponsorship notice 1016 to clearly indicate that it is an advertisement, thereby making it possible to distinguish Advertisement 1003 from the response text 1001.

[0058] On this screen, the user can enter a new query in the query input field 1004 and select the submit button 1005. When the user enters a new query and selects the submit button 1005, the user terminal 103 sends the entered query to the question answering server 101. The submitted query is received by the reception module 210.

[0059] When the reception module 210 receives a new query (YES in step 708), the judgment module 215 calculates the relevance between the received new query and each advertisement identified in step 706 (step 709). In doing so, the module first converts the new query and each advertisement (where an advertisement is a set of company name, ad headline, and description) into vector representations. For this conversion to vector representations, for example, Sentence Transformers is used. The module then calculates the relevance (for example, cosine similarity) between the generated vector representation of the new query and the vector representation of each advertisement.

[0060] After calculating the relevance, the module determines whether the calculated relevance exceeds a predetermined threshold (step 710). If the result of this determination is that the calculated relevance does not exceed the predetermined threshold (NO in step 710), the process returns to step 703. On the other hand, if the result of this determination is that the calculated relevance exceeds the predetermined threshold (YES in step 710), the module increments the number of times the ad with that relevance has been prompted (see Figure 6) (step 711). This is because it is presumed that the ad was viewed by the user, since a query related to that ad was entered.

[0061] Then, the third acquisition module 216 acquires the detailed advertisement from the auxiliary storage device 202 (step 712). The detailed advertisement acquired here includes the logo image, company name, display URL, advertisement headline, description, and site links (see Figure 5). Furthermore, as detailed advertising data obtained here, information extracted from the landing page may be used, as will be explained later.

[0062] After obtaining the detailed advertisement, the first acquisition module 211 inputs the query received in step 708 into the generation AI module 310 (step 713). The generation AI module 310 outputs a response corresponding to the input query. The first acquisition module 211 retrieves the outputted response (step 714).

[0063] After obtaining the response text, the second display module 217 updates the chat screen 1000 and displays the chat screen 1100 on the user terminal 103 (step 715). The chat screen 1100 contains the response text obtained in step 714 and the detailed advertisement obtained in step 712.

[0064] Figure 11 shows an example of a chat screen 1100. The screen shown in the figure includes a query 1101, a response text 1102, a detailed advertisement 1103, a query input field 1104, and a send button 1105.

[0065] Of these, the answer text 1102 is displayed in the answer text area 1106 on the left side of the screen. In contrast, the detailed advertisement 1103 is displayed in the advertisement area 1107 on the right side of the screen. In other words, the answer text 1102 and the detailed advertisement 1103 are displayed separately.

[0066] The detailed advertisement 1103 is displayed to the right of the answer text 1102. In other words, the detailed advertisement 1103 is displayed in the vicinity of the answer text 1102.

[0067] Detailed advertisement 1103 consists of a sponsored label 1108, a logo image 1109, a company name 1110, a display URL 1111, an ad headline 1112, a description 1113, and a site link 1114. This detailed advertisement 1103 includes a sponsored label 1108 to clearly indicate that it is an advertisement, thereby making it possible to distinguish the detailed advertisement 1103 from the response text 1102.

[0068] On this screen, the user can enter a new query in the query input field 1104 and select the submit button 1105. When the user enters a new query and selects the submit button 1105, the user terminal 103 sends the entered query to the question answering server 101. The submitted query is received by the reception module 210.

[0069] When the query is accepted by the reception module 210 (YES in step 702), the process in step 703 is executed again. The above is an explanation of question answering process 700.

[0070] The question-answering process 700 described above is intended to be executed on a user terminal 103 with a sufficiently large display space (for example, a desktop PC). If the user terminal 103 does not have a sufficiently large display space (for example, if the user terminal 103 is a smartphone), a different chat screen 1200 may be displayed in step 707 above.

[0071] Figure 12 shows an example of another chat screen 1200. The screen shown in this figure includes an advertisement area 1201, a response text area 1202, a scroll bar 1203, a query input field 1204, and a send button 1205.

[0072] Of these, advertisement 1206 is displayed in the advertising area 1201 at the top of the screen. Advertisement 1206 consists of a sponsored notice 1207 and an advertisement headline 1208. Advertisement 1206 includes a sponsored notice 1207 to clearly indicate that it is an advertisement, thereby making it possible to distinguish the response text 1209 from advertisement 1206.

[0073] The answer text area 1202 is located below and adjacent to the advertisement area 1201. The answer text 1209 is displayed in this answer text area 1202. On this screen, the advertisement area 1201 and the answer text area 1202 are separated, so the advertisement 1206 and the answer text 1209 are distinguished. Note that answer 1209 can be scrolled vertically using the scroll bar 1203.

[0074] The response text area 1202 has a display ad determination area 1210. This display ad determination area 1210 is an area for determining which ads to display in the ad area 1201. The first display module 214 displays ads in the ad area 1201 that match the chunks contained in this display ad determination area 1210. The ad 1206 displayed in the ad area 1201 in Figure 12 matches the chunk 1211 contained in the display ad determination area 1210.

[0075] Furthermore, since the display advertisement determination area 1210 is located adjacent to the lower part of the advertisement area 1201, the advertisement 1206 will be displayed adjacent to its corresponding chunk 1211.

[0076] Figure 13 shows an example of the chat screen 1300 that appears when the response text 1209 is scrolled down in the chat screen 1200. In the screen shown in the figure, an advertisement 1301 is displayed in the advertisement area 1201. This advertisement 1301 matches the chunk 1302 contained in the display advertisement determination area 1210.

[0077] In this way, the first display module 214 can switch the advertisements it displays in response to scrolling of the displayed answer text.

[0078] Furthermore, when performing the above question-answering process 700 on a user terminal 103 with limited display space, the process for displaying detailed advertisements (steps 712-715) may be omitted due to the limited display space.

[0079] According to the embodiment described above, relevant advertisements can be displayed for the response text generated by the generation AI. In addition, even if an advertisement is not selected by the user, the advertisement can be considered viewed if a query related to that advertisement is entered, and the advertiser can be charged for the cost of that advertisement.

[0080] 2. Variations The above embodiment may be modified as follows. The following modifications may be combined with each other.

[0081] 2-1. Related Information In the above embodiment, advertisements are assumed as related information and their detailed information. However, related information is not limited to advertisements. Related information does not have to be provided for commercial purposes, as long as it is information related to the answer.

[0082] 2-2. How to identify advertisements In the ad identification process 800 described above, ads are identified based on chunks of the response text (see step 802). Alternatively, ads may be identified based on the response text and the query.

[0083] In this case, the second acquisition module 213 acquires relevant information (in other words, advertisements) based on the answer text and the question text (in other words, the query). At that time, the module refers to the advertisement-related information 600 and determines whether the combination of query and chunk contains phrases determined by the advertisement's keywords and match type. If the result of this determination is positive, the module identifies the advertisement as matching the chunk. On the other hand, if the result of this determination is negative, the module does not identify the advertisement as matching the chunk. If the module identifies multiple advertisements, it refers to the advertisement-related information 600 and identifies the advertisement with the highest ad rank among the identified advertisements. This method allows for the identification of ads by taking queries into account.

[0084] Alternatively, in step 706 of the question-answering process 700, advertisements may be identified based solely on the query. In this case, the second retrieval module 213 retrieves relevant information (in other words, advertisements) based on the question (in other words, the query). At that time, the module refers to the advertisement-related information 600 and identifies advertisements in which the query contains phrases determined by keywords and match types. If the module identifies multiple advertisements, it refers to the advertisement-related information 600 and identifies the advertisement with the highest ad rank among the identified advertisements. On the other hand, if the query does not contain phrases determined by keywords and match types, the module does not identify any advertisements.

[0085] This method makes it possible to identify ads based solely on the query. The identified ads are displayed near the response text.

[0086] Furthermore, if this method is adopted, it is not necessary to identify advertisements for each chunk, so step 705 of the question answering process 700 may be omitted.

[0087] 2-3. Methods for identifying advertisements (Part 2) In the ad identification process 800 described above, keywords are used to identify ads that match the chunk (see step 802). Alternatively, vector representation may be used to identify ads that match the chunk.

[0088] In this case, the second acquisition module 213 first converts the chunks and each advertisement (where an advertisement is a set of company name, ad headline, and description) into vector representations. For this conversion to vector representations, for example, Sentence Transformers is used. The module then calculates the degree of relationship (for example, cosine similarity) between the generated vector representations of the chunks and the vector representations of each advertisement.

[0089] After calculating relevance, the module determines whether the calculated relevance exceeds a predetermined threshold. If the calculated relevance exceeds the predetermined threshold, the module identifies the ad with that relevance as a match for the chunk. On the other hand, if the calculated relevance does not exceed the predetermined threshold, the module does not identify the ad with that relevance as a match for the chunk. If the module identifies multiple ads, it refers to ad-related information 600 and identifies the ad with the highest ad rank among the identified ads. This method also allows us to identify ads that match the chunk.

[0090] 2-4. Ad placement In the chat screen 1000 shown in Figure 10, advertisements 1002 and 1003 are displayed to the right of the response text 1001. However, the display position of advertisements 1002 and 1003 is not limited to the position shown in this example. Advertisements 1002 and 1003 may be displayed in other locations as long as they are displayed near their corresponding chunks. For example, advertisement 1002 may be displayed to the left, above, or below its corresponding chunk 1008. If displayed above or below chunk 1008, advertisement 1002 will be displayed sandwiched between chunk 1008 and its adjacent chunks.

[0091] In the chat screen 1100 shown in Figure 11, the detailed advertisement 1103 is displayed to the right of the response text 1102. However, the display position of the detailed advertisement 1103 is not limited to the position shown in this example. The detailed advertisement 1103 may be displayed in a different location as long as it is near the response text 1102. For example, the detailed advertisement 1103 may be displayed to the left, above, or below the response text 1102.

[0092] In the chat screen 1200 shown in Figure 12, the advertisement 1206 is displayed above the response text 1209. However, the display position of the advertisement 1206 is not limited to the position shown in this example. The advertisement 1206 may be displayed in a different location as long as it is displayed near the corresponding chunk. For example, by setting up an advertisement area 1201 below the response text area 1202, and further setting up a display advertisement determination area 1210 below the response text area 1202, the advertisement 1206 may be displayed below the corresponding chunk.

[0093] 2-5. Unit of division of the answer text In the above embodiment, it is assumed that the response text is divided into paragraphs. However, the unit of chunking is not limited to paragraphs. As another example, the response text may be divided into sentences, fixed-length units (in other words, a predetermined number of characters or words), or thematic units.

[0094] 2-6. Splitting the answer text In the question-answering process 700 described above, the answer is divided into multiple chunks, and an advertisement is identified for each chunk (see steps 705 and 706). Alternatively, the advertisement may be identified based on the entire answer without dividing it. In this case, step 705 is omitted, and in step 706, the second acquisition module 213 acquires advertisements based on the entire answer. At that time, the module refers to the ad-related information 600 and identifies advertisements in the answer that contain phrases determined by keywords and match types. If the module identifies multiple advertisements, it refers to the ad-related information 600 and identifies the advertisement with the highest ad rank among the identified advertisements. On the other hand, if the answer does not contain phrases determined by keywords and match types, the module does not identify any advertisements.

[0095] This method makes it possible to identify advertisements based on the entire response text. Identified advertisements are displayed in the vicinity of the response text.

[0096] 2-7. Chunking In the question-answering process 700 described above, the question-answering server 101 performs the division of the answer text (see step 705). Alternatively, the LLM server 102 may be instructed to perform the division of the answer text. In this case, the division module 212 instructs the generation AI module 310 to divide the answer text and obtain multiple chunks.

[0097] 2-8. Detailed Advertisement In the question-answering process 700 described above, the detailed advertisement is obtained from the auxiliary storage device 202 (see step 712). Alternatively, the detailed advertisement may be obtained from the landing page of the advertisement. In this case, the third acquisition module 216 accesses the landing page using the display URL of the advertisement (see Figure 5) and obtains the information to be displayed as the detailed advertisement from the accessed page.

[0098] 2-9. Chat screen 1200, 1300 The chat screens 1200 and 1300 shown in Figures 12 and 13 are intended to be displayed on a user terminal 103 with limited display space (e.g., a smartphone). However, the same display method may also be used for user terminals 103 with sufficient display space (e.g., a desktop PC).

[0099] 2-10.Functional layout In the above embodiment, the question answering process 700 is executed by two servers, the question answering server 101 and the LLM server 102. However, this functional arrangement is merely an example. The functions of each server may be appropriately arranged according to the execution environment of the question answering process 700.

[0100] 2-11. System Components Each device constituting the above-described information processing system 100 may be, for example, a mobile device such as a smartphone, tablet, cell phone, or personal digital assistant (PDA), or a wearable device such as glasses, a wristwatch, or clothing. Each device may also be a stationary or portable computer, or a server located on the cloud or a network. Functionally, each device may be a VR (Virtual Reality) terminal, an AR (Augmented Reality) terminal, or an MR (Mixed Reality) terminal. Alternatively, a combination of multiple such terminals may be used. For example, a combination of one smartphone and one wearable device can logically function as a single terminal. Other types of information processing terminals may also be used.

[0101] 2-12. Others It should be noted that the present invention is not limited to the embodiments described above, and various modifications are included. For example, the embodiments described above are described in detail to make the present invention easier to understand, and are not necessarily limited to those having all the configurations described. Furthermore, it is possible to replace parts of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add configurations from other embodiments to the configuration of one embodiment. In addition, it is possible to add, delete, or replace parts of the configuration of each embodiment with other configurations.

[0102] Furthermore, each of the above configurations, functions, processing units, and processing means may be implemented in hardware, either partially or entirely, by designing them as integrated circuits, for example. Alternatively, each of the above configurations and functions may be implemented in software by having the processor interpret and execute programs that implement each function. Information such as programs, tables, and files that implement each function can be stored in memory, a recording device such as a hard disk or SSD (Solid State Drive), or a recording medium such as an IC card, SD card, or DVD.

[0103] Furthermore, the control lines and information lines shown are those deemed necessary for explanatory purposes, and not all control lines and information lines are necessarily shown in the actual product. In reality, it is safe to assume that almost all components are interconnected. Furthermore, the above-described embodiments disclose at least the configuration described in the claims. [Explanation of Symbols]

[0104] 100... Information processing system, 101... Question answering server, 102... LLM server, 103... User terminal

Claims

1. A means of receiving input from users regarding questions, A first acquisition means for obtaining an answer sentence generated by a generation AI based on the aforementioned question sentence, A second acquisition means for acquiring relevant information based on at least one of the aforementioned question and answer, A first display means for displaying the related information near a portion of the answer text related to the related information, An information processing system equipped with the following features.

2. The second acquisition means acquires the related information based on the answer text. The information processing system according to claim 1.

3. A division means for dividing the aforementioned answer into multiple parts including the aforementioned part, The second acquisition means acquires the related information based on the portion. The information processing system according to claim 1.

4. The division means instructs the generating AI to divide the answer sentence and obtains the plurality of parts. The information processing system according to claim 3.

5. The first display means displays the related information separately from the answer text. The information processing system according to claim 1.

6. The receiving means receives another question from the user, and if the other question is related to the related information, the means further includes a determination means for determining that the user has viewed the related information. The information processing system according to claim 1.

7. When the receiving means receives another question from the user, and the first acquisition means acquires another answer generated by the generation AI based on the other question, and the other question is related to the related information, the system further includes a second display means for displaying detailed information of the related information near the other answer. The information processing system according to claim 1.

8. The second display means does not display any related information other than the detailed information in the vicinity of the other answer sentence. The information processing system according to claim 7.

9. A division means for dividing the aforementioned answer into multiple parts including the aforementioned part, The second acquisition means acquires relevant information for each of two or more parts, including the aforementioned part, among the plurality of parts. The first display means switches the displayed related information from among the related information acquired for each of the two or more parts, in accordance with the scrolling of the displayed answer text. The information processing system according to claim 1.

10. A method of information processing performed by a computer, The steps include receiving a question from the user, The steps include obtaining an answer generated by a generation AI based on the aforementioned question, A step of obtaining relevant information based on at least one of the above question and answer statements, The steps include displaying the relevant information near a portion of the response text that is related to the relevant information, Information processing methods including

11. On the computer, The steps include receiving a question from the user, The steps include obtaining an answer generated by a generation AI based on the aforementioned question, A step of obtaining relevant information based on at least one of the above question and answer statements, The steps include displaying the relevant information near a portion of the response text that is related to the relevant information, A program to execute.

Images