Information processing system, information processing method, and program

Abstract

To provide an information processing system, an information processing method, and a program that can make operations efficient in planning for transaction targets such as commodities and services and planning for sales promotion activities.SOLUTION: An information processing system comprises: a segment generation unit 12 that extracts network users corresponding to conditions for specifying the population of consumer layers to appeal, and generates segment data in which the network users extracted on the basis of the details of posts on the SNS are classified by attributes; a campaign generation unit 13 that generates campaign data on sales promotion activities for appealing the details for appealing input from the user, for appealing to the consumer layer having a specific attribute in the segment data selected by a user; and a concept generation unit 22 that generates concept data that embodies information on transaction targets input from the user.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present invention relates to an information processing system, an information processing method, and a program for supporting planning of trading objects such as products and services, and planning of sales promotion activities. [Background technology]

[0002] Technologies for improving the efficiency of planning and developing business transactions, such as products and services, and planning and developing sales promotion activities for those transactions, by using a computer system, are being considered. For example, Patent Document 1 describes a system for determining a product or service suitable for a user by using a computer system. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2022 / 191210 Summary of the Invention [Problem to be solved by the invention]

[0004] In planning business transactions and sales promotion activities, there is a need for greater efficiency in order to respond to rapid changes in the market and the diversification of consumer needs. [Means for solving the problem]

[0005] An information processing system for solving the above problem comprises a segment generation unit that extracts from a search target a plurality of network users who meet conditions for identifying a population of a consumer segment to be appealed to, acquires behavioral information on the plurality of network users from the search target, and generates segment data that classifies the plurality of network users by attribute based on the behavioral information via a large-scale language model; a campaign generation unit that generates, via a large-scale language model, campaign data for sales promotion activities to appeal to the appeal content input by the user, with a consumer segment having a specific attribute in the segment data selected by the user as the appeal target; and a concept generation unit that generates, via a large-scale language model, concept data that concretizes the information on the transaction target input by the user.

[0006] An information processing method for solving the above problem is an information processing method executed by an information processing system having a control unit, and executes the following steps: extracting from a search target multiple network users who meet conditions for identifying a population of a consumer segment to be appealed to, acquiring behavioral information about the multiple network users from the search target, and generating segment data that classifies the multiple network users by attribute based on the behavioral information via a large-scale language model; generating campaign data for sales promotion activities to appeal to a consumer segment having a specific attribute in the segment data selected by a user as the appeal target, via a large-scale language model; and generating concept data that specifies information about the transaction target entered by the user, via a large-scale language model. A program for solving the above problem causes a control unit included in the information processing system to function as a means for executing the above information processing method.

[0007] The segment generation process by the segment generation unit can improve the efficiency of the work of considering target purchasing groups when planning a transaction target, and the work of considering targets to appeal to when planning sales promotion activities. The campaign generation process by the campaign generation unit can improve the efficiency of the work of planning sales promotion activities. The concept generation process by the concept generation unit can improve the efficiency of the work of planning a transaction target. Therefore, the above information processing system, information processing method, and program can improve the efficiency of the work of planning a transaction target and sales promotion activities.

[0008] In the information processing system, if there is already generated concept data for the appeal content input by the user, the campaign generation unit may generate the campaign data using the already generated concept data. With this configuration, it is possible to improve the consistency and consistency of content between the concept data for a specific transaction object and the campaign data that appeals to the transaction object.

[0009] In the information processing system, the concept generation unit may generate the concept data by defining a consumer demographic having a specific attribute in the segment data as a target purchasing demographic for the transaction object. With the above configuration, the concept data for the transaction object can be generated after clarifying the consumer demographic that will be the target purchasing demographic.

[0010] In the information processing system, the concept data may include a concept image of the transaction object, and the concept generation unit may generate the concept image through an image generation model, the image generation model being fine-tuned using images related to the business activities of a specific business entity as training data. According to the above configuration, the concept image may reflect the image designs of images related to products or services previously provided by the specific business entity. This may improve the completeness of the concept image.

[0011] In the information processing system, the campaign data may include campaign images related to the sales promotion activities, and the campaign generation unit may generate the campaign images via the image generation model. This configuration allows a specific business entity's previous product designs or service-related image designs to be reflected in the campaign images, thereby improving the completeness of the campaign images. Furthermore, generating campaign images using an image generation model that generates concept images can improve the sense of unity and consistency between the concept images and the campaign images.

[0012] The information processing system may further include an image evaluation unit, wherein the information on the transaction target for generating the concept data includes target information for identifying a target demographic of the transaction target, and the image evaluation unit executes at least one of the following processes: scoring the campaign image via an image evaluation model so that the more the campaign image is preferred by the target consumer demographic, the higher a first scoring result is, and causing the campaign generation unit to generate a regenerated campaign image so that the first scoring result is higher when the first scoring result is equal to or lower than a first threshold; and scoring the concept image via the image evaluation model so that the more the concept image is preferred by the target consumer demographic corresponding to the target information, the higher a second scoring result is, and causing the concept generation unit to generate a regenerated concept image so that the second scoring result is higher when the second scoring result is equal to or lower than a second threshold. According to the above configuration, the image evaluation unit scores the campaign image and regenerates the campaign image so that the first scoring result is higher when the first scoring result is equal to or lower than the first threshold, thereby providing a campaign image preferred by the target consumer demographic. Similarly, the image evaluation unit scores the concept image and, if the second scoring result is below a second threshold, regenerates the concept image so that the second scoring result is higher, thereby providing a concept image that is preferred by the target purchasing demographic.

[0013] In the information processing system, the image evaluation model may be configured to have undergone a learning process using the segment data as training data. With this configuration, the accuracy of image evaluation by the image evaluation model can be improved as the segment data is accumulated.

[0014] In the information processing system, the search target may include one or more SNS (Social Networking Services), and the behavioral information may include content posted on the SNS. With this configuration, segment data can be created based on content posted on the SNS.

[0015] In the information processing system, the behavioral information may be text data created using a large-scale language model and may include a description explaining the content of the image posted to the SNS. According to the above configuration, segment data can be created based on the content of the image posted to the SNS. This allows the creation of segment data that reflects insights into the target consumer demographic.

[0016] In the information processing system, the information on the transaction object includes an input text related to a concept of the transaction object, and the concept data includes concept text data that embodies the input text. The concept generation unit may receive input of information for identifying a viewpoint to be emphasized as content of the concept data along with the information on the transaction object, input the input text and the information for identifying the viewpoint into a sentence scoring model to obtain emphasized text from the sentence scoring model that emphasizes the viewpoint of information contained in the input text, and input the input text and the emphasized text into a large-scale language model to obtain the concept text data from the large-scale language model. With the above configuration, it is possible to generate concept text data in which the viewpoint is more emphasized.

[0017] In the information processing system, the sentence scoring model may be configured to have undergone a learning process using the segment data as training data. With this configuration, trends in the segment data can be reflected in the emphasized text generated by the sentence scoring model. [Effects of the Invention]

[0018] According to the present invention, it is possible to improve the efficiency of work in planning and developing trading objects such as products and services, and in planning and developing sales promotion activities. [Brief explanation of the drawings]

[0019] [Figure 1] FIG. 1 is a diagram illustrating a configuration of an information processing system. [Figure 2] FIG. 2 is a diagram illustrating a configuration of a user terminal in the information processing system. [Figure 3] FIG. 3 is a diagram illustrating a configuration of a server in the information processing system. [Figure 4] FIG. 4 is a diagram showing the configuration of the planning support server unit and the database unit in the information processing system. [Figure 5] FIG. 5 is a flowchart of the segment generation process. [Figure 6] FIG. 6 is a diagram illustrating an example of segment data. [Figure 7] FIG. 7 is a flowchart of the campaign generation process. [Figure 8] FIG. 8 is a diagram illustrating an example of campaign data. [Figure 9] FIG. 9 is a flowchart of the first learning process. [Figure 10] FIG. 10 is a flowchart of the concept generation process. [Figure 11] FIG. 11 is a diagram showing an example of an input screen displayed on a user terminal. [Figure 12] FIG. 12 is a diagram illustrating an example of concept data. [Figure 13]FIG. 13 is a diagram illustrating an example of concept data. [Figure 14] FIG. 14 is a flowchart of the first process. [Figure 15] FIG. 15 is a flowchart of the second process. DETAILED DESCRIPTION OF THE INVENTION

[0020] An information processing system, an information processing method, and a program to which the present invention is applied will be described below with reference to the drawings. <Summary> An information processing system to which the present invention is applied is, for example, a system operated by a group of companies that are economically and organizationally related, or by a corporate entity consisting of a single company. Users who are employees of the corporate entity use this system when planning and developing transactions for trading objects, such as tangible goods or intangible services. Furthermore, users use this system when planning and developing sales promotion activities (hereinafter referred to as "sales promotion activities") to appeal to target consumer segments for trading objects, such as goods or services, using the trading objects as the appeal content. In other words, the information processing system is a computer system that supports the planning and development of trading objects and sales promotion activities.

[0021] <Configuration of Information Processing System 1> 1, the information processing system 1 includes a user terminal 2, a planning support server unit 3, a database unit 4, an LLM server 5, and an external server 6. These components are interconnected via a network line so that they can communicate with each other.

[0022] The user terminal 2 is a computer terminal managed by a user such as an employee of a company. The user terminal 2 is, for example, a mobile terminal such as a tablet or smartphone that is compatible with a mobile communication system. The user terminal 2 may also be a desktop or laptop personal computer. The user terminal 2 has installed thereon an application program for supporting the planning of business transactions and the planning of sales promotion activities, as well as a web page browser.

[0023] The planning support server unit 3 is composed of one or more servers managed by the company to which the user belongs. The planning support server unit 3 is accessed from the user terminal 2. The planning support server unit 3 is connected to a database unit 4, an LLM server 5, and an external server 6.

[0024] The planning support server unit 3 includes, for example, a purchasing desire acquisition server 10, a concept generation server 20, and an image evaluation server 30. The purchasing desire acquisition server 10 assists in understanding the purchasing desires of the target consumer segment. The purchasing desire acquisition server 10 also assists in planning promotional activities for the target consumer segment. The concept generation server 20 assists in planning trading objects such as products and services. The image evaluation server 30 evaluates images generated by the purchasing desire acquisition server 10 and the concept generation server 20. The planning support server unit 3 may be configured to realize the functions of the purchasing desire acquisition server 10, the concept generation server 20, and the image evaluation server 30 using one or more servers.

[0025] The database unit 4 is configured, for example, by one or more servers managed by the company to which the user belongs. The database unit 4 stores data necessary for performing various processes in the planning support server unit 3. The server that configures the database unit 4 may be a server separate from the planning support server unit 3. At least a part of the server that configures the database unit 4 may be shared with the server that configures the planning support server unit 3. Furthermore, at least a part of the server that configures the database unit 4 may be provided on an external server separate from the company to which the user belongs.

[0026] The LLM server 5 may be a server managed by the company to which the user belongs, or may be an external server. The LLM server 5 includes a large-scale language model. The large-scale language model is a model trained from a large text corpus. The large-scale language model is used to perform natural language understanding tasks. The large-scale language model has the ability to interpret sentences given in response to prompts and generate appropriate responses in that context.

[0027] The LLM server 5 may include a chat completion API. The chat completion API has a function of, for example, processing input sentences to generate prompts and inputting the generated prompts into a large-scale language model to obtain answers. The chat completion API complements chats between the planning support server unit 3 and the large-scale language model included in the LLM server 5.

[0028] The external server 6 is a server that is searched on an external network when the planning support server unit 3 searches for various information. For example, the external server 6 that is searched when the purchasing desire acquisition server 10 searches for information is a server of a social networking service (SNS). For example, the external server 6 that is searched when the concept generation server 20 searches for information is a server of an SNS, blog, e-commerce site, etc.

[0029] <User device 2> As shown in FIG. 2, the user terminal 2 includes a communication IF 2A, a touch panel 2B serving as an input unit and an output unit, a memory 2C, a storage unit 2D, and a control unit 2E.

[0030] The communication IF 2A is an interface for inputting and outputting signals for communication with external devices. The touch panel 2B is a device that combines a display panel as a display unit that displays images and a touchpad as an input unit that accepts user operations. The display panel is a liquid crystal display panel, an organic EL panel, or the like. The touch panel 2B constitutes a user terminal screen. The user terminal 2 may also include a keyboard, a mouse, or the like as an input unit. The user terminal 2 may also include a display panel without a touchpad as a display unit. The input unit may also be a microphone for voice input. The memory 2C temporarily stores programs and data processed by the programs, etc. The memory 2C is, for example, a volatile memory such as a DRAM (Dynamic Random Access Memory). The memory unit 2D is storage for saving data. The memory unit 2D is, for example, a flash memory, an HDD (Hard Disc Drive), etc. The control unit 2E is hardware for executing an instruction set described in a program. The control unit 2E is a processor configured by an arithmetic unit, a register, a peripheral circuit, and the like.

[0031] <Planning Support Server Unit 3> As shown in FIG. 3, the purchasing desire acquisition server 10, the concept generation server 20, and the image evaluation server 30, which are servers included in the planning support server unit 3, each include a communication IF 3A, an input / output IF 3B, a memory 3C, and a storage unit 3D.

[0032] The communication IF3A is an interface for inputting and outputting signals for communication with external devices. The input / output IF3B is an interface that functions as an input device for receiving operation input from an administrator of the information processing system 1 and as an output device for presenting information to the administrator. The memory 3C temporarily stores programs and data processed by the programs, etc. The memory 3C is, for example, a volatile memory such as a DRAM. The memory unit 2D is a storage for saving data. The memory unit 2D is, for example, a flash memory, an HDD, etc.

[0033] Furthermore, the purchasing desire acquisition server 10 includes a control unit 11. The concept generation server 20 includes a control unit 21. The image evaluation server 30 includes a control unit 31. The control units 11, 21, and 31 are hardware for executing an instruction set written in a program. The control units 11, 21, and 31 are processors configured with an arithmetic unit, registers, peripheral circuits, etc.

[0034] 4, the control unit 11 of the purchase desire acquisition server 10 executes a purchase desire acquisition program to function as a segment generation unit 12. The purchase desire acquisition program is stored in the storage unit 3D of the purchase desire acquisition server 10.

[0035] The segment generation unit 12 extracts a plurality of network users who meet specific conditions, and generates segment data by classifying the extracted network users by attribute based on behavioral information on the network.

[0036] For example, the segment generation unit 12 extracts multiple SNS users who meet specific conditions, and generates segment data by classifying the extracted SNS users by attribute based on the content of their posts on the SNS. The content of posts on the SNS is an example of consumer behavior information. The SNS users are an example of network users.

[0037] The specific conditions under which the segment generation unit 12 generates segment data are conditions set by the user via the user terminal 2, and are conditions for identifying a target consumer population. In other words, the segment data includes information that indicates the correspondence between the conditions for identifying a target consumer population and the attributes of each SNS user that makes up the population.

[0038] The generated segment data is stored in the segment database 41 provided in the database unit 4. The segment data can be used as consideration when selecting a consumer segment to appeal to in sales promotion activities. The segment data can also be used as consideration when selecting a target purchasing segment to transact with. The segment generation unit 12 uses the large-scale language model provided in the LLM server 5 when generating segment data.

[0039] The control unit 11 of the purchase desire acquisition server 10 executes a campaign generation program to function as a campaign generation unit 13. The campaign generation program is stored in the storage unit 3D of the purchase desire acquisition server 10.

[0040] The campaign generation unit 13 generates campaign data for sales promotion activities to appeal to consumer groups with specific attributes in the segment data generated by the segment generation unit 12. The appeal content is a product or service of the business entity to which the user belongs. The appeal content may be an existing product or service, or a product or service in the planning stage or under development. The campaign generation unit 13 generates the campaign data using information on the appeal content input from the user terminal 2.

[0041] In the campaign data, the target consumer group is selected from the segment data by the user, i.e., by input from the user terminal 2. The campaign data includes campaign text data for promoting the promotional content as a sales promotion activity and campaign images related to the sales promotion activity. The campaign text data includes, for example, the campaign name and detailed campaign content.

[0042] The generated campaign data is stored in a campaign database 42 provided in the database unit 4. When generating the campaign data, the campaign generation unit 13 uses a large-scale language model provided in the LLM server 5 and an image generation model 45 provided in the database unit 4.

[0043] The control unit 21 of the concept generation server 20 executes a concept generation program to function as a concept generation unit 22. The concept generation program is stored in the storage unit 3D of the concept generation server 20.

[0044] The concept generation unit 22 generates concept data by embodying information about a trading object that belongs to a predetermined product category or service category, input from the user terminal 2, as a concept proposal. The trading object is a product or service of the business entity to which the user belongs. The trading object may be, for example, a product or service that is in the planning stage or under development, but may also be a product or service that is on sale. In the information processing system 1, the product or service categories of trading objects for which concept data can be generated are set in advance by the administrator of the information processing system 1.

[0045] The information on the transaction target includes input text, which is text data related to the concept of the transaction target. The input text may be, for example, the category of the transaction target, a description of the transaction target, etc. The information on the transaction target may also include an image of the transaction target, target information for identifying a target demographic, etc. The target information may include, for example, the age and gender of the target demographic of the transaction target.

[0046] Concept data includes concept text data for explaining the concept of the transaction target and concept images of the transaction target. Concept text data is text data that embodies the input text as a concept proposal. Concept text data includes detailed information such as the name, tagline, ACB (Accepted Consumer Belief), RTB (Reason to Believe), and product specifications. The name here refers to the product name or service name of the transaction target. The tagline is a phrase that succinctly expresses the value and concept of the transaction target. The ACB is content that anticipates consumer problems related to the transaction target. The RTB is an idea that solves anticipated consumer problems. The specifications are information such as the specifications and price of the transaction target.

[0047] The generated concept data is stored in a concept database 43 provided in the database unit 4. When generating concept data, the concept generation unit 22 uses a large-scale language model provided in the LLM server 5 and an image generation model 45 provided in the database unit 4. The image generation model 45 used by the concept generation unit 22 is the same as the image generation model 45 used by the campaign generation unit 13.

[0048] Furthermore, in order to emphasize a particular viewpoint in the content of the concept data as a process for generating concept data, the concept generation unit 22 executes a sentence scoring process using the sentence scoring model 44 provided in the database unit 4. The viewpoint here corresponds to the numerical dimension when a word or part of a sentence is represented as a numerical vector by embedding.

[0049] In the information processing system 1, perspectives that can be emphasized as content of concept data are determined in advance. Perspectives that can be emphasized as content of concept data include, for example, a novelty perspective, which is how novel the transaction object is compared to conventional ones, and a purchase intention perspective, which is how likely the transaction object is to be accepted by consumers. Information for specifying a perspective to be emphasized as content of concept data is input from the user terminal 2.

[0050] In the text scoring process, the concept generation unit 22 inputs text about a transaction object, such as a product or service, input from the user terminal 2 and information for identifying a viewpoint to be emphasized as the content of concept data to the text scoring model 44. The text scoring model 44 is a machine learning model. The concept generation unit 22 then acquires, from the text scoring model 44, emphasized text in which information contained in the input text about the concept of the transaction object is emphasized with respect to that viewpoint. The concept generation unit 22 then inputs the input text about the concept of the transaction object and the emphasized text acquired from the text scoring model 44 into the large-scale language model, thereby acquiring concept data from the large-scale language model.

[0051] The control unit 21 of the concept generation server 20 executes a first learning program, thereby functioning as a first learning processing unit 23. The first learning program is stored in the storage unit 3D of the concept generation server 20.

[0052] The first learning processing unit 23 executes a first learning process S3 (see FIG. 9) of the sentence scoring model 44 included in the database unit 4. The first learning processing unit 23 uses, for example, information listed on an SNS, blog, e-commerce site, or the like of the external server 6 as training data for the first learning process S3 in the sentence scoring model 44.

[0053] Furthermore, the first learning processing unit 23 uses the segment data stored in the segment database 41 as training data for the first learning process S3 in the sentence scoring model 44. That is, the sentence scoring model 44 uses, as training data, the correspondence between the conditions for identifying the target consumer demographic obtained as segment data and the attributes of each SNS user constituting that population.

[0054] The control unit 31 of the image evaluation server 30 executes an image evaluation program to function as an image evaluation unit 32. The image evaluation program is stored in the storage unit 3D of the image evaluation server 30.

[0055] The image evaluation unit 32 uses the image evaluation model 46 provided in the database unit 4 to evaluate the campaign images generated by the campaign generation unit 13 and the concept images generated by the concept generation unit 22. Furthermore, when the evaluation value of the campaign image or concept image is equal to or less than a threshold, the image evaluation unit 32 instructs the campaign generation unit 13 or the concept generation unit 22 to regenerate an image that will have a higher evaluation value.

[0056] The control unit 31 of the image evaluation server 30 executes the second learning program, thereby functioning as a second learning processing unit 33. The second learning program is stored in the storage unit 3D of the image evaluation server 30.

[0057] The second learning processing unit 33 performs a second learning process for the image evaluation model 46 included in the database unit 4. The second learning processing unit 33 uses the segment data stored in the segment database 41 as training data for the second learning process for the image evaluation model 46. That is, the image evaluation model 46 uses, as training data, the correspondence between the conditions for identifying a target consumer demographic obtained as segment data and the attributes of each SNS user constituting that population.

[0058] <Database Section 4> As shown in FIG. 4, the database unit 4 includes a segment database 41, a campaign database 42, and a concept database 43.

[0059] The segment database 41 stores the segment data generated by the segment generation unit 12 in association with an ID for identifying the segment data. The segment database 41 stores, as segment data, conditions for identifying a target consumer population and attributes of each SNS user who makes up that population. The segment database 41 may also store the content of posts on SNS by SNS users used to generate the segment data, as well as summaries of those posts.

[0060] The campaign database 42 stores the campaign data generated by the campaign generation unit 13, an ID for identifying the campaign data, information on the appeal content input by the user, and information on the appeal target of the campaign data, in association with each other. The campaign database 42 stores, as campaign data, campaign text data on sales promotion activities and campaign images related to sales promotion activities.

[0061] The concept database 43 stores the concept data generated by the concept generation unit 22, an ID for identifying the concept data, and information on the transaction object input by the user in association with each other. The concept database 43 stores, as concept data, concept text data for explaining the concept of the transaction object and a concept image of the transaction object.

[0062] The database unit 4 also includes a text scoring model 44, an image generation model 45, and an image evaluation model 46. The sentence scoring model 44 is a machine learning model used by the concept generation unit 22 in the process of generating concept data. The sentence scoring model 44 captures the meaning and relationships between words and sentences by converting sentences or parts of sentences into numerical vectors. The sentence scoring model 44 receives input text related to the concept of the transaction target and information for identifying a perspective to be emphasized as content of the concept data, and outputs emphasized text that emphasizes the information contained in the input text with respect to that perspective. The sentence scoring model 44 preferably outputs a list of emphasized texts including multiple emphasized texts, but it is sufficient to output at least one emphasized text.

[0063] Specifically, the sentence scoring model 44 converts input text related to the concept of the transaction into a numerical vector and outputs emphasized text adjusted to increase the dimension value of the numerical vector for the viewpoint to be emphasized. The emphasized text may be a word, a part of a sentence, or an entire sentence. Note that emphasizing any text from a specific viewpoint means that the dimension value for the specific viewpoint is increased when the text is converted into a numerical vector.

[0064] In sentence scoring model 44, the learning content is updated periodically, for example, once a week or once a month, through a first learning process S3 by first learning processing unit 23, which uses information from external server 6 and segment data stored in segment database 41 as training data. The first learning process S3 of sentence scoring model 44 by first learning processing unit 23 will be described later.

[0065] The image generation model 45 is a machine learning model for image generation used in the process of generating campaign data by the campaign generation unit 13 and in the process of generating concept data by the concept generation unit 22. When an image generation prompt is input, the image generation model 45 outputs image data corresponding to the input image generation prompt.

[0066] The image generation model 45 is fine-tuned using images related to the business activities of a specific business entity as training data. The specific business entity here refers to the business entity to which the user, the owner of the user terminal 2, belongs. The images related to the business activities here also include images of trading objects, such as products and services, created by the business entity. In other words, images created using the image generation model 45 reflect the trends in images of trading objects created by the specific business entity up to that point. Furthermore, the image created by the image generation model 45 may use the business entity's logo or the like in the design of the trading object.

[0067] The image evaluation model 46 is a machine learning model used in the process of evaluating campaign images and concept images by the image evaluation unit 32. In the image evaluation model 46, the learning content is updated periodically, for example, weekly or monthly, through a second learning process by the second learning processing unit 33, using the segment data stored in the segment database 41 as training data.

[0068] <Segment generation process S1> The segment generation process S1 by the segment generation unit 12 will be described with reference to Figures 5 and 6. The segment generation process S1 is performed, for example, when a user makes a promotion of any product or service, with the aim of examining the target consumer demographic. The segment generation process S1 is performed according to the flowchart shown in Figure 5.

[0069] 5, the segment generation process S1 includes steps S1-1 to S1-6. First, in step S1-1, the segment generation unit 12 receives input of conditions for identifying a target consumer population from the user terminal 2. The conditions for identifying a target consumer population include, for example, the age, gender, occupation, income, residential area, household composition, hobbies, interests, and lifestyle of the consumer population.

[0070] Next, in step S1-2, the segment generation unit 12 acquires a search word for extracting, from the SNS server of the external server 6, SNS users who meet the conditions input from the user terminal 2.

[0071] In detail, the segment generation unit 12 inputs a prompt including an instruction for generating search words for extracting SNS users who meet the conditions for identifying a target consumer demographic, along with the conditions, into the large-scale language model of the LLM server 5. Then, the segment generation unit 12 acquires the search words from the large-scale language model. In other words, in step S1-2, the segment generation unit 12 causes the large-scale language model to generate search words for extracting SNS users who meet the conditions input from the user terminal 2.

[0072] Next, in step S1-3, the segment generation unit 12 uses the search word generated in step S1-2 to extract SNS users who meet the conditions entered from the user terminal 2 from the SNS server of the external server 6. The extracted SNS users constitute a population of the target consumer segment. Information on the extracted SNS users is stored in the segment database 41.

[0073] Then, all posted content of the extracted SNS user is extracted from the SNS server of the external server 6. At this time, the segment generation unit 12 may extract only posted content for a specific period in response to input from the user terminal 2. The posted content to be extracted includes image data, such as still images or videos, and text data. The extracted posted content is stored in the segment database 41.

[0074] Next, in step S1-4, the segment generation unit 12 acquires a caption for the image data acquired in step S1-3. The caption is an explanatory text that explains the content of the image data. The caption is an example of behavioral information of the SNS user.

[0075] In detail, the segment generation unit 12 inputs a prompt including an instruction sentence for generating a caption for the image data, together with the image data acquired in step S1-3, into the large-scale language model of the LLM server 5. Then, the segment generation unit 12 acquires a caption for the image data acquired in step S1-3 from the large-scale language model. In other words, in step S1-4, the segment generation unit 12 causes the large-scale language model to generate a caption for the image data acquired in step S1-3. The caption for the acquired image data is stored in the segment database 41.

[0076] Next, in step S1-5, the segment generation unit 12 acquires a summary of the posted content for each extracted SNS user based on the text data acquired in step S1-3 and the caption of the image data acquired in step S1-4.

[0077] In detail, the segment generation unit 12 inputs a prompt including an instruction to generate a summary of the posted content, together with the text data included in the posted content and the captions of the image data included in the posted content, into the large-scale language model of the LLM server 5. Then, the segment generation unit 12 acquires a summary of the posted content for each SNS user extracted in step S1-3 from the large-scale language model. In other words, in step S1-5, the segment generation unit 12 causes the large-scale language model to generate a summary of the posted content for each SNS user extracted in step S1-3. The acquired summary of the posted content is stored in the segment database 41.

[0078] Next, in step S1-6, the segment generation unit 12 acquires segment data in which the SNS users extracted in step S1-3 are classified by attribute, based on the summary of the content posted by each SNS user acquired in step S1-5.

[0079] In detail, the segment generation unit 12 inputs a prompt including an instruction sentence for classifying the extracted SNS users by arbitrary attribute, together with a summary of the posted content of each SNS user acquired in step S1-5, into the large-scale language model of the LLM server 5. Then, the segment generation unit 12 acquires segment data obtained by classifying the SNS users extracted in step S1-3 by attribute from the large-scale language model. In other words, in step S1-6, the segment generation unit 12 causes the large-scale language model to generate segment data obtained by classifying the SNS users extracted in step S1-3 by attribute. The acquired segment data is stored in the segment database 41 and is displayed on the user terminal 2.

[0080] The attributes for classifying the extracted SNS users are derived, for example, from the most frequently posted content of all posts by each SNS user. Therefore, in one piece of segment data, the attributes for classifying the extracted SNS users may include a variety of categories such as age, gender, occupation, income, residential area, household composition, hobbies, interests, and lifestyle.

[0081] In the segment generation process S1, segment data is created by classifying target consumers into attributes based on the content of their posts on SNS, according to the procedure of steps S1-1 to S1-6 described above. The output format of the segment data may be a pie chart, as described below, or a table.

[0082] 6 shows a pie chart 100 as an example of segment data. The pie chart 100 represents segment data when the lifestyle condition "health-conscious people who prioritize organic and natural products" is input as the condition input in step S1-1. In other words, the pie chart 100 is segment data in which SNS users who belong to the consumer demographic of "health-conscious people who prioritize organic and natural products" are classified by attribute based on the content of their posts.

[0083] For example, area 101 in pie chart 100 shows that 11.8% of SNS users belonging to the consumer demographic of "health-conscious people who prioritize organic and natural products" have the attribute "coffee lovers." Also, area 102 in pie chart 100 shows that 4.7% of SNS users belonging to the same demographic have the attribute "dog owners."

[0084] By referring to the pie chart 100 displayed on the user terminal 2, the user can understand what attributes the consumer population that meets the conditions entered in step S1-1 has. Therefore, by referring to the pie chart 100 displayed on the user terminal 2, the user can carry out sales promotion activities that target consumer groups with specific attributes.

[0085] <Campaign generation process S2> The campaign generation process S2 performed by the campaign generation unit 13 will be described with reference to Figures 7 and 8. In the campaign generation process S2, campaign data is generated for sales promotion activities to promote any desired content, such as a product or service, targeting a consumer group having a specific attribute in the segment data selected by the user. Therefore, the campaign generation process S2 is performed after the segment generation process S1. The campaign generation process S2 is performed according to the flowchart shown in Figure 7.

[0086] 7, the campaign generation process S2 includes steps S2-1 to S2-6. First, in step S2-1, the campaign generation unit 13 receives input of various conditions from the user terminal 2 when generating campaign data.

[0087] The conditions for generating campaign data include information on the products or services to be promoted and information on the target consumer demographic. The information on the promotion includes the product or service name and a description of the product or service. The information on the target consumer demographic includes information for specifying a consumer demographic with specific attributes in the segment data of a specific population generated in the segment generation process S1. In other words, in the campaign data generated by the campaign generation unit 13, the consumer demographic with specific attributes in the target segment data is set by user selection, i.e., input from the user terminal 2.

[0088] Furthermore, the information on the target consumer group may include the age, gender, etc. of the target consumer group. In this case, the target consumer group is a consumer group having a specific attribute in the segment data, and is further limited in terms of age, gender, etc. In addition to the above information, detailed campaign information such as the campaign concept and purpose may be input as a condition for generating campaign data.

[0089] Next, in step S2-2, the campaign generation unit 13 acquires campaign text data for promoting the appeal content to the target of the appeal, depending on various conditions when generating the campaign data input from the user terminal 2 in step S2-1.

[0090] In detail, the campaign generation unit 13 inputs a prompt including an instruction sentence for generating campaign text data, together with the information input from the user terminal 2 in step S2-1, to the large-scale language model of the LLM server 5. Then, the campaign generation unit 13 acquires the campaign text data from the large-scale language model. In other words, in step S2-2, the campaign generation unit 13 causes the large-scale language model to generate campaign text data for promoting the appeal content to the appeal target.

[0091] Next, in step S2-3, the campaign generation unit 13 acquires an image generation prompt for causing the image generation model 45 to generate a campaign image in accordance with the content of the campaign text data.

[0092] In detail, the campaign generation unit 13 inputs a prompt including an instruction sentence for generating an image generation prompt for causing the image generation model 45 to generate a campaign image to the large-scale language model of the LLM server 5. At this time, the campaign generation unit 13 inputs the campaign text data acquired in step S2-2 together with the above-mentioned prompt to the large-scale language model of the LLM server 5. The campaign generation unit 13 may also input information input from the user terminal 2 in step S2-1 to the large-scale language model of the LLM server 5. Then, the campaign generation unit 13 acquires the image generation prompt from the large-scale language model. In other words, in step S2-3, the campaign generation unit 13 causes the large-scale language model to generate an image generation prompt for causing the image generation model 45 to generate a campaign image corresponding to the campaign text data.

[0093] Next, in step S2-4, the campaign generation unit 13 acquires a campaign image by inputting the image generation prompt acquired in step S2-3 into the image generation model 45. In other words, in step S2-4, the campaign generation unit 13 causes the image generation model 45 to generate a campaign image.

[0094] Next, in step S2-5, the campaign generation unit 13 inputs the campaign image generated in step S2-4 into the image evaluation server 30, thereby evaluating the campaign image.

[0095] In detail, first, the campaign generation unit 13 inputs the campaign image generated in step S2-4 together with information on the target consumer segment corresponding to the campaign image to the image evaluation server 30. Then, the image evaluation unit 32 of the image evaluation server 30 inputs the campaign image input from the campaign generation unit 13 and the information on the target consumer segment corresponding to the campaign image to the image evaluation model 46, thereby scoring the campaign image.

[0096] The image evaluation model 46 scores the campaign image based on information about the target consumer segment, so that the higher the scoring result, the more the input campaign image is preferred by the target consumer segment. Hereinafter, the scoring result of the campaign image by the image evaluation model 46 will be referred to as the first scoring result. The image evaluation model 46 outputs, together with the first scoring result, first guideline information, which is text data including improvements to make the campaign image more preferred by the target consumer segment. The image evaluation unit 32 obtains the first scoring result of the campaign image and the first guideline information from the image evaluation model 46.

[0097] When the first scoring result by the image scoring model 46 is equal to or less than a predetermined first threshold, the image evaluation unit 32 inputs, together with the first guideline information, a command to the purchase desire acquisition server 10 to cause the campaign generation unit 13 to regenerate a campaign image so as to increase the first scoring result. The campaign generation unit 13 inputs the first guideline information to the image generation model 45 in addition to the image generation prompt acquired in step S2-3, thereby acquiring a regenerated campaign image that is a campaign image reflecting the first guideline information.

[0098] The campaign generation unit 13 inputs the campaign regenerated image to the image evaluation server 30. Then, the image evaluation unit 32 inputs the campaign regenerated image and information on the target consumer group corresponding to the campaign regenerated image to the image evaluation model 46, thereby scoring the campaign regenerated image.

[0099] In step S2-5, the campaign image evaluation process and regeneration process are repeated until the first scoring result exceeds a predetermined first threshold. When the first scoring result of the campaign image or the campaign regenerated image exceeds the predetermined first threshold, the image evaluation unit 32 outputs a signal indicating that the first scoring result has exceeded the predetermined first threshold to the purchasing desire acquisition server 10. The campaign generation unit 13 of the purchasing desire acquisition server 10 is triggered by the signal to end the process of step S2-5 and proceed to the next step S2-6.

[0100] Next, in step S2-6, the campaign generation unit 13 outputs campaign data to the user terminal 2. The campaign data output by the campaign generation unit 13 to the user terminal 2 includes the campaign text data acquired in step S2-2 and the campaign image acquired in step S2-4 or the campaign reproduced image acquired in step S2-5. In addition, the campaign data output by the campaign generation unit 13 to the user terminal 2 is stored in the campaign database 42.

[0101] In the campaign generation process S2, campaign data is created for sales promotion activities to appeal to consumer groups with specific attributes in the segment data through the procedure of steps S2-1 to S2-6 described above.

[0102] FIG. 8 shows a promotional image 200 as an example of campaign data. The promotional image 200 is campaign data for a consumer demographic of "coffee lovers" selected from the population of "health-conscious individuals who prioritize organic and natural products" in the segment data of the pie chart 100 shown in FIG. 6. The promotional content in this case is a lactic acid bacteria drink with the product name "□△ lactic acid bacteria drink," which is sold by the company to which the user belongs. In other words, the promotional image 200 is campaign data for a promotional activity to promote the lactic acid bacteria drink to the consumer demographic of "health-conscious individuals who prioritize organic and natural products" and "coffee lovers."

[0103] The promotional image 200 includes a text area 201. The text area 201 contains the promotional logo, the campaign title, campaign text data, and other content. The campaign text data here includes text to promote the lactic acid bacteria drink to consumers who are "health-conscious people who prioritize organic and natural products" and "coffee lovers." The text area 201 may also contain tags for classifying keywords and topics on social media, etc.

[0104] The promotional image 200 includes an image area 202. In the image area 202, a campaign image is displayed that shows the campaign content described in the text area 201. The campaign image here includes a bottle of the lactic acid bacteria drink that is the promotional content, coffee beans, a cup filled with the lactic acid bacteria drink alone, a cup filled with coffee made using the lactic acid bacteria drink, etc. Using this promotional image 200, users can carry out promotional activities to promote the lactic acid bacteria drink that is the promotional content to consumers who are "health-conscious people who prioritize organic and natural products" and "coffee lovers."

[0105] <First learning process S3 of sentence scoring model 44> 9, the first learning process S3 of the sentence scoring model 44 by the first learning processing unit 23 will be described. In the first learning process S3, the sentence scoring model 44 learns words and sentence expressions for emphasizing a particular viewpoint in each category of transaction objects set in advance by the administrator of the information processing system 1. In the first learning process S3, the words and sentence expressions on the external server 6 are used as training data. The first learning process S3 of the sentence scoring model 44 is performed according to the flowchart shown in FIG.

[0106] 9, the first learning process S3 of the sentence scoring model 44 includes steps S3-1 to S3-4. First, in step S3-1, the first learning processing unit 23 sets a category of the transaction object to be learned by the sentence scoring model 44 in accordance with input from the administrator of the information processing system 1. For example, it is assumed here that "beer" has been set as the category of the transaction object to be learned by the first learning process S3.

[0107] Next, in step S3-2, the first learning processing unit 23 sets the viewpoint (evaluation axis, dimension) of sentence expression to be learned by the sentence scoring model 44 in accordance with input from the administrator of the information processing system 1. For example, it is assumed here that "novelty" is set as the viewpoint of sentence expression for the category of transaction object for which the first learning process S3 is performed.

[0108] Next, in step S3-3, the first learning processing unit 23 acquires from the external server 6 expressions relating to the viewpoint of the sentence expression input in step S3-2 for the category of the transaction object input in step S3-1.

[0109] For example, the first learning processing unit 23 acquires expressions related to "novelty" of "beer" from the external server 6. In this case, the expressions acquired from the external server 6 may be, for example, a product description or a catchphrase for a new beer product on an e-commerce site. The expressions acquired from the external server 6 may also be campaign information for a new beer product on a webpage operated by a beer manufacturer. The expressions acquired from the external server 6 may also be reviews of the new beer product by consumers on social media, blogs, or e-commerce sites.

[0110] Next, in step S3-4, first learning processing unit 23 uses the expressions acquired in step S3-3 as training data to train sentence scoring model 44. As described above, first learning processing unit 23 performs first learning process S3 of sentence scoring model 44 through the procedure of steps S3-1 to S3-4.

[0111] Once the transaction object category and the viewpoint of sentence expression are set in steps S3-1 and S3-2, expression extraction from the external server 6 in step S3-3 and model training in step S3-4 are performed periodically and automatically. That is, the first learning processing unit 23 automatically executes the first learning process S3 of the sentence scoring model 44 every time a predetermined period elapses.

[0112] <Concept generation process S4> 10 to 13, the concept generation process S4 performed by the concept generation unit 22 will be described. In the concept generation process S4, concept data for the transaction object is generated. The concept generation process S4 is performed according to the flowchart shown in FIG. 10. Also, FIG. 11 shows an example of a screen for starting the concept generation process S4 in application software for the information processing system 1 installed in the user terminal 2.

[0113] 10, the concept generation process S4 includes steps S4-1 to S4-8. First, in step S4-1, the concept generation unit 22 receives input of various conditions from the user terminal 2 for generating concept data.

[0114] 11 shows an input screen 300 for inputting various conditions for generating concept data from the user terminal 2 to the concept generation unit 22. The input screen 300 is an example of a screen realized by application software for the information processing system 1 installed in the user terminal 2.

[0115] As shown in FIG. 11, the input screen 300 includes a category selection object 301. The category selection object 301 is an object for inputting the category of the transaction object for which concept data is to be generated. The category selection object 301 is, for example, a drop-down list type object. In response to an operation input by the user, the category selection object 301 displays a list of categories of the transaction object for which concept data can be generated in the concept generation process S4. The categories of the transaction object for which concept data can be generated correspond to the categories for which the first learning process S3 of the above-mentioned sentence scoring model 44 was performed.

[0116] The input screen 300 includes a description input object 302. The description input object 302 is an object for inputting a description of a transaction object for which concept data is to be generated. The content of the description of the transaction object is not particularly limited as long as it is text data related to the transaction object. For example, if the category of the transaction object is beer, the description of the transaction object is a sentence about the taste and aroma.

[0117] The input screen 300 includes a sentence balance adjustment object 303. The sentence balance adjustment object 303 is an object for inputting the balance of the viewpoint to be emphasized as the content of the concept data. In this embodiment, the viewpoint of novelty and the viewpoint of purchase intention are set by the system administrator as viewpoints whose balance can be adjusted by the sentence balance adjustment object 303. Therefore, the sentence balance adjustment object 303 is used to input information indicating which viewpoint, novelty or purchase intention, is to be emphasized. The information input by the sentence balance adjustment object 303 is an example of information for specifying the viewpoint to be emphasized as the content of the concept data. Note that the viewpoints whose balance can be adjusted by the sentence balance adjustment object 303 correspond to the viewpoints of sentence expression that the sentence scoring model 44 was made to learn in the first learning process S3 of the sentence scoring model 44 described above.

[0118] The input screen 300 includes a trend reflection necessity selection object 304. The trend reflection necessity selection object 304 is, for example, a checkbox-type object. When the trend reflection necessity selection object 304 is checked, the sentence scoring model 44 adjusts the expression of the highlighted text so that concept data that reflects expressions in line with the most recent trends is generated.

[0119] The input screen 300 includes an industry selection object 305. The industry selection object 305 is an object for selecting an arbitrary industry by, for example, inputting text. When a specific industry is selected by the industry selection object 305, the sentence scoring model 44 adjusts the expression of the highlighted text so that concept data reflecting expressions specific to that industry is generated.

[0120] The input screen 300 includes a target information input object 306. The target information input object 306 is an object for inputting the age and gender of the target purchasing group. The target information input object 306 includes, for example, a checkbox-type object for inputting the gender and an object for inputting the age.

[0121] The input screen 300 includes a free keyword input object 307. The free keyword input object 307 is an object that accepts the input of any text. For example, text that further limits the target purchasing demographic may be input into the free keyword input object 307 as target information, or text that more specifically describes the transaction target may be input as input text related to the concept of the transaction target.

[0122] The input screen 300 includes an image template input object 308. The image template input object 308 is an object that accepts input of image data. The image template input object 308 is used when inputting image data that is to be reflected in concept data. For example, image data that serves as a reference when generating a concept image may be input into the image template input object 308.

[0123] The input screen 300 includes a visual keyword input object 309. The visual keyword input object 309 is an object for inputting conditions for generating a concept image in text format. The text input to the visual keyword input object 309 is added to the image generation prompt when generating a concept image.

[0124] The input screen 300 includes an input content deletion object 310 for deleting input content for each object, and a generation start object 311 for starting generation of concept data. After the user has completed inputting data for each object, the user operates the generation start object 311, whereby in step S4-1 the concept generation unit 22 accepts input of various conditions for generating concept data from the user terminal 2.

[0125] Returning to FIG. 10, next, in step S4-2, the concept generation unit 22 performs a sentence scoring process using the sentence scoring model 44 to obtain emphasized text by emphasizing a specific viewpoint from the input text entered in step S4-1.

[0126] In detail, the concept generation unit 22 inputs the input text about the transaction object input in step S4-1 and information for specifying a viewpoint to be emphasized as the content of the concept data input in step S4-1 to the sentence scoring model 44. Then, the concept generation unit 22 acquires emphasized text in which information included in the input text about the transaction object is emphasized with respect to that viewpoint from the sentence scoring model 44. For example, in step S4-1, the concept generation unit 22 acquires from the sentence scoring model 44 a list of emphasized text about the input text input in each object on the input screen 300.

[0127] Next, in step S4-3, the concept generation unit 22 further expands the list of emphasized texts by searching the external server 6 for the contents of the emphasized texts based on the list of emphasized texts acquired in step S4-2. That is, the concept generation unit 22 searches the external server 6 using words and phrases included in the list of emphasized texts acquired in step S4-2, thereby acquiring expressions related to the emphasized texts from the external server 6. This allows the list of emphasized texts to be further expanded. Note that the processing of step S4-3 may be omitted.

[0128] Next, in step S4-4, the concept generation unit 22 acquires text data for concepts via the large-scale language model of the LLM server 5 using the information on the trading object input in step S4-1.

[0129] In detail, the concept generation unit 22 inputs a prompt including an instruction sentence for generating text data for a concept to the large-scale language model of the LLM server 5. At this time, the concept generation unit 22 inputs the input text for the transaction object input in step S4-1 and the list of emphasized text acquired in steps S4-2 and S4-3, along with the prompt, to the large-scale language model of the LLM server 5. Then, the concept generation unit 22 acquires text data for a concept that embodies the input text from the large-scale language model. In other words, in step S4-4, the concept generation unit 22 causes the large-scale language model to generate text data for a concept.

[0130] Next, in step S4-5, the concept generation unit 22 acquires a prompt for generating a concept image in accordance with the content of the concept text data.

[0131] In detail, the concept generation unit 22 inputs a prompt including an instruction sentence for generating an image generation prompt for causing the image generation model 45 to generate a concept image to the large-scale language model of the LLM server 5. At this time, the concept generation unit 22 inputs the concept text data acquired in step S4-4 to the large-scale language model of the LLM server 5 together with the above-mentioned prompt. The concept generation unit 22 may also input the input text about the transaction object input in step S4-1 and the list of emphasized text acquired in steps S4-2 and S4-3 to the large-scale language model of the LLM server 5. Then, the concept generation unit 22 acquires the image generation prompt from the large-scale language model. In other words, in step S4-5, the concept generation unit 22 causes the large-scale language model to generate an image generation prompt for causing the image generation model 45 to generate a concept image corresponding to the concept text data.

[0132] Next, in step S4-6, the concept generation unit 22 acquires a concept image by inputting the image generation prompt acquired in step S4-5 into the image generation model 45. In other words, in step S4-6, the concept generation unit 22 causes the image generation model 45 to generate a concept image.

[0133] Next, in step S4-7, the concept generation unit 22 inputs the concept image generated in step S4-6 into the image evaluation server 30, thereby evaluating the concept image.

[0134] In detail, first, the concept generation unit 22 inputs the concept image generated in step S4-6 together with target information for identifying the target purchasing demographic corresponding to the concept image to the image evaluation server 30. Then, the image evaluation unit 32 of the image evaluation server 30 inputs the concept image input from the concept generation unit 22 and the target information for identifying the target purchasing demographic corresponding to the concept image to the image evaluation model 46, thereby scoring the concept image.

[0135] The image evaluation model 46 scores the campaign image based on the target information so that the more the input concept image is preferred by the target purchasing demographic corresponding to the target information, the higher the scoring result. Hereinafter, the scoring result of the concept image by the image evaluation model 46 will be referred to as the second scoring result. The image evaluation model 46 outputs, together with the second scoring result, second guideline information, which is text data including improvements to make the concept image more preferred by the target purchasing demographic corresponding to the target information. The image evaluation unit 32 obtains the second scoring result of the concept image and the second guideline information from the image evaluation model 46.

[0136] When the second scoring result by the image scoring model 46 is equal to or less than a predetermined second threshold, the image evaluation unit 32 inputs, together with the second guideline information, a command to the concept generation server 20 to cause the concept generation unit 22 to generate a concept image again so as to increase the second scoring result. The concept generation unit 22 inputs the second guideline information to the image generation model 45 in addition to the image generation prompt acquired in step S4-5, thereby acquiring a regenerated concept image, which is a concept image reflecting the second guideline information.

[0137] The concept generation unit 22 inputs the concept-regenerated image to the image evaluation server 30. Then, the image evaluation unit 32 inputs the concept-regenerated image and target information for identifying the target demographic corresponding to the concept-regenerated image to the image evaluation model 46, thereby scoring the concept-regenerated image.

[0138] In step S4-7, the evaluation process and regeneration process of the concept image are repeated until the second scoring result exceeds a predetermined second threshold. When the second scoring result of the concept image or the regenerated concept image exceeds the predetermined second threshold, the image evaluation unit 32 outputs a signal indicating that the second scoring result has exceeded the predetermined second threshold to the concept generation server 20. The concept generation unit 22 of the concept generation server 20 is triggered by the signal to end the process of step S4-7 and proceed to the next step S4-8.

[0139] Next, in step S4-8, the concept generation unit 22 outputs concept data to the user terminal 2. The concept data output by the concept generation unit 22 to the user terminal 2 includes the concept text data acquired in step S4-4, and the concept image acquired in step S4-6 or the concept-regenerated image acquired in step S4-7. In addition, the concept data output by the concept generation unit 22 to the user terminal 2 is stored in the concept database 43.

[0140] In the concept generation process S4, concept data is created based on various conditions input by the user via the input screen 300 displayed on the user terminal 2, according to the procedure of steps S4-1 to S4-8 described above.

[0141] 12 and 13 show a product concept screen 400 as an example of concept data. The product concept screen 400 is product concept data used when planning a banana-flavored beer to be traded. The product concept screen 400 is an example of a screen realized by application software for the information processing system 1 installed in the user terminal 2.

[0142] As shown in FIG. 12, the product concept screen 400 includes a name display object 401 , a tagline display object 402 , an ACB display object 403 , an RTB display object 404 , and a product specification display object 405 .

[0143] The name display object 401 displays a concept proposal for the product's name. The tagline display object 402 displays a concept proposal for the product's tagline. The ACB display object 403 displays a concept proposal for the ACB of the product. The RTB display object 404 displays a concept proposal for the RTB of the product. The product specification display object 405 displays a concept proposal for the product's specifications. For example, if the product being traded is banana-flavored beer, the product specifications include information such as the volume per bottle, the type of container, the price per bottle, and the alcohol content.

[0144] The contents of the proposed concept displayed in these objects 401-405 are based on the concept text data acquired by the concept generation unit 22 in step S4-4. Each of the objects 401-405 includes an evaluation object 400A for the user to evaluate the contents of the proposed concept, and a reproduction object 400B for outputting an alternative to the contents of the proposed concept. The evaluation results of the objects 401-405 by the evaluation object 400A are fed back to the large-scale language model of the LLM server 5.

[0145] 13, the product concept screen 400 includes an image object 406. The image object 406 displays a concept image of the product to be traded. For example, the image object 406 displays multiple concept image proposals. Like other objects, the image object 406 also includes an evaluation object 400A and a reproduction object 400B. The evaluation result of the image object 406 by the evaluation object 400A is fed back to the image generation model 45.

[0146] The product concept screen 400 includes a save object 407, a download object 408, and a send object 409. The save object 407 is an object for instructing the concept generation unit 22 to perform a process of storing the above-mentioned concept data in the concept database 43 in response to a user's operational input. The download object 408 is an object for downloading the above-mentioned concept data to the user terminal 2. The send object 409 is an object for instructing the concept generation unit 22 to perform a process of transmitting the above-mentioned concept data to another device such as a mail server. The user can use the concept data represented as the above-mentioned product concept screen 400 to plan a transaction target.

[0147] <1st usage example> A first use example of the information processing system 1 will be described with reference to Fig. 14. In the first use example, a user acquires concept data of a transaction object using the information processing system 1, and then acquires campaign data for conducting promotional activities for the transaction object. In the first use example, the information processing system 1 executes a first process that is performed in accordance with the flowchart shown in Fig. 14.

[0148] 14, in the first process, the information processing system 1 first generates concept data of the transaction object through a concept generation process S4. The generated concept data is output to the user terminal 2 and stored in the concept database 43. Hereinafter, the concept data stored in the concept database 43 will be referred to as generated concept data.

[0149] Next, the information processing system 1 generates segment data for examining target consumer groups through a segment generation process S1. The generated segment data is output to the user terminal 2 and stored in the segment database 41.

[0150] Next, the information processing system 1 generates campaign data in the campaign generation process S2, using the transaction target for which concept data was generated in the concept generation process S4 as the appeal content. At this time, the appeal target is selected to be a consumer group having a specific attribute from the segment data generated in the segment generation process S1.

[0151] In the first process, if generated concept data for the appeal content input from the user terminal 2 exists in the concept database 43, the campaign generation unit 13 generates campaign data using the generated concept data.

[0152] In detail, the campaign generation unit 13 determines whether the appeal content input from the user terminal 2 matches the transaction object of the generated concept data stored in the concept database 43. If the appeal content input from the user terminal 2 matches the transaction object of the generated concept data, the campaign generation unit 13 generates campaign data using the generated concept data. That is, the campaign generation unit 13 generates campaign text data and campaign images based on the contents of the concept text data and concept image included in the generated concept data.

[0153] According to this first process, the consistency and consistency of content and terminology between the concept data of a specific transaction object and the campaign data that appeals to that transaction object is increased. Note that if the appeal content input from the user terminal 2 does not match the transaction object of the generated concept data, the campaign data is generated according to the procedure shown in FIG. 7 above.

[0154] <Second usage example> A second use example of the information processing system 1 will be described with reference to Fig. 15. In the second use example, a user uses the information processing system 1 to consider target consumer segments using segment data, and then acquires concept data that sets consumer segments with specific attributes in the segment data as target purchasing segments for transactions. In the second use example, the information processing system 1 executes a second process that is performed in accordance with the flowchart shown in Fig. 15.

[0155] 15, in the second process, the information processing system 1 first generates segment data for examining target consumer groups through a segment generation process S1. The generated segment data is output to the user terminal 2 and stored in the segment database 41. Hereinafter, the segment data stored in the segment database 41 will be referred to as generated segment data.

[0156] Next, the information processing system 1 generates concept data for the transaction object through a concept generation process S4. In the second process, information for identifying a consumer group having a specific attribute in the generated segment data is input as target information to be input when generating the concept data. For example, as the target information, conditions for identifying a population of consumer groups in the generated segment data and information for identifying the attributes of SNS users who make up that population are input. As a result, the concept generation unit 22 generates concept data in which a consumer group having a specific attribute in the segment data is the target purchasing group of the transaction object. According to this second process, it is possible to generate concept data for the transaction object after clarifying the consumer group that will be the target purchasing group.

[0157] In the second process, the concept generation unit 22 may compare the target information input from the user terminal 2 with the generated segment data stored in the segment database 41. In this case, the concept generation unit 22 searches for a combination that matches or is similar to the target information input from the user terminal 2, among the combination patterns of the population conditions in the generated segment data and the attributes of the SNS users that make up the population. If the target information input from the user terminal 2 matches or is similar to the information included in the generated segment data, the concept generation unit 22 generates concept data using the generated segment data and the information used to generate it.

[0158] In the second process, similar to the first process, concept data for a transaction object may be generated by the concept generation process S4, and then campaign data for conducting sales promotion activities for the transaction object may be generated by the campaign generation process S2. In this case, similar to the first process, the campaign generation unit 13 generates campaign text data and campaign images based on the contents of the concept text data and concept images included in the generated concept data.

[0159] <Effects of the embodiment> (1) The information processing system 1 includes a segment generation unit 12, a campaign generation unit 13, and a concept generation unit 22. The segment generation unit 12 generates segment data for considering an appeal target. The campaign generation unit 13 generates campaign data regarding sales promotion activities. The concept generation unit 22 generates concept data regarding a transaction target.

[0160] The segment generation unit 12 can improve the efficiency of the work of considering target purchasing groups when planning transaction targets, and the work of considering appeal targets when planning sales promotion activities. The campaign generation unit 13 can improve the efficiency of the work of planning sales promotion activities. The concept generation unit 22 can improve the efficiency of the work of planning transaction targets. Therefore, the information processing system 1, the information processing method using the information processing system 1, and the program that executes the information processing method can improve the efficiency of the work of considering target purchasing groups and appeal targets, and the work of planning transaction targets and sales promotion activities.

[0161] The segment generation unit 12 can provide segment data that reduces the bias inherent in human judgment, as material for considering target purchasing demographics when planning business transactions and for considering target audiences when planning sales promotion activities. The campaign generation unit 13 creates campaign data targeting the consumer demographics considered using the segment data, thereby enabling sales promotion activities that reduce the bias inherent in human judgment. Therefore, the information processing system 1, the information processing method using the information processing system 1, and the program that executes the information processing method can provide deliverables based on more objective data.

[0162] (2) In the first process, if there is generated concept data for the appeal content input from the user terminal 2, the campaign generation unit 13 generates campaign data using the generated concept data. This first process improves the consistency and consistency of content and terminology between the concept data for a specific transaction object and the campaign data that appeals to that transaction object. Therefore, higher quality campaign data can be output.

[0163] (3) In the second process, the concept generation unit 22 generates concept data in which a consumer demographic with specific attributes in the segment data generated in advance is the target purchasing demographic of the transaction. According to this second process, it is possible to clarify the target purchasing demographic and then generate concept data suitable for the target purchasing demographic. Furthermore, by creating concept data in which the consumer demographic considered using the segment data is the target purchasing demographic of the transaction, it is possible to formulate a concept that reduces the bias inherent in human judgment.

[0164] (4) The image generation model 45 is fine-tuned using images related to the business activities of a specific business entity as training data. This allows the specific business entity's past product image designs or service image designs to be reflected in concept images and campaign images. This allows for improved completion of concept images and campaign images.

[0165] (5) The image evaluation unit 32 scores the campaign image via the image evaluation model 46 so that the higher the first scoring result, the more the campaign image is liked by the target consumer segment. If the first scoring result is equal to or less than the first threshold, the image evaluation unit 32 causes the campaign generation unit 13 to generate a campaign reproduction image so that the first scoring result becomes higher. This makes it possible to provide a campaign image that is liked by the target consumer segment.

[0166] Similarly, the image evaluation unit 32 scores the concept image via the image evaluation model 46 so that the more the concept image is preferred by the target purchasing demographic corresponding to the target information, the higher the second scoring result. If the second scoring result is equal to or less than the second threshold, the image evaluation unit 32 causes the concept generation unit 22 to generate a concept-regenerated image so that the second scoring result is higher. This makes it possible to provide a concept image preferred by the target purchasing demographic.

[0167] (6) The image evaluation model 46 undergoes a learning process using the segment data stored in the segment database 41 as training data. As a result, the accuracy of image evaluation by the image evaluation model 46 can be improved as segment data accumulates.

[0168] (7) In the segment generation process S1, when searching for network users who meet the criteria for identifying the target consumer demographic, the search targets on the network include one or more SNSs. Furthermore, the behavioral information used to generate segment data by classifying network users includes the content posted on the SNSs. This allows segment data to be created based on the content posted on the SNSs.

[0169] (8) In the segment generation process S1, the text data created using a large-scale language model is captions, which are explanatory texts explaining the content of images posted to the SNS. These captions are used as behavioral information when classifying SNS users. This allows segment data to be created based on the content of images posted to the SNS. This allows segment data to be created that reflects insights into the target consumer demographic.

[0170] (9) The concept generation unit 22 generates emphasized text that emphasizes, from a specific perspective, information included in the input text related to the concept of the transaction object input from the user terminal 2 via the sentence scoring model 44. The concept generation unit 22 inputs the input text and the emphasized text into a large-scale language model, thereby acquiring concept text data from the large-scale language model. This makes it possible to generate concept text data in which a specific perspective is more emphasized.

[0171] (10) The sentence scoring model 44 is subjected to a learning process using the segment data as training data, which allows the tendency of the segment data to be reflected in the emphasized text generated by the sentence scoring model 44.

[0172] <Example of change> The above embodiment can be modified as follows: The following modifications can be implemented in combination with each other within the scope of technical compatibility.

[0173] There are no particular limitations on the training data used in the first learning process S3 of the sentence scoring model 44. That is, it is not necessary to use segment data as training data, and for example, only expressions written in the external server 6 may be used as training data.

[0174] In the concept generation process S4, the text scoring process using the text scoring model 44 may be omitted. In this case, the concept generation process S4 only needs to be configured to be able to output concept text data that embodies at least the input text related to the concept of the transaction object input from the user terminal 2.

[0175] The behavioral information used to classify SNS users in the segment generation process S1 does not necessarily need to include captions that explain the content of images posted to the SNS. For example, the behavioral information may consist only of text posted to the SNS.

[0176] In the segment generation process S1, when searching for network users who meet the criteria for identifying the target consumer demographic, the search target on the network is not limited to social networking sites. For example, the search target on the network can be any platform on which content can be posted. For example, it can be a blog, a podcast, any website, or a web bulletin board.

[0177] The training data used in the learning process of the image evaluation model 46 is not particularly limited, and segment data does not necessarily have to be used as training data. In the campaign generation process S2, the process of evaluating the campaign image by the image evaluation unit 32 and regenerating the campaign image may be omitted. Similarly, in the concept generation process S4, the process of evaluating the concept image by the image evaluation unit 32 and regenerating the concept image may be omitted. That is, in the information processing system 1, the function of the image evaluation server 30 may be omitted.

[0178] The image generation model 45 does not necessarily need to be fine-tuned using images related to the business activities of a specific company as training data. For example, if the image generation model 45 is not fine-tuned as described above, it will output images with new designs that are unrelated to the designs of images of existing products or services. Therefore, it is possible to output images with a higher degree of novelty.

[0179] The second process is an example of a process that combines the segment generation process S1 and the concept generation process S4. Therefore, in the information processing system 1, it is not necessary to execute the second process, and the segment generation process S1 and the concept generation process S4 can be executed in any order.

[0180] The campaign generation unit 13 may not need to perform a process of determining whether or not generated concept data exists for the appeal content input from the user terminal 2. In this case, the campaign generation unit 13 may generate campaign data without using generated concept data.

[0181] The input screen 300 may include a target selection object for selecting a consumer group with specific attributes from the generated segment data as a means for inputting target information. The target selection object displays, for example, a segment list showing combinations of the conditions of the consumer group population in the generated segment data and the attributes of the SNS users that make up that population. In the target selection object, the user selects one combination from the combinations of the population conditions and SNS user attributes listed in the segment list, and the selected combination is input as target information. [Explanation of symbols]

[0182] S1...Segment generation process S2: Campaign generation process S4: Concept generation process 1. Information processing system 2...User terminal 3...Planning support server section 4. Database Department 5...LLM server 6...External server 11, 21, 31...Control section 12...Segment generation unit 13...Campaign Generation Unit 22...Concept generation section 32...Image evaluation unit 44...Text scoring model 45...Image generation model 46...Image evaluation model 100...pie chart 200...Promotional images 300...Input screen 400...Product concept screen

Claims

1. a segment generation unit that extracts from a search target a plurality of network users who meet conditions for identifying a target consumer demographic, acquires behavioral information about the plurality of network users from the search target, and generates segment data that classifies the plurality of network users by attribute based on the behavioral information via a large-scale language model; a campaign generation unit that generates, via a large-scale language model, campaign data for sales promotion activities to appeal to consumer groups having specific attributes in the segment data selected by the user as the appeal target, and to appeal to the appeal content input by the user; a concept generation unit that generates concept data that embodies information about the transaction object input by the user through a large-scale language model; Information processing system.

2. When generated concept data exists for the appeal content input by the user, the campaign generation unit generates the campaign data using the generated concept data. The information processing system according to claim 1 .

3. The concept generation unit generates the concept data by setting a consumer group having a specific attribute in the segment data as a target purchasing group of the transaction target. The information processing system according to claim 1 .

4. the concept data includes a concept image of the trading object; the concept generation unit generates the concept image via an image generation model; The image generation model is fine-tuned using images related to the business activities of a specific company as training data.

4. The information processing system according to claim 1.

5. the campaign data includes campaign images related to the sales promotion activity; The campaign generation unit generates the campaign image via the image generation model. The information processing system according to claim 4 .

6. further comprising an image evaluation unit; The information on the transaction target for generating the concept data includes target information for identifying a target demographic of the transaction target, The image evaluation unit scoring the campaign image via an image evaluation model so that the more the campaign image is preferred by the target consumer group, the higher the first scoring result; and when the first scoring result is equal to or less than a first threshold, causing the campaign generation unit to generate a campaign reproduction image so that the first scoring result is higher; scoring the concept image via the image evaluation model so that the more the concept image is preferred by a target purchasing group corresponding to the target information, the higher the second scoring result; and, when the second scoring result is equal to or less than a second threshold, causing the concept generation unit to generate a concept-regenerated image so that the second scoring result is higher. The information processing system according to claim 5 .

7. The image evaluation model has undergone a learning process using the segment data as training data. The information processing system according to claim 6.

8. the search target includes one or more SNS (Social Networking Service), The behavioral information includes the content posted on the SNS.

4. The information processing system according to claim 1.

9. The behavioral information is text data created using a large-scale language model, and includes a description explaining the content of the image posted on the SNS. The information processing system according to claim 8 .

10. the information on the trading object includes an input text related to a concept of the trading object; the concept data includes concept text data that embodies the input text, The concept generation unit Accepting input of information for specifying a viewpoint to be emphasized as content of the concept data together with information on the transaction object; By inputting the input text and information for identifying the viewpoint into a sentence scoring model, an emphasized text in which information included in the input text is emphasized with respect to the viewpoint is obtained from the sentence scoring model; The input text and the emphasized text are input into a large-scale language model, and the concept text data is obtained from the large-scale language model.

4. The information processing system according to claim 1.

11. The sentence scoring model has undergone a learning process using the segment data as training data. The information processing system according to claim 10.

12. An information processing method executed by an information processing system including a control unit, a segment generation process for extracting from a search target a plurality of network users who meet the conditions for identifying a target consumer demographic, acquiring behavioral information on the plurality of network users from the search target, and generating segment data by classifying the plurality of network users by attribute based on the behavioral information via a large-scale language model; a campaign generation process for generating, via a large-scale language model, campaign data for sales promotion activities to appeal to consumer groups having specific attributes in the segment data selected by the user as the target consumers, and to appeal to the content of the appeal input by the user; A concept generation process is performed to generate concept data that embodies the information on the transaction subject input by the user through a large-scale language model. Information processing methods.

13. The control unit provided in the information processing system functions as a means for executing the information processing method according to claim 12. program.

Images