Information processing device, control method, and program

Abstract

To provide an information processing device, a control method, and a program capable of generating prompts for generating advertisements that are more effective than those already delivered. [Solution] The information processing device includes: an acquisition unit that acquires appeal information indicating the degree to which a delivered advertisement appeals to users; a policy information determination unit that determines policy information indicating the degree to which the delivered advertisement should be improved according to the acquired appeal information; a content determination unit that determines the content of the next advertisement according to the determined policy information; and a prompt generation unit that generates a prompt including reference text to be used by the generation AI model when generating the next advertisement according to the determined content of the next advertisement.

Description

Technical Field

[0004] , ,

[0006] , , , ,

[0005] , , ,

[0003] , , ,

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

Background Art

[0002] Advertisements have the role of conveying to consumers the characteristics, values, usage methods, etc. of products and services and promoting their purchase and use. As a technology related to such advertisements, Patent Document 1 discloses an automatic generation system for advertisement means. Patent Document 1 discloses that scoring is performed based on the click-through rate of the distributed advertisement means, etc., and if it is above a threshold value, an image tag set and an interest feature tag set are integrated, the image tag set predicts the interest feature tag set, a prompt is generated from the interest feature tag set, and the generated prompt is input into GPT-4.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In order to perform appropriate marketing such as CRM (Customer Relationship Management) and sales promotion using advertisements, it is necessary to create advertisements regularly or frequently and also create advertisements corresponding to various consumers.

[0005] Under these circumstances, it is conceivable to generate advertisements using a generative AI model. However, for products and services for which advertisements have already been made, when generating advertisements again using a generative AI model, a prompt for generating a more effective advertisement than the already distributed advertisements has not been developed yet. [[ID=​One aspect of this disclosure provides an information processing device, a control method, and a program capable of generating prompts for generating advertisements that are more effective than those already delivered. [Means for solving the problem]

[0007] An information processing device according to one aspect of this disclosure includes: an acquisition unit that acquires appeal information indicating the degree to which a delivered advertisement appeals to users; a policy information determination unit that determines policy information indicating the degree to which the delivered advertisement should be improved according to the acquired appeal information; a content determination unit that determines the content of the next advertisement according to the determined policy information; and a prompt generation unit that generates a prompt including reference text to be used by the generation AI model when generating the next advertisement according to the determined content of the next advertisement. [Brief explanation of the drawing]

[0008] [Figure 1] This figure shows an example of an advertising generation system. [Figure 2A] The previous example of the structure of advertising information is shown in the figure below. [Figure 2B] This figure shows an example of the structure of conditional information. [Figure 2C] This figure shows an example of how improvement information can be structured. [Figure 2D] This figure shows an example of the structure of reference text information. [Figure 2E] This figure shows an example of how to configure enhancement information. [Figure 2F] This figure shows an example of the structure of product and service information. [Figure 3A] This figure shows an example of condition information. [Figure 3B] This figure shows an example of improvement information. [Figure 3C] This figure shows an example of reference text information. [Figure 3D] This figure shows an example of enhanced information. [Figure 4] This is a flowchart showing the processing flow of an information processing device. [Figure 5] This figure shows an example of a generated prompt. [Figure 6A] It is a diagram showing an example of a previously distributed advertisement example. [Figure 6B] It is a diagram showing an advertisement example generated as the next advertisement. [Figure 7] It is a diagram showing an example of the hardware configuration of an information processing apparatus, a knowledge server, and a generation apparatus.

Embodiments for Carrying Out the Invention

[0009] Hereinafter, embodiments according to one aspect of the present disclosure will be described with reference to the drawings. Note that the embodiments described below are examples, and the embodiments to which the present disclosure is applied are not limited to the following embodiments.

[0010] FIG. 1 is a diagram showing an example of an advertisement generation system 10 according to an embodiment of the present disclosure. The advertisement generation system 10 includes an information processing apparatus 100, a knowledge server 200, and a generation apparatus 300. The information processing apparatus 100 and the knowledge server 200 are communicable. The information processing apparatus 100 and the generation apparatus 300 are communicable.

[0011] The information processing apparatus 100 is an example of an information processing apparatus operated by an operator who causes a generative AI model to generate an advertisement, and is, for example, a personal computer, a smartphone, or a tablet.

[0012] The information processing apparatus 100 includes an acquisition unit 101, a policy information determination unit 102, a content determination unit 103, a prompt generation unit 104, a prompt transmission unit 105, an advertisement reception unit 106, and an advertisement distribution unit 107.

[0013] The acquisition unit 101 acquires appeal information indicating the degree of appeal of the distributed advertisement to users from the knowledge server 200. The appeal information is composed of items related to the display of the advertisement, items related to the purchase of goods or the conclusion of services, and values corresponding to each item. In the following description, the above items are expressed as "feedback items", and the values corresponding to the items are expressed as "feedback values". The feedback value is not limited to numerical values, and there are also texts such as character strings (for example, "The advertisement has a good reputation", "The advertisement is popular among young people", etc.). Also, the feedback value is an example of the degree of appeal.

[0014] Among the feedback items related to the display of the former advertisement, there are those indicating the number of times the advertisement was displayed, those indicating the number of users who viewed the advertisement, those indicating the number of times the advertisement was opened when distributed by email or short message, or those indicating the number of times the page containing the advertisement was displayed.

[0015] In the following description, the feedback item indicating "the number of times the advertisement was displayed" may be expressed as "IMP". The feedback item indicating "the number of users who viewed the advertisement" may be expressed as "reach". The feedback item indicating "the number of times the advertisement was opened when distributed by email or short message" may be expressed as "opened". The feedback item indicating "the number of times the page containing the advertisement was displayed" may be expressed as "PV".

[0016] Among the feedback items related to the purchase of goods or the conclusion of services of the latter, there are those indicating the number of goods purchased, those indicating the number of service contracts, those indicating the purchase rate or contract rate, those indicating the ratio of users who purchased or concluded a service among the users who viewed the advertisement, those indicating the change in sales before and after the advertisement, those indicating the change in the number of sales or contracts before and after the advertisement, or those indicating the evaluation result of users on the advertisement for the goods or services targeted by the advertisement.

[0017] In the following explanation, feedback items indicating "number of product purchases" may be referred to as "number of purchases." Feedback items indicating "number of service contracts" may be referred to as "number of contracts." Feedback items indicating "change in sales or contract numbers before and after advertising" may be referred to as "change." Feedback items indicating "user evaluation results" may be referred to as "evaluation."

[0018] The policy information determination unit 102 determines policy information indicating the degree to which the delivered advertisement should be improved, in accordance with the acquired appeal information. Here, the "policy information" is indicated in three stages in this embodiment: "Improvement Required (High)", "Improvement Required (Normal)", and "Improvement Required (Low)", but it is not limited to three stages, and may be two stages or four or more stages. Furthermore, "Improvement Required (Low)" also includes a policy of not changing the content of the delivered advertisement and maintaining the advertisement content as it is.

[0019] The policy information determination unit 102 determines policy information according to the conditions for associating the degree of appeal indicated by the appeal information with the policy information. For example, the lower the degree of appeal of the delivered advertisement, the greater the need for improvement, and therefore the policy information determined to be "Improvement Required (High)". Conversely, the higher the degree of appeal of the delivered advertisement, the less the need for improvement, and therefore the policy information determined to be "Improvement Required (Low)". Details of the method for determining policy information will be explained in the flowchart described later.

[0020] The content determination unit 103 determines the content of the next advertisement according to the determined policy information. Specifically, the content determination unit 103 determines the content of the next advertisement using the content of past advertisements, the content of improved advertisements, and improvement information that shows the policy information of past advertisements. Details of the content determination method will be explained in the flowchart below.

[0021] The prompt generation unit 104 generates a prompt that includes reference text, which the generation AI model will use as a reference when generating the next advertisement, according to the determined content of the next advertisement. Specifically, the prompt generation unit 104 generates a prompt using reference text information, which associates the content of the advertisement with policy information and reference text. Details of the prompt generation method will be explained in the flowchart below.

[0022] The prompt generation unit 104 uses RAG (Retrieval-Augmented Generation), a type of prompt augmentation technology, to generate prompts. Here, a prompt is an instruction or question that a user inputs when interacting with a generative AI model such as a Large Language Model (LLM). A common application of RAG is when a user searches for similar documents in a search system beforehand and requests those documents together with the LLM. This allows the LLM to generate answers based on specific documents, such as company information.

[0023] The prompt transmission unit 105 transmits the prompt generated by the prompt generation unit 104 to the generation device 300. The advertisement receiving unit 106 receives the advertisement generated by the generation AI model based on the transmitted prompt from the generation device 300. The advertisement distribution unit 107 provides the received advertisement to, for example, an advertisement distribution server. As a result, the advertisement is delivered to the user or displayed in the advertising section of a social networking service or web page.

[0024] Next, the generation device 300 will be described. The generation device 300 includes a prompt receiving unit 301, a generated AI model 302, and an advertisement transmission unit 303.

[0025] The prompt receiving unit 301 receives the prompt sent by the information processing device 100 and inputs it to the AI ​​model 302. The AI ​​model 302 functions as a generating AI model and outputs an advertisement corresponding to the prompt. The advertisement sending unit 303 sends the advertisement output by the AI ​​model 302 to the information processing device 100.

[0026] Next, the knowledge server 200 will be described. The knowledge server 200 includes a memory unit 202. The memory unit 202 stores previous advertisement information 211, condition information 212, improvement information 213, reference text information 214, enhancement information 215, and product information 216.

[0027] Each of these pieces of information will be explained. Figure 2A shows an example of the structure of the previous advertisement information 211. The previous advertisement information 211 includes "appeal information," "previous content," and "delivery method." Of these, "appeal information" consists of "feedback items" and "feedback values," as described above. Note that the previous content is not limited to the previous advertisement, but can be any content delivered in the past, and may be content from multiple deliveries prior to the previous one.

[0028] The feedback items include, for example, "IMP" as shown above, and the corresponding feedback value is "the number of times the ad was displayed." "Previous content" indicates the content of the previous ad that was delivered. "Previous content" can include text such as "Receive a luxurious prize when you purchase XX cosmetics for 10,000 yen or more," as well as ad images (videos). "Delivery method" indicates how the ad will be delivered, and examples include email, short message, social media, magazines, TV commercials, media, Instagram, XX's web page, etc. This "Delivery method" may also include information such as the ad's file format and file size.

[0029] Figure 2B shows an example of the configuration of condition information 212. The condition information includes "feedback items," "policy conditions," and "policy information." Of these, "policy conditions" are the conditions that associate the feedback values ​​corresponding to the feedback items with the policy information.

[0030] Figure 2C shows an example of the structure of improvement information 213. The improvement information includes "feedback items," "previous content," "application conditions," "policy information," "next content," and "delivery method." Of these, "feedback items" and "previous content" are as described above. "Application conditions" indicate the conditions for determining whether the advertisement to be improved is applicable to the content shown in "next content." "Next content" indicates the content of the next advertisement to be delivered. "Delivery method" indicates the delivery method for the next advertisement to be delivered.

[0031] Figure 2D shows an example of the structure of Reference Text Information 215. Reference Text Information 215 includes "Enhancement Items," "Next Content," and "Reference Text." "Enhancement Items" indicate the feedback items to be enhanced through improvement. "Next Content" is as described above. "Reference Text" is, as described above, the reference text used by the generation AI model when generating the advertising copy for the next advertisement.

[0032] Figure 2E shows an example of the configuration of the enhancement information 216. The enhancement information 216 is provided in response to feedback items and is an instruction to the generative AI model to improve the feedback value corresponding to the feedback item. The enhancement information 216 is divided into two parts: advertising enhancement information and purchase contract enhancement information. The advertising enhancement information includes "IMP enhancement information," "reach enhancement information," "open enhancement information," and "PV enhancement information." The purchase contract enhancement information includes "purchase quantity enhancement information," "contract quantity enhancement information," "change enhancement information," and "evaluation enhancement information."

[0033] Figure 2F shows an example of the structure of product / service information 217. Product / service information 217 includes "Description," "PR Information," "Campaign Information," "Company Information," and "Advertising Distribution Period." "Description" shows a description of the product or service. "PR Information" shows the content of public relations for the product or service. "Campaign Information" shows information about a series of promotions aimed at promoting the sale of the product or service. "Company Information" shows information about the company that provides the product or service. "Advertising Distribution Period" shows the period during which the advertisement will be distributed.

[0034] Next, specific examples of the above-mentioned information types—conditional information, improvement information, reference text information, and reinforcement information—are shown. Figure 3A shows an example of conditional information. In the example shown in Figure 3A, the feedback item is IMP, and the condition is defined using the number of times the advertisement was displayed per day. If the number of times the advertisement was displayed per day is greater than 1000, the policy information is determined to be "No improvement needed (low)." On the other hand, if the number of times the advertisement was displayed per day is 500 or less, the policy information is determined to be "Improvement needed (high)." In the example shown in Figure 3A, only the feedback item IMP is shown, but other feedback items and their corresponding conditions also exist in the conditional information. Furthermore, conditions do not necessarily use numerical values; for example, some use text. Specifically, if the feedback value is expressed in text, such as "The advertisement has a good reputation," an example of a condition might be whether or not the feedback value contains words such as "good."

[0035] Figure 3B shows an example of improvement information. In the example shown in Figure 3A, for example, if the enhancement item for the advertisement to be improved (hereinafter sometimes referred to as the "advertisement to be improved") is IMP, and the advertisement to be improved has already had "High Appeal for Value" applied to it in the previous step, and the number of times the advertisement is displayed per day is greater than 1000, and the determined policy information is "No Improvement Needed (Low)", then the next content will be "High Appeal for Experience (Normal) + High Appeal for Value (Normal) (Experience is conveyed through video)". Note that (High) and (Normal) shown in Figure 3B are abbreviations for "No Improvement Needed (High)" and "Improvement Needed (Normal)", respectively. Also, as shown in "High Appeal for Experience (Normal) + High Appeal for Value (Normal) (Experience is conveyed through video)", multiple elements ("Experience", "Value") may be included in the next content.

[0036] Furthermore, "Distribution Method" can be "Same as Last Time," which indicates that the previous distribution method will be used again. For example, "Add Magazine" means that the previous distribution method will be used again, and that the distribution will be done using "Magazine." Also, if "Distribution Method" is simply "Magazine" instead of "Add Magazine," it indicates that the distribution using the previous method will be discontinued and distribution will be done using "Magazine." Note that the distribution method may indicate multiple methods, such as "Magazine" and "SNS."

[0037] Figure 3C shows an example of sample text information. In the example shown in Figure 3C, when the "Enhancement Item" is IMP and the "Next Content" is "Emphasize Value (Normal)", the "Sample Text" is "Buy Now and Get 〇 Points!". Also, when the "Emphasize Value (High)" is selected, the "Sample Text" is "Buy Now and Get 〇 Points! Plus, get □% cashback for every 1000 yen spent. Earn 〇 points for a year by spending 〇 yen per month!". Compared to the "Needs Improvement (Normal)" case, this text is more appealing and helps generate a more effective advertisement.

[0038] Figure 3D shows an example of enhancement information. In Figure 3D, when the "enhancement item" is "open," the instruction to the generative AI model to improve the feedback value corresponding to the feedback item is, "Please make the subject line concise so that the content of the ad can be roughly understood." This is because users who receive ads often decide whether or not to open them based on the subject line.

[0039] Figure 4 is a flowchart showing the processing flow of the information processing device 100. In Figure 4, the acquisition unit 101 acquires appeal information by acquiring previous advertising information from the knowledge server 200 (step S101). Next, the policy information determination unit 102 determines policy information indicating the degree to which the delivered advertisement should be improved, according to the acquired appeal information (step S102). Specifically, the policy information determination unit 102 acquires feedback items and feedback values ​​from the acquired appeal information and determines the policy information by referring to the condition information. For example, if the condition information is as shown in Figure 3A, the feedback item acquired by the policy information determination unit 102 is IMP, and the feedback value is 600, the policy information determination unit 102 determines "Improvement Required (Normal)" as the policy information.

[0040] The content determination unit 103 determines the content of the next advertisement according to the determined policy information (step S103). For example, if the improvement information is the content shown in Figure 3B, the previous content shown in the previous advertisement information is, for example, "Highly emphasize functionality", the feedback item is IMP as before, the feedback value is 600, and the determined policy information is "Needs improvement (normal)", then this corresponds to the second line of the example improvement information shown in Figure 3B, and the content determination unit 103 determines "Highly emphasize value" as the content for the next advertisement.

[0041] The prompt generation unit 104 generates a prompt that includes reference text for the generation AI model to use as a reference when generating the next advertisement, according to the content of the next advertisement that has been determined (step S104). For example, if the reference text information is as shown in Figure 3C, the enhancement item is IMP, and the determined content for the next advertisement is "Highly appealing to value", then this corresponds to the second line of the example reference text information shown in Figure 3C, and the prompt generation unit 104 generates a prompt that includes the reference text, "Buy now and receive ○ points! Plus, get □% back for every 1000 yen spent. Earn ○ points for a year by spending ○ yen per month!"

[0042] In addition to this reference text, the prompt generation unit 104 generates prompts that include information about previously delivered advertisements (previous content, delivery method, previous policy). Furthermore, depending on the enhancement item, the prompt generation unit 104 generates prompts that include instructions to the generating AI model to improve the feedback value. For example, if the enhancement information is as shown in Figure 3D and the enhancement item is IMP, the prompt generation unit 104 generates a prompt that includes "Please make the subject line concise so that the content of the advertisement can be roughly understood." The prompt generation unit 104 also obtains product and service information from the knowledge server 200 and generates prompts that include information that constitutes the product and service information ("Description," "PR Information," "Campaign Information," "Company Information," and "Advertising Delivery Period").

[0043] The information processing device 100 sends the generated prompt to the generation device 300 (step S105). The generation device 300 gives the received prompt to the generation AI model 302, and the generation AI model 302 generates an advertisement. The generation device 300 sends the generated advertisement to the information processing device 100. The information processing device 100 receives the transmitted advertisement received from the generation device 300 (step S106). The received advertisement is provided to, for example, a distribution server that distributes advertisements, and is then delivered to users or displayed in the advertising section of social media or web pages.

[0044] Figure 5 shows an example of a generated prompt. As shown in Figure 5, prompt 400 consists of previous advertisement information 401 and next advertisement instruction 402. The following strings written in prompt 400 other than the previous advertisement information 401 and next advertisement information 402 are standard phrases. "role" "You are an advertising creator." "task" "Please generate an ad based on the following conditions." "Please create an ad based on the revised ad creation policy as follows."

[0045] The previous ad information 401 includes "Previous Content," "Delivery Method," and "Policy." "Previous Content" indicates the content of the previously delivered ad. In Figure 5, the content of the previous ad was "You can also use it for mobile payments." "Delivery Method" indicates the delivery method of the previously delivered ad. In Figure 5, the delivery method of the previous ad was "SNS." "Policy" indicates the policy used when generating the previously delivered ad. In Figure 5, the policy for the previous ad was "Highlight the features."

[0046] The next ad instruction 402 includes "Delivery Method," "Policy," "Reference Text," "Enhancement Items," "Enhancement Information," and "Product Information." "Delivery Method" indicates how the next ad will be delivered. In Figure 5, the delivery method for the next ad is "Email." "Policy" indicates the policy for generating the next ad. In Figure 5, the policy for the next ad is "Emphasis on value."

[0047] As mentioned above, the "reference text" is the text that the generating AI model uses as a reference when generating the next advertisement. In Figure 5, the reference text is "Buy now and receive 〇 points!". The "enhancement item" indicates the feedback item to be enhanced through improvement. In Figure 5, the enhancement item is "open". The "enhancement information" is the instruction to the generating AI model to improve the feedback value corresponding to the feedback item. In Figure 5, the enhancement information is "Please make the subject line concise so that the content of the advertisement can be roughly understood". The "product and service information" is the product and service information obtained from the knowledge server 200. In Figure 5, the instruction is given to make the advertisement include the "description" and "PR information" from the product and service information.

[0048] Figure 6A shows an example of a previously delivered advertisement. Figure 6B shows an example of an advertisement generated for the next advertisement. The advertisement example shown in Figure 6A indicates that product XX is on sale, that points are doubled, that it can be used for mobile payments, and the date and venue of the event.

[0049] In contrast to the previously delivered ad example, the ad example 500 shown in Figure 6B was generated by the AI ​​model 302 using the prompt 400 described above. The subject line 500 is written in accordance with the prompt 400's "Enhancement Item" ("Open") and "Enhancement Information" ("The subject line should be concise so that the user can roughly understand the content of the ad"), and is designed to strongly encourage the user to open the ad.

[0050] Similarly, ad copy 502 follows the "Policy" of prompt 400, which is to "emphasize the value," and the "Reference Text," which is to "Get X points when you buy now!" Furthermore, ad copy 503 also follows the "Product / Service Information" of prompt 400, and includes "Description" and "PR Information."

[0051] <Effects> As explained above, this embodiment makes it possible to generate prompts to create advertisements that are more effective than those already delivered. Specifically, for example, Patent Document 1 discloses changing the design of advertising policies, but does not describe the details of the advertising policies. On the other hand, this embodiment determines policy information indicating the degree to which delivered advertisements should be improved, depending on the degree to which delivered advertisements appeal to users. As a result, if delivered advertisements are not very effective, the degree of improvement can be increased, and if delivered advertisements are effective, the degree of improvement can be decreased, making use of delivered advertisements. Therefore, regardless of the degree of effectiveness of delivered advertisements, it becomes possible to generate prompts to create advertisements that are more effective than those already delivered.

[0052] <Example 1> Modification 1 describes an embodiment using feature information that indicates the characteristics, trends, or requirements that an advertisement should meet. First, an example of the above feature information will be described.

[0053] (Example of feature information 1) The feature information may also be information that indicates the characteristics (or tendencies) of Japanese expression. For example, it may be information that indicates characteristics such as politeness, cheerfulness, energy, or cuteness. For example, specific examples of "politeness" may include characteristics of using Japanese expression such as "using the 'desu' and 'masu' forms as polite language" or "using honorific language when speaking to customers." By generating a prompt containing such feature information that indicates the characteristics of Japanese expression that should be noted, the prompt generation unit 104 can convey the characteristics of Japanese expression that should be noted to the generating AI model 302.

[0054] (Example of feature information 2) The characteristic information may also be information indicating the level of detail in the product description. For example, it could be information that describes the description in any way, such as "explain in detail" or "explain briefly," or it could be an example sentence indicating the degree of detail.

[0055] For example, an example sentence for "a simple explanation" could be "○○ is a communication plan," and an example sentence for "a detailed explanation" could be "○○ is a simple one-plan that allows you to use a generous 30GB for a monthly fee of 2,970 yen (tax included)..."

[0056] By generating prompts containing feature information indicating the level of detail in the product information description using the prompt generation unit 104, the AI ​​model 302 can be informed of the features indicating the level of detail in the product information description. Specifically, subjective expressions such as "detailed" or "easy to understand" can be conveyed in a way that the AI ​​model can understand.

[0057] Furthermore, you may specify what needs to be explained in detail, linking it to the characteristic of detail itself. For example, you could describe the characteristic as "explain the product's performance simply" or "explain how to use the product carefully."

[0058] (Example of feature information 3) In addition to the feature information examples 1 and 2 above, the feature information may also include any or a combination of the following: (a) number of characters, (b) whether or not emojis are used or their usage rate / degree, (c) regional dialects (Kansai dialect, Tohoku dialect, Kyushu dialect, etc.), (d) language (English, German, Japanese, etc.), (e) additional information such as warnings / notes / remarks / supplementary information based on laws, etc., and descriptions guiding users to website links for more detailed information.

[0059] Next, we will explain the process of generating effective advertisements using the feature information described above. First, the following metadata is stored in the previous advertisement information 211, corresponding to the previous content. • Target user attributes (age, gender, hobbies, places they frequent, things they often buy, income, family, education level, etc.) • Attributes of the product campaign to be distributed (product category, price, online / physical sales, genre, brand, country of manufacture, campaign type and content such as discount / free / point rewards, etc.) • Characteristics / trends / requirements set during generation • Distribution results (If distribution results by user attributes are available, please include those results. For example, information showing the percentage increase in sales, such as "sales increased among men in their 20s after distribution," is also acceptable.) As mentioned above, the "previous content" does not have to be limited to the previous broadcast; any content previously broadcast is acceptable, and it doesn't have to be the content from the immediate preceding broadcast, but can be from several broadcasts prior to the previous one.

[0060] In this modified example, it is assumed that the previous ad information 211 stores user attributes and the results of previously delivered ads (sales increase rate). Based on the characteristic information and metadata, the system delivers ads with high similarity / commonality to the target user attributes (target product campaign attributes) of the next ad.

[0061] Therefore, the information processing device 100 searches for past advertisements delivered to a target audience that matches the target audience for the next advertisement, and uses the characteristic information that indicates the features, trends, or requirements set when generating the advertisement that resulted in a predetermined delivery outcome from the searched advertisements as the characteristic information that the next advertisement should satisfy.

[0062] Specifically, the information processing device 100 searches the knowledge server 200 for past advertisements that match the user attributes of the next advertisement (S1). The information processing device 100 selects the content of the advertisements that have been delivered in the past and whose delivery results (sales increase rate) have been determined to be a predetermined delivery result (for example, exceeding a predetermined threshold) from among the advertisements that match the user attributes, etc., and sets them as the "previous content" to be used as a reference for the next advertisement (S2). The information processing device 100 sets the feature information (information that describes features of Japanese expression, information that indicates the thoroughness of the product description, etc.) that indicates the features, trends, or requirements set at the time of generation corresponding to the "previous content" as the "features, trends, and requirements that the advertisement should satisfy" when generating the next advertisement (S3). Then, the prompt generation unit 104 generates a prompt that includes this feature information, causing the generation AI model 302 to generate an advertisement that reflects the feature information.

[0063] Note that the above example searches for cases with matching user attributes, but this is not limited to this. Also, although the search is performed for matching cases, user attributes may be information expressed in text. For example, it could be "a working adult in his 20s who likes ●●". In this case, a text similarity search (a publicly known search tool) may be performed and similar "previous content" may be selected. Furthermore, there may be multiple pieces of characteristic information that satisfy the above (S1) to (S3). In this case, the information processing device 100 may select only the top predetermined number of delivery results (e.g., sales growth rate).

[0064] By doing so, the prompt generation unit 104 can generate a prompt that includes the following information. • Use polite language, such as the "desu / masu" form. • Use respectful language when speaking to customers. • Briefly explain the product's performance. • Carefully explain how to use the product.

[0065] These prompts enable AI model 302 to generate effective advertisements using feature information that takes into account the previous content.

[0066] The information processing device 100 performs the above processes (S1) to (S3) independently of the processes S101 to S103 shown in Figure 4, and the prompt generation unit 104 generates a prompt using the results of the processes S101 to S103 and the characteristic information that the next advertisement should satisfy. If the results of the processes S101 to S103 and the characteristic information that the next advertisement should satisfy contradict each other, one of them is prioritized. Which one to prioritize can be set in advance, or it can be selected by the user each time.

[0067] <Modification 2> In the embodiment described above, the information processing device 100 retrieved various information from the knowledge server 200 to generate recommendation information. However, the information processing device 100 may also store the various information stored in the knowledge server 200 in its own storage device. This configuration can be achieved by changing the search destination of the information processing device 100 from the knowledge server 200 to its own storage device.

[0068] Furthermore, in the system configuration of the advertising generation system 10 described above, the generation AI model 302 is implemented in a device connected to the information processing device 100 via a different network, but it may also be implemented in the information processing device 100. Also, the generation unit 107 that generates prompts using RAG is implemented in the information processing device 100, but it may also be implemented in the device on which the generation AI model is implemented, or in a device other than the device on which the generation AI model is implemented. In addition, the various information stored in the memory unit of the knowledge server 200 may be stored in the information processing device 100 or the generation device 300.

[0069] Furthermore, each functional unit of the information processing device 100 and the generation device 300 may be implemented in a distributed manner on the cloud. Also, each functional unit may be implemented in multiple information processing devices. Furthermore, the same functional unit may be realized by multiple information processing devices. The various types of information stored in the memory unit 202 of the knowledge server 200 may be implemented in a distributed manner on the cloud.

[0070] <Example Hardware Configuration> The information processing device 100, knowledge server 200, and generation device 300 in one embodiment of the present disclosure may function as a computer that processes the wireless communication method of the present disclosure. Figure 7 is a diagram showing an example of the hardware configuration of the information processing device 100, knowledge server 200, and generation device 300 according to one embodiment of the present disclosure. The above-described information processing device 100, knowledge server 200, and generation device 300 may be physically configured as a computer device including a processor 1001, memory 1002, storage 1003, communication device 1004, input device 1005, output device 1006, bus 1007, etc.

[0071] In the following explanation, the term "device" can be replaced with "circuit," "device," "unit," etc. The hardware configuration of the information processing device 100, the knowledge server 200, and the generation device 300 may include one or more of the devices shown in the figure, or it may be configured to omit some of the devices.

[0072] The functions of the information processing device 100, the knowledge server 200, and the generation device 300 are realized by loading predetermined software (programs) onto hardware such as the processor 1001 and memory 1002, which allows the processor 1001 to perform calculations, control communication by the communication device 1004, and control at least one of data reading and writing in the memory 1002 and storage 1003.

[0073] The processor 1001 controls the entire computer, for example, by running an operating system. The processor 1001 may consist of a central processing unit (CPU) that includes interfaces with peripheral devices, control units, arithmetic units, registers, and so on.

[0074] Furthermore, the processor 1001 reads programs (program code), software modules, data, etc., from at least one of the storage 1003 and the communication device 1004 into the memory 1002 and executes various processes accordingly. The program used is one that causes the computer to execute at least a part of the operations described in the above embodiment. For example, each part of the information processing device 100 may be stored in the memory 1002 and implemented by a control program that runs on the processor 1001, and other functional blocks may be implemented similarly. The above-described various processes have been explained as being executed by one processor 1001, but they may be executed simultaneously or sequentially by two or more processors 1001. The processor 1001 may be implemented by one or more chips. The program may also be transmitted from a network via a telecommunications line.

[0075] Memory 1002 is a computer-readable recording medium and may consist of at least one of the following: ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), RAM (Random Access Memory), etc. Memory 1002 may also be called a register, cache, main memory, etc. Memory 1002 can store executable programs (program code), software modules, etc., for carrying out a wireless communication method according to one embodiment of the present disclosure.

[0076] Storage 1003 is a computer-readable recording medium and may consist of at least one of the following: an optical disc such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., a compact disc, a digital multipurpose disc, a Blu-ray® disc), a smart card, flash memory (e.g., a card, a stick, a key drive), a floppy® disk, a magnetic strip, etc. Storage 1003 may also be called an auxiliary storage device. The above-mentioned storage medium may be, for example, a database, server, or other suitable medium including at least one of memory 1002 and storage 1003.

[0077] The communication device 1004 is hardware (transceiver / receiver device) for communicating between computers via at least one of a wired network and a wireless network, and is also referred to as a network device, network controller, network card, communication module, etc.

[0078] The input device 1005 is an input device that accepts input from an external source (e.g., a keyboard, mouse, microphone, switch, button, sensor, etc.). The output device 1006 is an output device that outputs to an external source (e.g., a display, speaker, LED lamp, etc.). The input device 1005 and the output device 1006 may be configured as an integrated unit (e.g., a touch panel).

[0079] Furthermore, each device, such as the processor 1001 and memory 1002, is connected by a bus 1007 for communicating information. The bus 1007 may be configured using a single bus, or different buses may be configured for each device.

[0080] Furthermore, the information processing device 100, the knowledge server 200, and the generation device 300 may include hardware such as a microprocessor, a digital signal processor (DSP), an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), and an FPGA (Field Programmable Gate Array), and some or all of each functional block may be realized by such hardware. For example, the processor 1001 may be implemented using at least one of these hardware components. [Industrial applicability]

[0081] One aspect of this disclosure is useful for marketing. [Explanation of Symbols]

[0082] 10 Ad Generation System 100 Information Processing Devices 101 Acquisition Department 102 Policy and Information Decision-Making Department 103 Content Determination Department 104 Prompt generation unit 200 Knowledge Servers 300 generator 302 Generative AI Models 1001 Processor 1002 memory 1003 Storage 1004 Communication device 1005 Input device 1006 Output device 1007 Bus

Claims

1. An acquisition unit that acquires appeal information indicating the degree to which delivered advertisements appealed to users, A policy information determination unit determines policy information indicating the degree to which the delivered advertisement should be improved, based on the acquired appeal information. A content determination unit determines the content of the next advertisement in accordance with the determined policy information, A prompt generation unit generates a prompt that includes reference text to be used by the generation AI model when generating the next advertisement, based on the content of the next advertisement that has been determined. Equipped with an information processing device.

2. The information processing apparatus according to claim 1, wherein the appeal information is information indicating the number of times the advertisement was displayed, information indicating the number of users who saw the advertisement, information indicating the number of times the advertisement was opened when delivered via email or short message, information indicating the number of times the page containing the advertisement was displayed, information indicating the number of product purchases, information indicating the number of service contracts, information indicating the purchase rate or contract rate, information indicating the percentage of users who viewed the advertisement and made a purchase or contracted a service, information on the change in sales before and after the advertisement, information on the change in the number of sales or contracts before and after the advertisement, or information indicating the user's evaluation result of the advertisement for the advertised product or service.

3. The information processing apparatus according to claim 1, wherein the policy information determination unit determines the policy information according to policy conditions that associate the degree of appeal indicated by the appeal information with the policy information.

4. The information processing apparatus according to claim 1, wherein the content determination unit determines the content of the next advertisement using the content of past advertisements, the content of an improved advertisement based on the past advertisements, and improvement information indicating the policy information of past advertisements.

5. The information processing apparatus according to claim 1, wherein the prompt generation unit generates the prompt using reference document information, in which the content of the advertisement, the policy information, and the reference document are associated.

6. The system searches for past advertisements delivered to the same target audience as the next advertisement, and from the searched advertisements, it identifies the characteristic information that indicates the features, trends, or requirements set when generating the advertisement that achieved the predetermined delivery result, and uses this characteristic information as the characteristic information that the next advertisement should satisfy. The information processing apparatus according to claim 1, wherein the prompt generation unit generates a prompt using the characteristic information that the next advertisement should satisfy.

7. Information processing device, We obtain appeal information that shows the degree to which delivered ads appealed to users. Based on the acquired appeal information, policy information indicating the degree to which the delivered advertisement should be improved is determined. Based on the policy information decided above, the content of the next advertisement will be determined. Based on the determined content of the next advertisement, the generation AI model generates a prompt that includes reference text to be used when generating the next advertisement. Control method.

8. On the computer, We obtain appeal information that shows the degree to which delivered ads appealed to users. Based on the acquired appeal information, policy information indicating the degree to which the delivered advertisement should be improved is determined. Based on the policy information decided above, the content of the next advertisement will be determined. Based on the determined content of the next advertisement, the generation AI model generates a prompt that includes reference text to be used when generating the next advertisement. A program to execute a process.

Images