Information processing device, information processing method, and program

Abstract

[Problem] To enable generation of more effective advertisements. [Solution] An information processing device comprising a control unit that edits a tag group obtained by converting advertisement content to text, extracts a derivable element from the advertisement content, and generates, by deriving the corresponding element according to the edited tag group, derived advertisement content based on the advertisement content.

Description

Information processing device, information processing method, and program

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

[0002] In recent years, technology for generating image generation models using machine learning has become widespread. When performing machine learning, it is necessary to input a large amount of training data into the generated model.

[0003] For example, Patent Document 1 below discloses a technique for generating training data using a CG (Computer Graphics) model.

[0004] International Publication No. 2021 / 177324

[0005] Various forms of advertising have traditionally been used to publicize a company's products and services. While these advertisements require constant updating to continually capture consumer interest, generating them takes time and effort. While it is conceivable to automatically generate advertisements using the image generation model described above, this approach is still insufficient to generate more effective advertisements.

[0006] Therefore, the present disclosure proposes an information processing device, an information processing method, and a program that are capable of generating more effective advertisements.

[0007] According to the present disclosure, an information processing device is provided that includes a control unit that edits a group of tags obtained by converting advertising content into text, extracts derivable elements from the advertising content, and derives the corresponding elements according to the edited group of tags, thereby generating derived advertising content based on the advertising content.

[0008] The present disclosure also provides an information processing method including: a processor editing a group of tags obtained by converting advertising content into text; extracting derivable elements from the advertising content; and generating derived advertising content based on the advertising content by deriving the corresponding elements according to the edited group of tags.

[0009] Furthermore, according to the present disclosure, there is provided a program that causes a computer to function as a control unit that edits a group of tags obtained by converting advertising content into text, extracts derivable elements from the advertising content, and derives the corresponding elements according to the edited group of tags, thereby generating derived advertising content based on the advertising content.

[0010] FIG. 1 is an overall configuration diagram of an information processing system 1 according to an embodiment of the present disclosure. FIG. 2 is a block diagram showing an example of the configuration of a server 20 according to this embodiment. FIG. 3 is a diagram for explaining an example of an advertisement generation cycle according to this embodiment. FIG. 4 is a diagram for explaining details of a prompt according to this embodiment. FIG. 5 is a diagram for explaining extraction of a derivable element from an advertisement image 500 that is an original image. FIG. 6 is a diagram showing an example of a derived advertisement image. FIG. 7 is a flowchart showing a first example of a flow of a process for generating a derived advertisement image in the information processing system 1. FIG. 8 is a flowchart showing a second example of a flow of a process for generating a derived advertisement image in the information processing system 1.

[0011] Preferred embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. In this specification and drawings, components having substantially the same functional configurations are designated by the same reference numerals, and redundant description will be omitted.

[0012] The explanation will be given in the following order: 1. Overview 2. Configuration of the server 20 3. Advertisement generation 3-1. Advertisement generation cycle 3-2. Prompt generation patterns 3-3. Others 4. Operational processing 4-1. First example 4-2. Second example 4-3. Combinations 5. Supplementary information

[0013] <<1. Overview>> An information processing system according to an embodiment of the present disclosure will be specifically described with reference to the drawings.

[0014] Fig. 1 is an overall configuration diagram of an information processing system 1 according to an embodiment of the present disclosure. As shown in Fig. 1, the information processing system 1 includes user terminals 10 (10a to 10n), a server 20, and an advertiser terminal 30. The user terminals 10 and the advertiser terminal 30 can be connected to the server 20 via a network 40.

[0015] The user terminal 10 is an information processing terminal used by consumers (referred to herein as a user). The user terminal 10 may be realized by a smartphone, a tablet terminal, a personal computer (PC), a head-mounted display (HMD), a wearable device, or the like. Various forms of advertisements may be provided on the user terminal 10. Examples of "advertisements" include digital advertisements posted on web media, social networking sites (SNS), emails, games, or applications. When a user performs an operation such as clicking on such a digital advertisement image displayed on the screen of the user terminal 10, operation information (reaction information to the advertisement) is transmitted to a server 20 or the like. The provision of digital advertisements is not limited to the user terminal 10, but may also be provided on electronic bulletin boards installed in various facilities, on trains, on the street, in convenience stores, or via television broadcasts. Digital advertisements may be still images or videos. Videos may be accompanied by audio. Furthermore, in this specification, "advertisements" are not limited to images and may also refer to audio.

[0016] The advertiser terminal 30 is an information processing terminal used by the advertising subject (referred to as the advertiser in this specification), such as a company, a local government, or an individual. The advertiser terminal 30 can be realized by a smartphone, a tablet terminal, a PC, an HMD, a wearable device, or the like. The advertiser terminal 30 transmits information regarding advertisement generation input by the advertiser to the server 20.

[0017] The server 20 has a function of generating advertisements and providing them to the user terminal 10 as appropriate. The server 20 may be realized by an information processing device, or may be realized as an advertisement system on a network. Furthermore, the function of the server 20 may be part of an advertisement system on a network. The advertisement system may be realized by multiple servers (information processing devices), or may be realized by a single server.

[0018] The server 20 may generate an advertising image (an example of advertising content) using, for example, an image generation model and provide it to the user terminal 10. An example of an image generation model is an image generation AI (Artificial Intelligence) equipped with an AI model (Text-to-Image model) that generates an image from an input text-based request. The server 20 can generate an advertising image using the image generation AI in response to a text-based request transmitted from the advertiser terminal 30 as a request from the advertiser.

[0019] The image generation AI may be prepared in advance or may be generated by the server 20. The image generation AI may be constructed by performing machine learning using a dataset generated from still image data of thousands to tens of thousands of real-life human models. Alternatively, the image generation AI may be constructed by a method (so-called fine tuning) in which an already trained text-to-image model is additionally trained with the dataset.

[0020] The configuration of the information processing system 1 according to this embodiment has been described above. Note that the system configuration shown in FIG. 1 is an example, and this embodiment is not limited to this. For example, while FIG. 1 illustrates the advertiser terminal 30, the information processing system 1 may include an agent terminal used by an advertising agent, such as an advertising agency, that performs advertising business at the request of an advertiser, instead of or in addition to the advertiser terminal 30. The advertising agent terminal transmits information related to advertisement generation input by the advertising agent (such as the above-mentioned text-based request) to the server 20.

[0021] Next, the configuration of the server 20 that generates the advertisement will be specifically described.

[0022] 2 is a block diagram showing an example of the configuration of the server 20 according to this embodiment. As shown in FIG. 2, the server 20 includes a communication unit 210, a control unit 220, and a storage unit 230.

[0023] (Communication Unit 210) The communication unit 210 has a transmission unit that transmits data to an external device and a reception unit that receives data from an external device. The communication unit 210 may be communicatively connected to an external device or the Internet using, for example, a wired or wireless LAN (Local Area Network), Wi-Fi (registered trademark), Bluetooth (registered trademark), a mobile communication network (LTE (Long Term Evolution), 4G (fourth generation mobile communication system), 5G (fifth generation mobile communication system)), or the like.

[0024] (Control Unit 220) The control unit 220 functions as an arithmetic processing unit and a control device, and controls the overall operation of the server 20 in accordance with various programs. The control unit 220 is realized by an electronic circuit such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or a microprocessor. The control unit 220 may also include a ROM (Read Only Memory) that stores programs to be used, calculation parameters, etc., and a RAM (Random Access Memory) that temporarily stores parameters that change as appropriate.

[0025] The control unit 220 according to this embodiment can also function as an advertisement generation unit 221 , a text conversion unit 222 , a prompt generation unit 223 , an advertisement effect result analysis unit 224 , and a derivable element extraction unit 225 .

[0026] The advertisement generation unit 221 generates advertisements. The advertisement generation unit 221 generates advertisement images using, for example, image generation AI. "Advertisement images" according to this embodiment include still images and videos. Furthermore, "video" may include audio. The image generation AI is an example of advertisement generation AI, and the advertisement generation unit 221 may further use audio generation AI as the advertisement generation AI. By using the image generation AI and audio generation AI, the advertisement generation unit 221 can generate videos containing audio as advertisements, such as singing videos, music videos (MVs), commercial messages (CMs), and videos with narration. Furthermore, the advertisement generation unit 221 may use audio generation AI to generate advertisement audio that is only audio provided by radio, music apps, audio distribution services, etc.

[0027] As described above, the advertisement generation unit 221 may use an image generation AI equipped with an AI model that generates an image from a text-based request. The text input to the image generation AI, i.e., a so-called prompt, may be obtained from an external source or may be generated by the prompt generation unit 223. The generated advertisement is stored in the advertisement information DB 231. The generated advertisement may be distributed by the server 20, or a distribution server may be provided separately from the server 20, and the advertisement may be distributed from the distribution server.

[0028] The text generation unit 222 analyzes the advertisement image or audio and converts the image or audio into text. For example, the text generation unit 222 uses machine-learned image analysis AI to identify objects appearing in the advertisement image and generate an image description. The image description is generated in the form of a tag group listing elements of the advertisement. The elements of the advertisement include the background (such as the type of scenery, whether indoors or outdoors, etc.) and brightness (such as the time of day, such as morning, noon, evening, or night). For example, when converting an advertisement image of a woman with long hair sitting on a terrace and holding an iced coffee into text, a tag group such as "beautiful Japanese woman, alone, about 20 years old, black hair, long hair, smile, sitting, light blue dress, diamond ring, diamond earrings, pink lipstick, coffee shop, terrace seats, drinking iced coffee, summer, sun, upper body, ultra-high-definition photography" is output. The output image description data (specifically, the tag group) is stored in the advertisement information DB 231.

[0029] The prompt generation unit 223 generates a prompt to be used for generating an advertisement by an image generation AI, etc. The prompt may be in the form of, for example, a tag group that lists elements in the advertisement.

[0030] Various patterns of prompt generation according to this embodiment can be cited. For example, a prompt may be generated to generate an advertisement that reflects the user's preferences, or a prompt may be generated to generate an advertisement according to the advertiser's requests. Furthermore, since continuously providing advertisements that cater to the user's preferences can cause boredom or loss of interest, the prompt generation unit 223 according to this embodiment may generate a prompt to generate an advertisement that is new to the user, has a trendy or seasonal feel, or creates serendipity. Details of prompt generation will be described later.

[0031] The advertising effectiveness result analysis unit 224 analyzes the advertising effectiveness result. The advertising effectiveness result may be obtained from an external server that collects advertising effects, or may be obtained from each user terminal 10. The advertising effectiveness result is, for example, the click-through rate (CTR) of an advertisement provided (distributed, displayed) to the user terminal 10. The click-through rate of an advertisement is calculated by dividing the number of clicks on the advertisement by the number of times it is displayed. The calculation of the click-through rate may be performed by an external server or by the server 20. The advertising effectiveness result may be stored in the advertising information DB 231 as an example of advertising information in association with a corresponding advertising image.

[0032] The advertising effectiveness result analysis unit 224 performs a process of associating such advertising effectiveness results with image description data (tag group), the ID (UID) of the user to whom the advertising results are provided, etc. For example, based on the click rate of an advertising image provided to a certain user, the advertising effectiveness result analysis unit 224 assigns an advertising effectiveness result to each tag of the tag group obtained from the advertising image.

[0033] The advertising effectiveness results for each tag by UID (by user) are stored in the tag information DB 232. The advertising effectiveness result analysis unit 224 can continuously obtain advertising effectiveness results for a large number of advertising images provided to each user and assign advertising effectiveness results for each tag for each user. The advertising effectiveness results associated with tags may be updated with the highest value, or may be updated with the latest value at regular intervals.

[0034] An example of information on the advertising effect results for each tag UID stored in the tag information DB 232 is shown in Table 1 below.

[0035]

[0036] This allows the server 20 to analyze each user's preferred tag preferences.

[0037] The advertising effectiveness result associated with each tag may be the advertising effectiveness result value itself, or may be a value calculated from the advertising effectiveness result value using a predetermined calculation formula. Furthermore, the advertising effectiveness result value associated with each tag may reflect a weighting previously set for each tag.

[0038] In addition, while the above example illustrates an analysis of advertising effectiveness results obtained for each user, the present embodiment is not limited to this example. The advertising effectiveness result analysis unit 224 may obtain advertising effectiveness results for each user attribute and analyze the advertising effectiveness results for each user attribute of each tag. Examples of user attributes include age, gender, region, nationality, family structure, income, and occupation.

[0039] Furthermore, the advertising effectiveness result analysis unit 224 may associate advertising effectiveness results with each tag regardless of the user or attribute.

[0040] The derivable element extraction unit 225 extracts derivable elements from the advertising image or audio. More specifically, for example, the derivable element extraction unit 225 separates objects in the advertising image by segmentation, identifies the separated objects, and tags the separated objects according to the identification results. For example, objects in the advertising image may be tagged with "woman," "hair," "earrings," "necklace," "coffee," "dress," "background," etc. If a part can be separated by segmentation, that part can be changed, i.e., it can be said to be a derivable element. Parts separated by segmentation may also include text parts such as advertising copy.

[0041] The derivable element extraction unit 225 outputs the tagging results for the derivable elements in the advertisement image to the advertisement generation unit 221. The advertisement generation unit 221 can generate a derived advertisement image (an example of derived advertisement content) by changing (i.e., deriving) at least a part of the advertisement image.

[0042] Specific examples of derivable elements and derived content are shown in Table 2 below.

[0043]

[0044] In Table 2, a "general-purpose AI model" is a generative AI model trained on a dataset obtained primarily from facial images of ordinary people. A "specific AI model" is a generative AI model trained on a dataset obtained from facial images of specific human models, such as fashion models, influencers, or celebrities. "Outside the image" refers to the area outside the angle of view captured in the image, and derivatives include adding scenes outside the angle of view (i.e., outside the image).

[0045] (Storage Unit 230) The storage unit 230 is realized by a ROM that stores programs and calculation parameters used in the processing of the control unit 220, and a RAM that temporarily stores parameters that change as needed.

[0046] The storage unit 230 stores an advertisement information DB 231 , a tag information DB 232 , a user preference information DB 233 , an advertiser request information DB 234 , and an advertisement creation information DB 235 .

[0047] The advertisement information DB 231 stores advertisement information in which advertisement IDs, advertisement images or advertisement audio, advertisement effectiveness results, and distribution destination UIDs are associated with each other. The tag information DB 232 stores information such as advertisement effectiveness results for each user for each tag as tag information.

[0048] The user preference information DB 233 stores preference information about advertisements for each user (or for each user attribute). The preference information about advertisements is, for example, the advertising effectiveness results for each advertisement or tag. The user preference information may be stored by the advertising effectiveness result analysis unit 224. The preference information may include not only positive information but also negative information. For example, the user preference information DB 233 may also store information about advertising images that the user did not click at all (images that the user dislikes or is not interested in).

[0049] The advertiser request information DB 234 stores information requested by advertisers regarding their advertisements. The information requested regarding advertisements is, for example, a prompt (tag group) for image generation sent from the advertiser terminal 30. The DB 234 also stores information on parts of the advertisement that the advertiser has specified in advance and that the advertiser does not want to be derived from. The parts that the advertiser does not want to be derived can be, for example, a face (person), a product, or advertising copy.

[0050] The advertisement generation information DB 235 stores information used for advertisement generation. The advertisement generation information DB 235 may store, for example, various image generation AIs. Furthermore, the advertisement generation information DB 235 may store information such as product images to be advertised in advance.

[0051] Although the configuration of the server 20 has been specifically described above, the configuration of the server 20 according to the present disclosure is not limited to the example shown in FIG. 2. For example, the server 20 does not necessarily have all of the components shown in FIG. 2. The server 20 may also be realized by a plurality of devices. The server 20 may also be realized by a virtual server provided by a cloud service platform. At least a portion of the components of the server 20 may also be realized by an external server or external system.

[0052] <<3. Generation of Advertisement>> Next, generation of advertisement according to this embodiment will be described.

[0053] <3-1. Advertisement Generation Cycle> Fig. 3 is a diagram for explaining an example of an advertisement generation cycle according to this embodiment. As shown in Fig. 3, first, after an advertisement image 401 is distributed (step S1), the server 20 converts the advertisement image into text (image description) using the text conversion unit 222 (step S2). The advertisement image distributed in S1 may be generated by the advertisement generation unit 221 using an image generation AI, or may be generated by other means (including photography).

[0054] Next, the server 20 acquires the advertising effectiveness results of the advertising image 401 (specifically, the click-through rate of the advertisement, etc.), and the advertising effectiveness results analysis unit 224 associates the advertising effectiveness results with the IDs (UIDs) of the users to whom the advertising image 401 is delivered, for each tag included in the tag group 402 obtained by the text conversion in S2 (step S3).

[0055] Next, the server 20 generates a prompt by editing tags, such as adding or changing tags, in the tag group 402 obtained by the text conversion in S2 using the prompt generation unit 223 (step S4). The tag group 403 to be added or changed may be tags that match the preferences of the advertisement target user. For example, the prompt generation unit 223 may add to the tag group 402 tags that have advertising effect results equal to or greater than a predetermined threshold value, among the tags associated with the advertisement target user's UID.

[0056] Next, the server 20 generates a new advertisement image 404 using the advertisement generation unit 221 based on the prompt generated in S4 (step S5). The advertisement image 404 can be generated automatically by the image generation AI. As described above, by adding or changing tags that match the preferences of the advertisement target user, it is possible to generate an advertisement image that better matches the user's preferences.

[0057] The prompts input to the image generation AI include positive prompts and negative prompts. Figure 4 is a diagram for explaining the details of the prompts according to this embodiment. As shown in Figure 4, the prompts generated according to this embodiment may be, more specifically, positive prompts and negative prompts.

[0058] The positive prompt includes tags that are desired to be included in elements of the advertising image to be generated, and the negative prompt includes tags that are not desired to be included in elements of the advertising image to be generated. In step S4, the server 20 generates both prompts using the prompt generation unit 223. For example, the server 20 may add tags obtained from advertising images that have been clicked by the user to the positive prompt, and add tags obtained from advertising images that have not been clicked by the user to the negative prompt. In this way, including negative prompts in the prompt makes it possible to specify elements that are desired to be excluded from the advertising image to be generated, thereby improving the accuracy of generating advertising images.

[0059] Then, the server 20 distributes (provides to the user) the advertising image generated in S5 (step S6).

[0060] The information processing system 1 according to this embodiment repeats steps S2 to S6 above, thereby enabling continuous automatic updating of advertisement images and reducing the burden on the advertisement provider (advertiser) in generating advertisements. Furthermore, since user preferences based on the advertisement effectiveness results can be reflected in the advertisement generation, advertisements that are more suitable for the advertisement recipient user, i.e., advertisements that are expected to have increased advertising effectiveness, can be automatically generated.

[0061] <3-2. Prompt Generation Patterns> In this embodiment, there are several patterns for generating prompts by the prompt generating unit 223. Each pattern will be described in detail below.

[0062] (1) Addition of Tags Matching User Preferences, etc. The prompt generation unit 223 may generate a prompt by adding tags that match the user's preferences to a tag group obtained from an advertisement image (the original image) based on the advertising effectiveness results of various tags obtained by analyzing the advertising effectiveness results of advertisements provided to the user, or by changing the tags. For example, the prompt generation unit 223 may add tags that have advertising effectiveness results higher than a threshold, among various tags associated with the user, to the tag group obtained from the advertisement image (the original image). Furthermore, if a tag included in the tag group is related to a tag that has a high advertising effectiveness result for the user, the prompt generation unit 223 may change the tag to a tag that has a high advertising effectiveness result. Generating an advertisement image using such a prompt makes it possible to generate an advertisement image that matches the user's preferences.

[0063] (2) Addition of tags that are not influenced by user preferences, etc. The prompt generation unit 223 may generate a prompt by adding or changing the tags to the tag group obtained from the original image, i.e., the advertising image, such as tags randomly selected from various tags (other than those included in the tag group), seasonal tags, tags that are popular, tags that are trending on SNS, tags that are preferred by friends of the user to whom the advertisement is provided, etc.

[0064] In the pattern presented in (1) above, tags that match the user's preferences are added based on the user's (advertising) preference history. However, if the advertisement generation cycle continues, the user may repeatedly be presented with similar advertisements (elements of advertisement images) that match the user's preferences, which may lead to boredom. Therefore, the pattern presented in (2) makes it possible to provide advertisement images that include elements that are new to the user but are expected to attract their interest. The information processing system 1 can avoid continuously providing advertisements that are tailored to the user's preferences and provide more effective advertisements.

[0065] Here, the generation of an advertising image (derived advertising image) based on a prompt generated by pattern (2) will be described with reference to FIGS. 5 and 6. FIG.

[0066] First, the derivable element extraction unit 225 extracts derivable elements from the advertising image 500, which is an original image. FIG. 5 is a diagram for explaining the extraction of derivable elements from the advertising image 500, which is an original image. The derivable element extraction unit 225 separates objects in the advertising image 500 by segmentation. In the example shown in FIG. 5 , for example, a woman 501, hair 502, a face 503, clothes 504, shoes 505, a coffee cup 506, and a background 507 may be separated from the advertising image 500. The derivable element extraction unit 225 identifies the separated objects and tags the separated objects according to the identification results (such as woman, hair, face, clothes, and shoes).

[0067] Next, the advertising generation unit 221 uses image generation AI to derive (change) parts (derivable elements) of the advertising image 500 that correspond to tags that have been changed, added, etc. to the group of tags obtained from the original image, advertising image 500, in the generated prompt, so that they conform to the changed, added, etc. tags, thereby generating a derived advertising image derived from the original image, advertising image 500.

[0068] Fig. 6 is a diagram showing an example of a derived advertising image. The left side of Fig. 6 shows an original image, advertising image 500. From this advertising image 500, derived advertising images 500-1 to 500-3, 500-11 to 500-12, etc. are generated by deriving derivable elements corresponding to tags that have been changed, added, etc.

[0069] Derived advertising images 500-1 to 500-3 are example images derived from the color of the clothes 504 in the advertising image 500. For example, if "brown dress" included in the tag group obtained from the advertising image 500 is changed to "black dress," "blue dress," "red dress," or the like, derived advertising images 500-1 to 500-3 derived from the color of the clothes 504 in the advertising image 500 can be generated as shown in FIG.

[0070] Derived advertising images 500-11 to 500-12 are example images derived from the color of hair 502 in advertising image 500. For example, if "black hair" included in the tag group obtained from advertising image 500 is changed to "red hair," "blonde hair," or the like, derived advertising images 500-11 to 500-12 derived from the color of hair 502 in advertising image 500 can be generated as shown in FIG.

[0071] As another example of derivation, the advertisement generation unit 221 may perform product placement (PP). For example, if "brown dress" included in the tag group obtained from the advertisement image 500 is changed to "blue dress (PP)," the advertisement generation unit 221 may change the color of the clothes 504 in the advertisement image 500 to blue, and may also display the name of the product or company to be advertised in the area corresponding to the clothes 504.

[0072] (3) Addition of tags, etc., taking into consideration portions desired (permitted) to be changed by advertisers, etc. When performing (1) or (2) above, the prompt generation unit 223 may make changes within portions (portions desired to be changed) that are desired to be changed (derived) in advance by advertisers, advertising businesses, etc. Advertising content includes images (still images, videos) and audio, and advertisers and advertising businesses can specify in advance the desired changes (permitted changes) for the portions desired (permitted) to be changed (derived), such as allowing part or all of the image or audio, allowing advertising copy, allowing people, etc.

[0073] (4) Addition of tags taking into consideration parts prohibited from being changed by advertisers, etc. When performing (1) or (2) above, the prompt generation unit 223 may make changes other than parts (prohibited parts) that have been prohibited from being changed (derivated) in advance by the advertiser, advertising company, etc. The advertiser or advertising company can specify in advance the parts that are prohibited from being changed (derivated) such as all or part of an image or audio, advertising copy, people, etc.

[0074] (5) Adding tags based on general instructions from advertisers, etc. The prompt generation unit 223 can also generate tags to be added, changed, etc. based on general instructions from advertisers or advertising businesses. For example, the prompt generation unit 223 uses a natural language processing model to generate specific tags that can be used as prompts based on general instructions from advertisers or advertising businesses (abstract concepts, such as season, location, time, current events, atmosphere, etc.).

[0075] (6) Utilization of Negative Prompts The prompt generation unit 223 may add tags that the user dislikes or is indifferent to the negative prompt. Tags that the user dislikes or is indifferent to may be obtained from advertising images that the user did not click. The information processing system 1 adds tags obtained from advertising images that the user did not click to the negative prompt. Furthermore, advertising images that the user did not click may include images that are causing hallucination, such as images with an obviously incorrect number of fingers or images with a distorted face. Therefore, by adding tags obtained from such images to the negative prompt as tags that the user dislikes or is indifferent to, it is possible to reduce hallucination.

[0076] <3-3. Others> The server 20 may generate a large number of prompts using the prompt generation unit 223, and may generate a large number of advertising images (including derived advertising images) using the advertisement generation unit 221. When a large number of advertising images are generated, the control unit 220 of the server 20 may perform a process of selecting advertising images to be distributed. For example, the control unit 220 may randomly select advertising images to be distributed. Furthermore, the control unit 220 may use a prediction AI that predicts advertising effectiveness to predict the advertising effectiveness of the large number of advertising images generated by the advertisement generation unit 221, and select the top advertising images as the advertising images to be distributed.

[0077] In addition, the control unit 220 may use hallucination detection AI to determine whether an image is showing hallucination, such as an image with an obviously incorrect number of fingers or an image with a distorted face, and prevent the advertisement from being delivered.

[0078] In addition, the control unit 220 can create a judgment AI for each user to whom the images are to be distributed to determine which images the user dislikes and which images the user is not interested in, so that images that the user dislikes or is not interested in are not distributed to the user.

[0079] <<4. Operational Processing>> Next, the operational processing of this embodiment will be described with reference to the drawings.

[0080] 4-1. First Example FIG. 7 is a flowchart showing a first example of the flow of a process for generating a derivative advertising image in the information processing system 1. As shown in FIG.

[0081] As shown in FIG. 7, first, the information processing system 1 distributes an advertising image (step S103).

[0082] Next, the information processing system 1 converts the advertisement image into text and acquires a tag group as image description data (step S106).

[0083] Next, the information processing system 1 generates a prompt for generating an advertisement image by editing the group of tags obtained from the advertisement image (step S109). For example, the information processing system 1 adds or changes the group of tags obtained from the advertisement image to randomly selected tags, seasonal tags, popular tags, tags that are trending on SNS, tags that are popular among friends of the advertisement recipient user, etc. By generating a prompt by performing such editing that is not influenced by the user's (advertising) preferences, it is possible to avoid continuously providing advertisements that are biased toward the user's preferences.

[0084] Next, the information processing system 1 extracts derivable elements from the advertising image (step S112). The extraction of derivable elements may involve detecting and tagging (areas of) objects, etc., within the advertising image. The detected objects can be considered derivable elements. Furthermore, by identifying and tagging the detected objects, etc., it is possible to determine which parts of the advertising image should be changed (derived) when generating a derived advertising image, as described below.

[0085] Next, the information processing system 1 generates a derived advertising image by deriving derivable elements of the original advertising image based on the generated prompt (step S115). As described above, deriving derivable elements includes changing or adding elements of the advertising image. Advertising image elements may include objects, scenes, advertising copy, and the like. By generating the prompt used to generate the advertising image by adding or changing randomly selected tags or popular tags, as described above, it is possible to provide a more effective advertisement that is fresh and likely to attract the user's attention.

[0086] The information processing system 1 repeats the processes shown in steps S103 to S115 described above, and can automatically and continuously deliver and update advertising images.

[0087] <4-2. Second Example> The process of generating a derived advertising image in the information processing system 1 according to this embodiment is not limited to the first example described above, and for example, the analysis of advertising effectiveness results may be taken into account when generating a derived advertising image. Hereinafter, this will be described with reference to FIG. 8 .

[0088] FIG. 8 is a flowchart showing a second example of the flow of the process for generating a derivative advertising image in the information processing system 1.

[0089] As shown in FIG. 8, first, the information processing system 1 distributes an advertising image (step S203).

[0090] Next, the information processing system 1 acquires the advertising effectiveness result for the distributed advertising image (step S206). The advertising effectiveness result may be data for each user to whom the advertising image is provided (distributed).

[0091] Next, the information processing system 1 converts the advertisement image into text and acquires a tag group as image description data (step S209).

[0092] Next, the information processing system 1 associates each tag of the tag group obtained from the advertising image with the advertising effectiveness result (step S212). Information associating each tag with the advertising effectiveness result can be stored in a database (DB). The DB (tag information DB 232) can store information associating various tags with the advertising effectiveness result.

[0093] Next, the information processing system 1 generates a prompt for generating an advertising image by editing the tag group obtained from the advertising image (step S215). At this time, the information processing system 1 may add or change tags suitable for the user based on information relating to the association between various tags and advertising effectiveness results stored in the DB (tag information DB 232). A tag suitable for the user is assumed to be a tag that matches the user's preferences, more specifically, a tag whose advertising effectiveness result value exceeds a threshold.

[0094] Next, the information processing system 1 extracts derivable elements from the advertisement image (step S218).

[0095] Next, the information processing system 1 generates a derived advertising image by deriving derivable elements of the advertising image that is the original image based on the generated prompt (step S221). By generating the prompt used to generate the advertising image by adding or changing tags that match the user's preferences as described above, it is possible to provide an effective advertisement that matches the user's preferences.

[0096] The information processing system 1 repeats the processes shown in steps S203 to S221 described above, and can automatically and continuously deliver and update advertising images.

[0097] <4-3. Combination> The above describes the generation of derived advertising images using multiple examples. Note that the information processing system 1 may continuously deliver and update advertising images by appropriately switching between the first example and the second example. Furthermore, when repeating the second example, the information processing system 1 may appropriately switch the prompt generation pattern in step S215. Specifically, the information processing system 1 may appropriately switch between a prompt generation pattern in which tags that match the user's preferences are added or modified, and a prompt generation pattern in which a prompt is generated by editing the prompt without being influenced by the user's (advertising) preferences, such as randomly selected tags or tags that are trending on SNS.

[0098] Furthermore, when generating the prompt in the first or second example, the information processing system 1 may prevent tags from being changed in the tag group obtained from the original advertising image for portions (prohibited portions) designated by the advertiser or advertising business not to be changed (derivated). The advertiser or advertising business can specify in advance the portions that are prohibited from being changed (derivated) when updating an advertisement. Information on the prohibited portions can be stored in the advertiser request information DB 234. The advertiser or advertising business can specify, for example, the types of elements in the advertising image or audio that are prohibited from being changed. For example, it is conceivable to specify specific elements (tag specification) that are prohibited from being changed in the advertising image or audio, or to specify the portions that are prohibited from being changed (such as "face," "product," "background," or "hairstyle").

[0099] Furthermore, when generating the prompt in the first or second example, the information processing system 1 may change the portion (requested change portion) that the advertiser or advertising agent wishes to change (derive). The advertiser or advertising agent can specify in advance the portion they wish to change (derive) when updating the advertisement. Information on the requested change portion can be stored in the advertiser request information DB 234. The advertiser or advertising agent can specify, for example, the type of element of the advertising image or audio that they wish to change. For example, they may wish to change part or all of the advertising image or audio, or they may wish to change the advertising copy.

[0100] Furthermore, when generating the prompt in the first or second example, the information processing system 1 may generate tags to be added, changed, etc. based on general instructions from the advertiser or advertising agency.

[0101] Furthermore, when generating the prompt in the first or second example, the information processing system 1 may add tags that the user dislikes or is indifferent to to the negative prompt.

[0102] <<5. Supplementary Information>> Although preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings, the present technology is not limited to such examples. It is clear that a person skilled in the art of the present disclosure can conceive of various modified or altered examples within the scope of the technical idea described in the claims, and it is understood that these also naturally fall within the technical scope of the present disclosure.

[0103] For example, one or more computer programs can be created for hardware such as a CPU, ROM, and RAM built into the server 20 to perform the functions of the server 20. Also provided is a computer-readable storage medium storing the one or more computer programs.

[0104] Furthermore, the effects described herein are merely descriptive or exemplary and are not limiting. In other words, the technology according to the present disclosure may achieve other effects that will be apparent to those skilled in the art from the description of this specification, in addition to or in place of the above-described effects.

[0105] The present technology can also be configured as follows. (1) An information processing device including a control unit that edits a group of tags obtained by converting advertising content into text, extracts derivable elements from the advertising content, and generates derived advertising content based on the advertising content by deriving the corresponding elements in accordance with the edited group of tags. (2) The information processing device according to (1), wherein the extracted elements are tagged, and the control unit uses image generation AI to change the elements corresponding to added or changed tags in the edited group of tags in accordance with the added or changed tags, thereby generating the derived advertising content. (3) The information processing device according to (1) or (2), wherein the advertising content is an image or audio. (4) The information processing device according to any one of (1) to (3), wherein the control unit changes hair color, hairstyle, pose, facial expression, clothing, or person as a derivative of the elements of the advertising image, which is the advertising content. (5) The information processing device according to any one of (1) to (4), wherein the control unit acquires an advertising effect result of the advertising image for each user to whom the advertising image that is the advertising content is provided, and associates the advertising effect result with each tag included in the tag group obtained by converting the advertising image into text, together with a user ID. (6) The information processing device according to (5), wherein the control unit adds a tag to the tag group in accordance with the preference of a user to whom the advertising image is provided, or changes the tag to the tag group, based on accumulated association information between various tags and the advertising effect result. (7) The information processing device according to any one of (1) to (5), wherein the control unit adds a tag to the tag group in accordance with the preference of a user to whom the advertising image that is the advertising content is provided, or changes the tag to the tag group regardless of the preference of a user to whom the advertising image is provided. (8) The information processing device according to (7), wherein the control unit adds or changes at least one of tags selected at random from various tags, seasonal tags, popular tags, tags trending on social media, and tags preferred by friends of the user to whom the advertisement is provided to the tag group.(9) The information processing device according to any one of (1) to (8), wherein the control unit edits the tag group within a pre-specified portion desired to be changed. (10) The information processing device according to any one of (1) to (9), wherein the control unit edits the tag group taking into consideration a pre-specified portion prohibited from being changed. (11) The information processing device according to any one of (1) to (10), wherein the control unit uses the edited tag group as a prompt and generates the derived advertising content using an image generation AI. (12) The information processing device according to (11), wherein the prompt includes a positive prompt and a negative prompt. (13) An information processing method, wherein a processor edits tags obtained by converting advertising content into text; extracts derivable elements from the advertising content; and generates derived advertising content based on the advertising content by deriving the corresponding elements according to the edited tag group. (14) A program that causes a computer to function as a control unit that edits a group of tags obtained by converting advertising content into text, extracts derivable elements from the advertising content, and generates derived advertising content based on the advertising content by deriving the corresponding elements according to the edited group of tags.

[0106] REFERENCE SIGNS LIST 1 Information processing system 10 User terminal 20 Server 210 Communication unit 220 Control unit 221 Advertisement generation unit 222 Text conversion unit 223 Prompt generation unit 224 Advertising effectiveness result analysis unit 225 Derivable element extraction unit 230 Storage unit 231 Advertisement information DB 232 Tag information DB 233 User preference information DB 234 Advertiser request information DB 235 Advertisement generation information DB 30 Advertiser terminal 40 Network

Claims

1. An information processing device comprising: a control unit that edits a group of tags obtained by converting advertising content into text; extracts derivable elements from the advertising content; and generates derived advertising content based on the advertising content by deriving the corresponding elements according to the edited group of tags.

2. The information processing device of claim 1, wherein the extracted elements are tagged, and the control unit uses an image generation AI to modify the elements corresponding to added or changed tags in the edited tag group in accordance with the added or changed tags, thereby generating the derived advertising content.

3. The information processing device according to claim 1, wherein the advertising content is an image or sound.

4. The information processing device according to claim 1, wherein the control unit changes the hair color, hairstyle, pose, facial expression, clothing, or person as a derivation of the element of the advertising image which is the advertising content.

5. The information processing device of claim 1, wherein the control unit obtains advertising effectiveness results for each user to whom the advertising image, which is the advertising content, is provided, and associates the advertising effectiveness results with each tag included in the tag group obtained by converting the advertising image into text, together with a user ID.

6. The information processing device of claim 5, wherein the control unit adds or changes tags to the tag group according to the preferences of a user to whom the advertising image is to be provided, based on the accumulation of association information between various tags and the advertising effectiveness results.

7. The information processing device according to claim 1, wherein the control unit adds a tag to the tag group or changes the tag to the tag, regardless of the preferences of a user to whom the advertising image, which is the advertising content, is to be provided.

8. The information processing device of claim 7, wherein the control unit adds or changes to the tag group at least one of the following tags: a tag randomly selected from various tags, a seasonal tag, a popular tag, a tag that is trending on social media, and a tag preferred by friends of the user to whom the advertisement is provided.

9. The information processing device according to claim 1, wherein said control unit edits said tag group within a pre-specified portion desired to be changed.

10. The information processing device according to claim 1, wherein said control unit edits said tag group taking into consideration pre-specified portions that are prohibited from being changed.

11. The information processing device according to claim 1, wherein the control unit uses the edited tag group as a prompt and generates the derived advertising content using an image generation AI.

12. The information processing device of claim 11, wherein the prompts include positive prompts and negative prompts.

13. An information processing method comprising: a processor editing a group of tags obtained by converting advertising content into text; extracting derivable elements from the advertising content; and generating derived advertising content based on the advertising content by deriving the corresponding elements according to the edited group of tags.

14. A program that causes a computer to function as a control unit that edits a group of tags obtained by converting advertising content into text, extracts derivable elements from the advertising content, and generates derived advertising content based on the advertising content by deriving the corresponding elements according to the edited group of tags.

Images