Multimodal generative ai-based saas system for marketing content generation

Abstract

A SaaS service system for producing multimodal marketing content based on generative artificial intelligence is provided, comprising a solution providing server including a user terminal that inputs product data and outputs marketing content generated based on the product data, an input unit that receives product data, a generation unit that generates a prompt to generate marketing content for a product based on the product data, a calling unit that calls at least one generative artificial intelligence to generate marketing content, a management unit that causes at least one generative AI to generate marketing content based on the prompt, and a providing unit that provides the marketing content generated by at least one generative AI to the user terminal.

Description

Technology Field

[0001] The present invention relates to a SaaS service system for producing multimodal marketing content based on generative artificial intelligence, and provides a system that generates marketing content using product data and uploads the marketing content to advertising channels according to a schedule. Background Technology

[0002] Recently, the demand for online marketing among small businesses, including SMEs, one-person businesses, and freelancers, has been surging. In particular, content-based marketing via Social Network Services (SNS) is gaining popularity as a promotional tool characterized by low cost and high efficiency; however, hiring professional marketers or utilizing outsourced agencies entails high costs and the burden of long-term operations. The influence of SNS is absolute, with social media advertising—including mobile ads—accounting for the majority of total online advertising expenditures. Despite many companies recognizing digital transformation as an essential task, they are currently facing difficulties in implementation due to limitations such as insufficient marketing budgets and a lack of strategic planning capabilities. Furthermore, the number of one-person businesses is steadily increasing, estimated at 1.2 million by 2025. These companies record average sales of approximately 300 million won and most utilize SNS or online channels as essential marketing tools.

[0003] At this time, methods for performing large-media advertising, analyzing data using AI, or generating promotional content have been researched and developed in consideration of advertiser demand. In this regard, prior art, Korean Published Patent No. 2022-0028628 (published March 8, 2022) and Korean Registered Patent No. 10-2406442 (announced June 7, 2022), respectively disclose configurations for providing advertisements in various media, such as keyword advertising, SNS advertising, press release advertising, website advertising, targeting advertising, design advertising, online text advertising, and video content production advertising, based on advertiser requirements, as well as configurations for analyzing data and generating promotional content using AI, managing inbound and outbound schedules and inventory, processing orders, and providing customer service.

[0004] However, while the former is stated to provide advertisements through various media, it does not involve the automatic generation of advertising content; similarly, although the latter is stated to generate content using AI, it merely involves setting images and text and automatically combining them, rather than actually creating new content. Recently, GPT-based generative AI has been used to automatically generate images, but this is merely a configuration that creates images matching text, not one that generates marketing content. Furthermore, it lacks a post-implementation feature for content quality and suffers from the limitation of repeatedly using the same prompt structure. Therefore, research and development are required for a system that automatically plans and generates marketing content using product data, and automatically proposes or links upload schedules. The problem to be solved

[0005] One embodiment of the present invention provides a SaaS service system for creating multimodal marketing content based on generative artificial intelligence that helps one-person businesses or small business owners perform online marketing without a marketer by: when product data is uploaded from a user terminal, analyzing the text, photos, and target customers of the product data, configuring a prompt, and loading at least one generative artificial intelligence to generate marketing content; providing the generated marketing content to the user terminal or uploading it to a pre-configured social media account; and, after generating marketing content considering industry-specific seasonal events, scheduling the upload of the marketing content according to the schedule of a pre-stored event DB. However, the technical problem that this embodiment aims to solve is not limited to the technical problem described above, and other technical problems may exist. means of solving the problem

[0006] As a technical means for achieving the aforementioned technical task, one embodiment of the present invention includes a solution providing server comprising: a user terminal that inputs product data and outputs marketing content generated based on the product data; an input unit that receives product data; a generating unit that generates a prompt to generate marketing content for a product based on the product data; a calling unit that calls at least one Generative Artificial Intelligence (GAI) to generate marketing content; a management unit that causes at least one Generative AI to generate marketing content based on the prompt; and a providing unit that provides the marketing content generated by at least one Generative AI to the user terminal. Effects of the invention

[0007] According to any one of the means for solving the problem of the present invention described above, by providing an AI marketing automation system as a SaaS (Software as a Service) that automatically plans and generates marketing content based on user product data using generative AI, and automatically proposes and schedules social media uploads, the marketing burden of small business owners can be substantially reduced. Brief explanation of the drawing

[0008] FIG. 1 is a diagram illustrating a SaaS service system for creating multimodal marketing content based on generative artificial intelligence according to one embodiment of the present invention. Figure 2 is a block diagram illustrating a solution providing server included in the system of Figure 1. FIGS. 3 and 4 are drawings for illustrating an embodiment in which a marketing automation solution according to an embodiment of the present invention is implemented. FIG. 5 is a flowchart illustrating a method for providing a marketing automation solution according to an embodiment of the present invention. Specific details for implementing the invention

[0009] Embodiments of the present invention are described below with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0010] Throughout the specification, when a part is described as being "connected" to another part, this includes not only cases where they are "directly connected" but also cases where they are "electrically connected" with other elements interposed between them. Furthermore, when a part is described as "including" a component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components, and it should be understood that this does not preclude the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0011] Terms such as “about,” “substantially,” etc., used throughout the specification, are used to mean at or near the stated value when inherent manufacturing and material tolerances are presented in the stated meaning, and are used to prevent unscrupulous infringers from unfairly exploiting the disclosure in which precise or absolute values ​​are mentioned to aid in understanding the invention. Terms such as “step” or “step of” used throughout the specification of the invention do not mean “step for”.

[0012] In this specification, the term "part" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Additionally, one unit may be realized using two or more pieces of hardware, and two or more units may be realized by one piece of hardware. Meanwhile, "part" is not limited to software or hardware, and "part" may be configured to reside in an addressable storage medium or configured to run on one or more processors. Accordingly, as an example, "part" includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and "parts" may be combined into a smaller number of components and "parts" or further separated into additional components and "parts." In addition, the components and '~parts' may be implemented to play one or more CPUs within the device or secure multimedia card.

[0013] Some of the operations or functions described herein as being performed by a terminal, device, or device may instead be performed by a server connected to said terminal, device, or device. Likewise, some of the operations or functions described as being performed by a server may also be performed by a terminal, device, or device connected to said server.

[0014] In this specification, some of the operations or functions described as mapping or matching with a terminal may be interpreted as meaning mapping or matching the terminal's unique number or personal identification information, which is the terminal's identifying data.

[0015] The present invention will be described in detail below with reference to the attached drawings.

[0016] FIG. 1 is a diagram illustrating a SaaS service system for creating multimodal marketing content based on generative artificial intelligence according to an embodiment of the present invention. Referring to FIG. 1, the SaaS service system (1) for creating multimodal marketing content based on generative artificial intelligence may include at least one user terminal (100), a solution providing server (300), and at least one information providing server (400). However, since the SaaS service system (1) for creating multimodal marketing content based on generative artificial intelligence of FIG. 1 is merely an embodiment of the present invention, the present invention is not to be interpreted as being limited through FIG. 1.

[0017] At this time, each component of FIG. 1 is generally connected through a network (Network, 200). For example, as shown in FIG. 1, at least one user terminal (100) can be connected to a solution providing server (300) through the network (200). And, the solution providing server (300) can be connected to at least one user terminal (100) and at least one information providing server (400) through the network (200). Also, at least one information providing server (400) can be connected to the solution providing server (300) through the network (200). And, at least one information providing server (500) can be connected to at least one user terminal (100), the solution providing server (300), and at least one information providing server (400) through the network (200).

[0018] Here, a network refers to a connection structure capable of exchanging information among individual nodes, such as multiple terminals and servers. Examples of such networks include Local Area Networks (LANs), Wide Area Networks (WANs), the World Wide Web (WWW), wired and wireless data networks, telephone networks, and wired and wireless television networks. Examples of wireless data communication networks include, but are not limited to, 3G, 4G, 5G, 3GPP (3rd Generation Partnership Project), 5GPP (5th Generation Partnership Project), 5G NR (New Radio), 6G (6th Generation of Cellular Networks), LTE (Long Term Evolution), WIMAX (World Interoperability for Microwave Access), Wi-Fi, Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), RF (Radio Frequency), Bluetooth network, NFC (Near-Field Communication) network, satellite broadcasting network, analog broadcasting network, DMB (Digital Multimedia Broadcasting) network, etc.

[0019] In the following, the term "at least one" is defined as a term including both singular and plural forms, and it will be obvious that even if the term "at least one" does not exist, each component may exist in a singular or plural form and may mean singular or plural. Furthermore, whether each component is provided in a singular or plural form may be changed according to the embodiment.

[0020] At least one user terminal (100) may be a terminal of a small business owner, one-person business, freelancer, etc., that uploads product data using a web page, app page, program, or application related to a marketing automation solution, receives marketing content generated in response to the product data, or inputs a social media account to automatically upload marketing content.

[0021] Here, at least one user terminal (100) may be implemented as a computer capable of connecting to a remote server or terminal via a network. Here, the computer may include, for example, a navigation system, a laptop equipped with a web browser, a desktop, a laptop, etc. At this time, at least one user terminal (100) may be implemented as a terminal capable of connecting to a remote server or terminal via a network. At least one user terminal (100) may include all kinds of handheld-based wireless communication devices, such as navigation, PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminal, smartphone, smartpad, tablet PC, etc.

[0022] The solution providing server (300) may be a server that provides a marketing automation solution web page, app page, program, or application. Additionally, the solution providing server (300) may be a server that, when product data is input from a user terminal (100), extracts text describing the product, a photo of the product, and a target customer, generates a prompt, loads at least one generative AI, inputs the prompt, and outputs marketing content. Here, the solution providing server (300) may be implemented as a computer capable of connecting to a remote server or terminal via a network. Here, the computer may include, for example, a laptop, desktop, or laptop equipped with a navigation system and a web browser.

[0023] At least one information providing server (400) may be a server that provides data necessary for creating marketing content using a web page, app page, program, or application related to a marketing automation solution, and information for building an event database. Here, at least one information providing server (400) may be implemented as a computer capable of connecting to a remote server or terminal via a network. Here, the computer may include, for example, a laptop, desktop, or laptop equipped with a navigation system or a web browser.

[0024] FIG. 2 is a block diagram for explaining a solution providing server included in the system of FIG. 1, and FIG. 3 and FIG. 4 are drawings for explaining an embodiment in which a marketing automation solution according to an embodiment of the present invention is implemented.

[0025] Referring to FIG. 2, the solution providing server (300) may include an input unit (310), a generation unit (320), a calling unit (330), a management unit (340), a providing unit (350), a scheduling unit (360), an automatic upload unit (370), and a response reflection unit (380).

[0026] When a solution providing server (300) or another server (not shown) operating in conjunction with one embodiment of the present invention transmits a marketing automation solution application, program, app page, web page, etc. to at least one user terminal (100) and at least one information providing server (400), the at least one user terminal (100) and at least one information providing server (400) may install or open the marketing automation solution application, program, app page, web page, etc. Additionally, a service program may be operated on at least one user terminal (100) and at least one information providing server (400) using a script executed in a web browser. Here, a web browser refers to a program that enables the use of web (WWW: World Wide Web) services and receives and displays hypertext described in HTML (Hyper Text Mark-up Language), and includes, for example, Chrome, Microsoft Edge, Safari, Firefox, Whale, UC Browser, etc. In addition, "application" refers to an application on a terminal, and includes, for example, an app running on a mobile terminal (smartphone).

[0027] Referring to FIG. 2, the input unit (310) can receive product data. The user terminal (100) can input product data. The product data may include text describing the product, a photo of the product, and information on the target customers of the product. In this case, assuming that the user of the user terminal (100) sells women's cosmetics on Naver Smart Store, the input of product data may be substituted by uploading a link to the Smart Store where the product is sold. This is because the input unit (310) can access the page of the link and extract text and images. However, since the product page is mostly an image, if text is not detected, the description of the product can be extracted by extracting text through OCR. Also, if information on the target customers of the product needs to be received or extracted but is not input, a method of inferring it based on the text and images (photos) of the product data can be used.

[0028] For example, let's assume that the product page of a Smart Store includes information such as Table 1 below.

[0029] - Product Name: "Premium Functional Probiotics 60 Sachets (Health Management for the Whole Family)" - Product Category: Health Foods > Probiotics - Product Description: "Contains 3 patented strains for gut health and immunity; suitable for everyone from children to the elderly. Easily take care of your health with just one sachet a day." - Detailed Image: Scene of family members taking the product together, female model in her 50s

[0030] 1. Text-based Keyword Extraction (NLP Preprocessing) Subject of analysis Derived Keywords Product Name Premium, Probiotics, Whole Family, Health Explanatory text Gut health, immunity, children, seniors, 1 packet a day Text within images (or OCR) Family, easy to take

[0031] At this stage, after extracting each text, morphological analysis and TF-IDF can be applied, or keywords can be extracted based on KeyBERT and FastText and then vectorized. Next, when classifying customer groups, rule-based inference is used initially due to the limited amount of data; however, once sufficient data has accumulated to warrant modeling, the classification model developed through training, validation, and testing can be utilized. In other words, rule-based inference is used in the initial stages, while inference can be performed using a classification model once sufficient data has been accumulated. In this case, the classification model may be a class classification model fine-tuned based on historical product data and actual purchasing customer information.

[0032] Input: "Probiotics, Children, Easy to Consume, Patented Strain, Family"→ Output: ["Women in their 30s to 50s", "Housewife interested in family health", "Consumer of functional health foods"]

[0033] Classification models can utilize, for example, a combination of Sentence-BERT and XGBoost, or a combination of LLM and Few-shot Prompting. For instance, target customer information can be derived by processing a query in GPT such as, "Who is the consumer group most likely to purchase this product?" Alternatively, targets can be inferred based on images; target customer information can be supplementarily inferred by utilizing the age, mood, and color tones of the models appearing in the images. For instance, if a middle-aged model appears in a family image, the target can be inferred to be family-oriented consumers or women aged 40 or older. For this purpose, a multimodal model can utilize, for instance, CLIP-based image-to-text mapping, employing a method that analyzes similarity with target keywords. Furthermore, targets can be classified using NLP after automatically generating image captions based on BLIP2 or Flamingo.

[0034] Analysis Items Judgment result Text keyword analysis Family, Gut Health, Children, Convenient Image analysis Female model in her 50s, family appears Learning based on past similar products Women in their 30s to 50s, interested in family health ▶ Final Target Profile Housewives in their 40s interested in family health, online health food consumer base

[0035] Alternatively, methods for inferring customer personas can be utilized. That is, by analyzing product descriptions, categories, and review data—not just the product page descriptions—the target customer base (age group, gender, interests, consumption patterns, etc.) can be automatically inferred, and the tone, writing style, keywords, and image style can be adjusted based on this information when generating marketing content.

[0036] step detail ① Enter product information - Input Items: Product Name, Description, Category, Price, Key Keywords - Technologies Used: Frontend Form Processing + Backend API ② Collection of reviews for similar products Method for extracting similar products - Category-based search - K-NN extraction based on cosine similarity after product description embedding Data source - Collection of reviews based on proprietary DB crawler (Coupang, Naver Shopping, Amazon, etc.) Technology used - Sentence-BERT or KoSimCSE-based embedding models - Elasticsearch, FAISS, Milvus (embedding search engines) ③ Review Data Analysis Natural language processing elements - Extract noun / adjective keywords → Identify interests - Sentiment analysis → Verify customer group emotional responses - Subject / person analysis → Infer age and gender from "I," "my child," "my mom," etc. Representative algorithm - Sentiment / topic classifiers based on KoBERT, EtriETRI, and HuggingFace Transformers - Interest classification based on LDA and BERTopic - Rule-based keyword matching: "Mom", "boyfriend", "elementary school student", etc. ④ Target Customer Profiling Create profile based on output attributes - Purpose of prompt optimization - Provides a guide for selecting the style of the generated result Usage examples "Women in their 30s + Emotional keywords + Beauty interests" → Instagram-style vibe + Soft writing + Includes emojis ⑤ Generative AI Content Creation underlying technology - Prompt Engineering: "Introduce natural ingredient skincare products for women in their 30s. Write in an emotional style." ▶ Control Token Method: Insert tokens such as [Women][30s][Emotional Tone] and pass them to a fine-tuned model ▶ Fine-tuning-based LLM (e.g., GPT-3.5 + LoRA, KoAlpaca, etc.) Example of results Has your skin become sensitive lately? Protect your own time with gentle, natural hydration.

[0037] In this case, text embeddings for calculating product description similarity can utilize, for example, Sentence-BERT or KoSimCSE; review crawling to secure similar reviews can utilize, for example, BeautifulSoup or Scrapy; sentiment topic analysis to infer review-based tendencies can utilize KoBERT, LDA, or BERTopic; and interest keyword matching for rule-based inference can utilize KoNLPy or Keyword-Matcher. Additionally, to automate customer inference, a hybrid approach combining rule-based and machine learning can be used, for example, LightGBM or Scikit-Learn.

[0038] Additionally, GPT API, LLaMA2, and KoAlpaca can be used for LLM connection, and LoRA and Prompt Injection can be used for token-based style reflection corresponding to style control, but the available tools are not limited to the tools described above. The generation unit (320) can generate a prompt to generate marketing content for a product based on product data. For example, if product data is entered as shown in Table 6 below, a template-type prompt can be configured based on this input value.

[0039] ▶ Product Data - Product Name: Cowhide Bucket Bag - Brief Description: A classic bag perfect for the commute - Target Customers: Working women in their 30s - Features: Italian leather, refined design - Link: https: / / smartstore.naver.com / xyz ▶ Template-based prompt You are a marketing content generation AI. The following is product information for a client: - Product Name: Cowhide Bucket Bag - Target Audience: Working women in their 30s - Features: Refined design, Italian leather - Sales Purpose: Descriptive text and hashtags to use with images for social media uploads Request: Based on this information, please generate one piece of text and five hashtags for Instagram content.

[0040] The prompt template can be composed of LangChain or Jinja2, and the generative AI invoked may be, for example, OpenAI GPT-4, KoGPT, or Claude, depending on the type of content being generated, but is not limited to these.

[0041] The calling unit (330) can call at least one Generative Artificial Intelligence (GAI) to generate marketing content. The text of the marketing content generated by the Generative AI, for example, GPT-4, based on the above-described prompt may be as shown in Table 7 below.

[0042] An 'Italian Cowhide Bucket Bag' to add sophistication to your commute today. It captures a refined silhouette and a luxurious mood. #BucketBag #ItalianLeather #OfficeLookFor30s #DailyBag #WorkOutfit

[0043] In addition, for images, a generative AI, such as Stable Diffusion or DALL·E, can be used to provide a prompt such as [A leather bucket bag on a white marble table, elegant, studio lighting], and the generated image can be output as a thumbnail and provided as marketing content along with the text of Table 6. Alternatively, existing images can be extracted; for example, when a product page is entered as a link on the user terminal (100), the generation unit (320) can extract a thumbnail based on Open Graph tags. In addition, Cheerio.js or Puppeteer can be used when extracting images. Also, if there are no thumbnails to extract or if a new image is to be used, a method of searching for images can be used; the generation unit (320) can derive keywords from the aforementioned prompts, search for images using the Pexels API or Unsplash API based on the keywords, and then provide the search results to the user terminal (100).

[0044] The management unit (340) can enable at least one generative AI to generate marketing content based on a prompt. In this case, the generative AI may be a model finely tuned to be optimized for generating marketing content. For example, marketing content with similar yet different moods or styles is used on platforms such as Instagram, Naver Blog, Facebook, and TikTok. By building a dataset corresponding to each domain (Instagram, Naver Blog, Facebook, etc.) and using it to fine-tune the system so that marketing content can be generated according to the characteristics of each platform or channel, it is possible to generate marketing content of different styles depending on each channel or platform even with the same prompt. That is, domain-specific micro-fine-tuning or style adaptation can be performed.

[0045] Same prompt → Create marketing content with different styles for each platform input output of power "Recommend some moisturizing summer cosmetics." → Instagram: "The hydration is different?? I only use this these days! #SummerMustHave #MoistureBomb" → Naver Blog: "Moisture products are important in this weather. The cosmetic I tried this time is..." → TikTok: "Summer moisture recommendation! Everyone gets a dewy glow if they apply this, so check it out"

[0046] Each channel has different writing styles, tones, emojis, sentence lengths, hashtags, and link patterns; by learning these individually, it is possible to generate different marketing content for each channel.

[0047] step detail particular Phase 1 Building domain-specific datasets -Source: Crawling or affiliate account-based data collection-Data structure: ▶Platform (Instagram / Blog / TikTok / Facebook) ▶Input prompt (or product description) ▶Output content (actual uploaded sentence) Phase 2 Text Style Conversion Learning (Domain Conditioning) Method A: Platform Tag Injection Method (Control Token) <instagram> Recommend some moisturizing summer cosmetics → The hydration is different?? I only use this these days!< / instagram> Method B: LLM fine-tuning (LoRA, PEFT, DPO, etc.) ▶ Domain-specific parameter branching in LLMs based on OpenAI, LLaMA2, Mistral, etc. ▶ "Platform-Aware Text Generation" e.g.: "Prompt + [Platform=TikTok]" → Outputs TikTok style text

[0048] For data crawling, examples such as Selenium, BeautifulSoup, Instaloader, YouTube API, and TikTok Scraper can be used; for data labeling, ChatGPT Labeler or PromptLab can be used; for preprocessing, spaCy, KoNLPy, fasttext, and SentencePiece can be used; and for vectorization, BERT, SBERT, KoELECTRA, and OpenKoLLM can be used. Additionally, for training frameworks, examples such as HuggingFace Transformers, PEFT, DeepSpeed, and LoRA can be used. In this case, the platform tag injection method of Method A is a lightweight fine-tuning method that trains to output different content styles depending on each platform (Instagram / blog / TikTok, etc.) even for the same prompt input.

[0049] For example, the input prompt is [ <instagram>If you say, "[Recommend some moisturizing summer cosmetics.]", LLM is <instagram>It accepts the control token as a style signal and, during training, remembers what style, tone, length, and emojis are needed based on that control token. Accordingly, the output comes out in an Instagram style, for example, it can be output as ["Moisture bursting!, This is all you need for summer! #SummerItem #DewySkin"].

[0050] The learning structure may be as shown in Table 10 below, but is not limited thereto.

[0051] [Prompt: " <tiktok> "Recommend a serum for sensitive skin"]→ [Output: "You'll regret it if you don't use this serum! OK for sensitive skin too!"][Prompt: "<NAVER_BLOG> "Recommend a serum for sensitive skin"]→ [Output: "Introducing a serum I would like to recommend to those who need skin soothing."]< / tiktok>

[0052] By securing approximately 1,000 to 5,000 pairs of examples per domain in the format of Table 10, domain-segment learning can be performed by applying the LoRA Adapter over the existing LLM. In this context, LoRA (Low-Rank Adaptation) is a technique that enables lightweight fine-tuning by inserting small additional parameters (low-rank matrices) without fine-tuning the entire parameters of the existing large language model (LLM), and control tokens refer to keywords that act as indicators influencing style or format by inserting additional input tokens that imply a specific domain, style, or purpose before or in the middle of the prompt. Furthermore, Platform-Specific Prompt Injection refers to inserting " before the prompt <instagram>" or " <tiktok>It refers to a method of inserting control tokens for presenting domain conditions, such as ”, to induce the LLM to output in that style.

[0053] The providing unit (350) can provide marketing content generated by at least one generative AI to a user terminal (100). The user terminal (100) can output marketing content generated based on product data. The marketing content can be provided as SaaS (Software as a Service). That is, marketing content suitable for each channel can be generated by providing only a link containing a product page on the user terminal (100) without installing a program or application on the user terminal (100). In this case, if the event described later is also taken into account, the marketing content is generated to match each date by taking into account the seasonal event as well. In this case, if the SaaS is provided as B2B, billing must be carried out according to the subscription type, so the usage of how much data the user has used can be tracked and a billing system can be applied accordingly.

[0054] At this time, an ideation function may be further provided when producing marketing content. Ideation is the starting point of creative thinking that derives the core theme, campaign message, keywords, or storyline of marketing content, and refers to the process of automating human brainstorming in an AI model based on data. This can be summarized as shown in Table 11 below, but each process or step is not limited to this.

[0055] step explanation technology 1 Industry and timing recognition Industry (e.g., Beauty, Food) + Month (e.g., June) 2 Monthly Major Issue Mapping Season / Calendar-based issue extraction (e.g., rainy season, summer vacation) 3 Identify product characteristics / target information Extract Customer Groups / USPs through Product Description + Review Analysis 4 Deriving Marketing Topic Candidates Generate topic / copy / campaign ideas with GPT or LLM 5 Prompt-based conversion Convert to the form "Write this phrase targeting these customers" 6 Multi-style content creation Create content drafts reflecting platform-specific tone & style

[0056] For example, if the industry is beauty, the product is moisturizing cream, and the time is June, ideas can be generated in the flow shown in Table 12 below.

[0057] 1. Industry + Time Period → Key Issues in the Beauty Industry for June: Skin troubles during the rainy season, lack of moisture, dryness caused by air conditioning 2. Product Data Analysis → Products emphasizing "long-lasting moisture" and "soothing for sensitive skin" 3. Target Review Analysis → Women in their 20s and 30s, sensitive skin, preference for natural ingredients 4. Ideation Output Examples: - "Stay hydrated even during the rainy season" - "Skin hydration protected even in air-conditioned environments" - "June hydration routine for sensitive skin"

[0058] These ideas are used as core messages to generate prompts for each marketing content. At this time, public calendar APIs, weather APIs, and wiki / trend DBs can be used for issue mapping, and GPT-4, KoGPT, Claude, and Gemini can be used for idea generation. Additionally, KoBERT, keyword extraction, and NER can be used for product analysis; LoRA, StyleTokens, and prompt injection methods for style matching; and LangChain and Prompt Template Engine for automatic branching. The results of ideation can be utilized, for example, as shown in Table 13.

[0059] Types of deliverables example Content Topic Rainy Season Hydration Routine Keywords "Moisture Soothing", "Air Conditioning Dryness", "June Moisturizing" prompt Please introduce a moisturizing cream for sensitive skin, even during the rainy season in June. Generation result Multi-channel output such as card news for social media, blog drafts, and scripts for shorts

[0060] Ideation does not simply mean creating sentences, but rather setting the direction of content based on [situation awareness + customer group inference + campaign design]. In one embodiment of the present invention, an artificial intelligence model can automatically perform this, thereby significantly reducing the user's planning burden.

[0061] When the automatic upload function is used on the user terminal (100), the scheduling unit (360) can extract the period and reason for which the product of the user terminal (100) should be advertised based on the event DB that stores industry-specific seasonal events, save it to a calendar, and generate a prompt by adding the reason for which the product should be advertised during the period in which the product should be advertised, and then generate marketing content. This can be named a monthly SNS content calendar, which is an SNS marketing schedule management tool that automatically generates and arranges marketing content generated by a generative AI on a daily basis based on product data, which is product or service information entered by the user.

[0062] The reason this tool is needed is that consistency is key to social media. In other words, no matter how good the content is, if uploads stop, it will fall behind in the algorithm and customer interest will drop along with it. However, it is difficult for one-person businesses or small business owners to plan, write, and upload content every day. To solve this, generative AI can generate a monthly content schedule. For example, assuming seasonal events are considered, a month's worth of content can be distributed according to upload dates. For example, it could be as shown in Table 14 below.

[0063] date Content Title Content Type Recommended hashtags Upload status 3 / 1 Back-to-school discounts start! Event Information #SpringSale #SemesterItems complete 3 / 3 Our customer said this Review-based #CustomerReview #ProofShot Unuploaded 3 / 8 Women's Day, a bag for you Themed content #WomensDay #WorkOutfit Reserved

[0064] At this time, the event database may consist of events based on industry and season, but events may be listed without distinction between industry or season. In this case, an event may refer to a specific user action, time condition, or external situation that triggers the creation or distribution of marketing content. For example, it may be as shown in Table 15 below, but is not limited thereto.

[0065] classification explanation example Time-based events A specific date or a schedule that occurs periodically Black Friday, Christmas, end-of-month sales, brand launch dates, etc. Product / Service-Based Events Product registration, inventory changes, surge in popularity, etc. New product registration, restock notifications, entry into bestsellers User behavior-based events Occurs based on user actions Cart abandonment, sign-up completion, first purchase API / External Integration Event Signal received from an external system Special exhibition information registered in ERP, schedules linked to the calendar AI Insight-based Event Generated based on AI analysis results Warnings such as "surge in review count," "increase in response rate for relevant keywords," and "low ad CTR"

[0066] To generate marketing content based on these events, a structure of [Event Listener + Trigger + Action Handler] can be used.

[0067] { "event_type": "Product_Registration", "event_triggered_at": "2025-06-26T14:00:00Z", "associated_product_id": 12345, "action": "Content_Automatic_Generation"}

[0068] For example, even if only a product page is entered on the user terminal (100), the scheduling unit (360) can extract brand names, target customers, keywords, etc., and generate and place marketing content based on [seasonal issues + industry-specific trends]. For example, if the bag sold by the user is a "bag for women in their 30s," marketing content can be generated and placed for the spring season, new semester, work look, Women's Day, Parents' Day, and summer preparation. Text and images containing keywords for the new semester can be generated for the new semester, and text and images containing keywords for summer can be generated for summer preparation, so that they are uploaded according to the schedule. At this time, the industry-specific seasonal events may be as shown in Table 17 below, for example, but are not limited thereto.

[0069] Industry month event Beauty March New semester start, White Day makeup, skin for the changing seasons fashion May Spring New Arrivals Lookbook, Trench Coat Recommendations, Work Look Food and beverages May Family Month Gift Sets, Children's Day Lunch Boxes Fitness / Exercise June Diet Season, Summer Home Workout Challenge education february New Semester Study Tips, New School Year Preparation Content pets december Winter Walking Precautions, Christmas Community / Agency November Year-end tax settlement guide, year-end party gathering

[0070] Of course, events are not concentrated in a single month for any one industry. For example, in the case of yoga studios, events are held every month: in December, an event wishing for success on the CSAT for students and a Holiday Appreciation Sale; in January, an exclusive January special offer for the New Year Healthy Package; in February, a special offer with double gift tickets for Lunar New Year reservations and an upgrade event; in March, spring discount coupons; in April, benefits for the first-time membership; in May, the launch of a surprise special pass for summer preparation; in June, a "Starting is Half" first-half clearance discount coupon event; in July, a summer event for dieters; in August, a cool double event; in September, a Chuseok benefit set; and in November, a free class event until the end of the year. Accordingly, it is possible to set themes for each month or reflect trends to create and upload new marketing content tailored to each industry.

[0071] The automatic upload unit (370) can upload marketing content to the social media account according to a schedule listed in the calendar when a social media account is entered on the user terminal (100). At this time, if only marketing content is uploaded, it may feel like there are too many advertisements, so information about products or services sold by the user can also be provided, or information about trends can be generated and provided together. For example, in the case of the marketing content of the yoga studio mentioned above, by providing various information related to mind and body training, such as the truth about cholesterol, the importance of body shape correction and adductor muscle exercises, and drinking water, in between the marketing content, the number of uploads to SNS can be met and the selection by the algorithm can be avoided.

[0072] The response reflection unit (380) can calculate the response index of the marketing content when the marketing content is uploaded to social media, and extract the prompts of the marketing content in order of highest response index to reflect them in the generation of the next prompt. That is, dynamic prompt engineering based on performance feedback can be used instead of fixed prompt engineering. While general prompt engineering is a process in which a person manually designs prompt sentences so that the AI ​​can produce the desired output well, the prompt engineering of the present invention is configured in the manner shown below and in Table 18.

[0073] division explanation Level of automation Prompts are dynamically generated and modified based on user input and content responses. core technology Prompt generation template + reinforcement learning (RL)-based feedback loop Application flow Product Data → Target Analysis → Extract Prompt Components → Automatic Generation of Prompts for LLM Invocation Performance reflection Updates the prompt component itself based on social media reactions (likes, comments, etc.) Prompt form "Generate content based on {emotional keywords} for the {target age group} according to {sentence length rules}"

[0074] Hereinafter, the operation process according to the configuration of the solution providing server of FIG. 2 described above will be explained in detail with reference to FIG. 3 and FIG. 4. However, it is obvious that the embodiment is merely one of the various embodiments of the present invention and is not limited thereto.

[0075] Referring to FIG. 3, (a) a solution providing server (300) collects product data from a user terminal (100), and (b) based on this, generates a prompt to call a generative AI, and then generates marketing content to provide to the user terminal (100) or automatically upload it to social media. At this time, (c) to ensure the continuity of uploading marketing content, the solution providing server (300) may generate marketing content according to industry-specific seasonal events and automatically schedule and upload the marketing content to a social media account designated by the user terminal (100) at that time. The process of the present invention is summarized as FIG. 4a, and user scenarios may be as FIG. 4b to 4e. A platform (tentative name, current Almond, formerly Marmond) according to an embodiment of the present invention has the purpose as in FIG. 4g and FIG. 4h, a vision as in FIG. 4i, and a configuration as in FIG. 4j. Additionally, a revenue structure may be created with a plan as in FIG. 4k. In addition, as shown in FIG. 41, content such as FIG. 4m can be generated on the start screen, and it can be automatically scheduled and uploaded as shown in FIG. 4n. Also, as shown in FIG. 4o, if there are keywords or points to be emphasized, content reflecting the user's intent can be generated by allowing input. Of course, FIG. 4a to FIG. 4o are merely explanatory screens or images, and the screens or images of the platform of the present invention are not limited thereto.

[0076] Content Creation Pipeline

[0077] A content creation pipeline according to one embodiment of the present invention may preferably be as shown in FIG. 3b.

[0078] 1. User Information Collection and Processing Step As a starting point for a SaaS-based multimodal marketing content creation system, it is a process of receiving essential information from users and automatically collecting and processing key data for targeting and content planning based on this. (1) User Information Input - Enter basic information such as marketing categories, product descriptions, personal websites, social media URLs, and marketing assets. - This information will be used as reference data for automatic content generation and target setting in subsequent stages. (2) Automatic extraction of keyword libraries optimized by SNS channel and target - Automatically sets the appropriate target age group for the relevant industry (e.g., cafes, gyms, cosmetics, etc.) based on marketing category data entered by the user. - Automatically extracts keywords from frequently used hashtags and popular caption lists related to the entered category by linking with an in-house keyword library optimized for each social media channel and target. - This process utilizes a word bank and external social media data and is performed without user intervention. (3) Automatic extraction of brand assets based on personal websites - Based on the personal website URL entered by the user, brand logos and design images with high marketing relevance are automatically extracted from the website. - The extraction process is performed automatically through crawling, Open Graph tag reading, and image relevance analysis. (4) Automatic data acquisition based on SNS URLs - Based on the SNS URL entered by the user, this system automatically collects SNS-related data such as the number of followers, registered logos, and past posts of the account. - This process is carried out through SNS API integration or web scraping, systematically securing the user's SNS operation status and branding elements. (5) Select user photo - Allows users to directly select photos they wish to use in social media designs from automatically extracted brand images and marketing-related images. - Users can actively incorporate visual elements that align with their branding intentions. (6) SNS account linking and data storage - It links acquired social media-related data to user accounts and stores all information entered by users and data automatically extracted by the system in an integrated database. - Through this, when users re-log in with the same account, they can receive continuous service based on their past input history and collected data, and establish an environment for data restoration and continuous marketing automation.

[0079] Through this, one embodiment of the present invention may include a function that supports a regular and periodic content creation schedule. Accordingly, data linked to a user account is utilized to automatically generate content on a monthly or weekly basis, and the user can continuously receive personalized marketing content without separate manual input. The data linkage structure according to one embodiment of the present invention can automate the user's marketing activities and minimize repetitive content creation tasks, thereby maximizing the productivity and efficiency of SNS marketing. Furthermore, one embodiment of the present invention can continuously enhance user-customized automatic recommendation and content optimization functions by learning usage patterns such as target customer base, preferred keywords, and image styles based on stored data, and reflecting this in content to be generated in the future.

[0080] 2. Proposal Creation and Revision Stage (1) Generative AI call and branding plan generation - The Generative AI (GPT) is invoked based on previously collected and processed data, as well as optimal information curated for the user. - The system comprehensively delivers multimodal data to the Generative AI and requests the generation of a branding plan based on that data. - The delivered data includes complex marketing elements such as product information, target age groups, social media data, hashtags, and design images. (2) Provide multiple 'Today's Feed' proposals - Color Scheme - Layout - Message - Caption - Highlight Point Each provided plan is optimized for the target audience and marketing category, and users can select their desired plan from multiple options. (3) Support for user modification and natural language-based editing - Users can select one of the proposals presented by the system or directly modify detailed elements of the proposal (color, wording, emphasis points, etc.). - Modifications are easily performed via natural language input, and the system supports the generative AI to immediately update the proposal upon user requests. - This enables users to efficiently complete a personalized final version of the proposal. (4) In-house SNS Template Library - Establishes and manages an in-house database of social media image templates created and classified by target audience and category. - Templates are pre-classified by target age group, industry, and keyword category, and the system is designed to automatically search and recommend them. (5) Automatic selection of optimal SNS template - Once the user finalizes the proposal, the system requests the generative AI to automatically select the most suitable social media template for the proposal. - The generative AI refers to a social media image template database, considering factors such as target customer groups, keyword categories, and layout conditions, and automatically matches and selects a template optimized for the final proposal. - This process is designed to enable the system to generate and recommend high-quality social media marketing content with minimal user intervention, and follows a data-driven structure that is continuously improved and optimized even with repeated use.

[0081] 3. Content Creation, Recurring Edits, and Scheduled Upload Management Steps (1) Automatic generation of SNS marketing images - Generative AI automatically generates marketing images based on the social media plan finalized by the user and the selected social media template. - The generation process includes detailed data from the plan, such as target customer groups, keywords, color combinations, layout, and emphasis points, and the final marketing image is provided to the system. (2) Support for iterative modifications based on user feedback - Generated marketing images are provided to users in real time, and if users input improvements or modification requests in natural language, the system immediately collects them and supports iterative modification and improvement through stock price prompts. - Rather than simple editing, changes are reflected through the feedback loop of generative AI, enabling continuous modification until the desired result is achieved. (3) SNS account linking and upload schedule management - Provides a user-specific social media content calendar function, enabling users to view the schedule of reserved marketing content at a glance. Users can freely perform actions such as checking the upload status of reserved marketing content, modifying content before uploading, and readjusting upload schedules. (5) Continuous content management and automation support When a user logs in with the same account, it provides a structure that continuously manages and connects previously created marketing content, reservation schedules, and social media integration information based on a database, and automates the regular creation and uploading of marketing content.

[0082] Accordingly, users can easily perform high-quality marketing content creation, repetitive schedule management, and social media optimization uploads with minimal intervention, and can drastically reduce the time, effort, and cost of social media marketing operations.

[0083] As for the details regarding the method of providing marketing automation solutions shown in FIGS. 2 to 4 that are not described, they are identical to or can be easily inferred from the details described above regarding the method of providing marketing automation solutions shown in FIG. 1, so further explanation will be omitted.

[0084] FIG. 5 is a diagram illustrating the process of data transmission and reception between each component included in the generative AI-based multimodal marketing content creation SaaS service system of FIG. 1 according to an embodiment of the present invention. Hereinafter, an example of the process of data transmission and reception between each component will be described through FIG. 5, but the present invention is not to be interpreted as being limited to such an embodiment, and it is obvious to those skilled in the art that the process of data transmission and reception shown in FIG. 5 may be changed according to various embodiments described above.

[0085] Referring to FIG. 5, the solution providing server receives product data (S5100).

[0086] Then, the solution providing server generates a prompt to generate marketing content for a product based on product data (S5200), and calls at least one Generative Artificial Intelligence to generate marketing content (S5300).

[0087] Additionally, the solution providing server enables at least one generative AI to generate marketing content based on a prompt (S5400), and provides the marketing content generated by at least one generative AI to a user terminal (S5500).

[0088] The order of the steps described above (S5100~S5500) is merely an example and is not limited thereto. That is, the order of the steps described above (S5100~S5500) may vary, and some of these steps may be executed simultaneously or deleted.

[0089] As for matters not described regarding the method of providing a marketing automation solution in Fig. 5, they are identical to or can be easily inferred from the description of the method of providing a marketing automation solution in Figs. 1 to 4, so further explanation will be omitted.

[0090] A method for providing a marketing automation solution according to one embodiment described through FIG. 5 may also be implemented in the form of a recording medium containing computer-executable instructions, such as an application or program module executed by a computer. A computer-readable medium may be any available medium accessible by a computer and includes both volatile and non-volatile media, and both removable and non-removable media. Additionally, a computer-readable medium may include all computer storage media. Computer storage media include both volatile and non-volatile, removable and non-removable media implemented by any method or technique for storing information such as computer-readable instructions, data structures, program modules, or other data.

[0091] The method for providing a marketing automation solution according to one embodiment of the present invention described above may be executed by an application basically installed on a terminal (which may include a program included in a platform or operating system, etc., basically installed on the terminal), or by an application (i.e., a program) directly installed by a user on a master terminal through an application providing server, such as an application store server, an application, or a web server related to the service. In this sense, the method for providing a marketing automation solution according to one embodiment of the present invention described above may be implemented as an application (i.e., a program) that is basically installed on a terminal or directly installed by a user, and may be recorded on a computer-readable recording medium such as a terminal.

[0092] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0093] The scope of the present invention is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present invention.< / tiktok> < / instagram> < / instagram> < / instagram>

Claims

Claim 1 A user terminal that inputs product data and outputs marketing content generated based on the product data; and an input unit that receives the above product data; a generation unit that generates a prompt to generate marketing content for the product based on the above product data; a calling unit that calls at least one Generative Artificial Intelligence (GAI) to generate the marketing content; a management unit that causes the at least one Generative AI to generate marketing content based on the prompt; a providing unit that provides the marketing content generated by the at least one Generative AI to the user terminal; a scheduling unit that, considering the characteristics of SNS where continuity is key, when the user terminal uses an automatic upload function, extracts the period and reason for the product of the user terminal to be advertised based on an event DB storing industry-specific seasonal events, saves it to a calendar, generates the above prompt by adding the reason for the product to be advertised during the period the product is to be advertised, generates the marketing content, and generates a monthly content schedule; an automatic upload unit that, when a social media account is entered on the user terminal, uploads the marketing content to the social media account according to the schedule listed in the calendar; when the marketing content is uploaded to social media, calculates the reaction index of the marketing content, and sorts the marketing content in order of highest reaction index A solution providing server including a response reflection unit that extracts a prompt and reflects it in the generation of the next prompt; wherein the product data includes text describing the product, a photo of the product, and target customer information of the product; wherein the marketing content is provided as SaaS (Software as a Service); and wherein the input unit analyzes product description, category, and review data to automatically infer the age group, gender, interests, and consumption patterns of the main customer base of the product.When generating marketing content based on the information regarding the inferred target customer base, the system is configured to adjust the tone, writing style, keywords, and image style; the management unit generates the marketing content based on the prompt through at least one fine-tuned generative AI using a dataset built corresponding to each domain so that marketing content can be generated according to the characteristics of each platform or channel including Instagram, Naver Blog, Facebook, and TikTok; and generates different marketing content for each channel by learning different writing styles, tones, emojis, sentence lengths, hashtags, and link patterns for each channel including Instagram, Naver Blog, Facebook, and TikTok, respectively; the providing unit generates multi-style content by recognizing the industry and time period related to the product, mapping major monthly issues, identifying product characteristics and target information, deriving marketing topic candidates, and converting them based on the prompt in order to derive the core theme, campaign message, and keywords or storyline of marketing, sets themes for each month by industry, generates new marketing content by reflecting trends, and uploads the generated new marketing content; and the event is the creation of marketing content or A generative AI-based multimodal marketing content creation SaaS service system, which includes time-based events that are specific dates or periodically occurring schedules, referring to specific user actions, time conditions, and external circumstances that trigger distribution; product / service-based events including product registration, inventory fluctuations, and surges in popularity; user behavior-based events generated based on user actions; API / external integration events generated based on signals received from external systems; and AI insight-based events generated based on AI analysis results; wherein the automatic upload unit is characterized by generating and uploading information about products or services sold by the user and information about trends together during the upload of the marketing content. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete Claim 6 delete

Images