Generative artificial intelligence-based interactive contents planning solution providing system

Abstract

A system for providing a generative AI-based interactive content planning solution is provided, comprising: a user terminal that inputs keywords, responds to queries generated from keywords, and outputs a planning proposal generated based on keywords, queries, and responses; an input unit that receives keywords from the user terminal; a configuration unit that generates a prompt to branch out stepwise based on a Seed-to-Tree structure using keywords as a Seed, and at least one element required for the planning proposal; a query generation unit that calls a pre-established LLM (Large Language Model) to generate queries and options using the generated prompt; a storage unit that receives and stores a response when the user terminal selects an option; and an integrated management server that generates a planning proposal based on keywords, queries, and responses after loading a pre-established planning proposal template.

Description

Technology Field

[0001] The present invention relates to a generative AI-based interactive content planning solution provision system, and provides a system that generates a query for at least one element required for a planning proposal based on keywords and generates a planning proposal using the response to the query. Background Technology

[0002] Recently, as the drama market has grown explosively, the demand for content continues to rise; however, there is a relative shortage of writers and new ideas capable of stably supplying this content. In this environment, ordinary individuals and novice creators, despite possessing the will to create, face high barriers to entry as they are hindered in the process of developing simple ideas into concrete deliverables such as proposals, synopses, and scripts. Meanwhile, AI-based creative tools focus on mass-producing outputs without user intervention, leading to issues such as similar or standardized content and limitations in fully reflecting the creator's intentions or direction. Consequently, there is a growing need for a new type of planning environment where creators are not merely passive recipients of simple deliverables, but can directly participate in the creative process by selecting and modifying elements during the story's development.

[0003] At this time, a method for generating a story based on data input by a user has been researched and developed. In this regard, prior art Korean Registered Patent No. 10-2640438 (published February 23, 2024) and Korean Published Patent No. 2021-0096749 (published August 6, 2021) respectively disclose a configuration in which artificial intelligence generates characters, events, and sequences based on ideas and loglines input by a user, and automatically generates the sequence of a story by analyzing extreme pattern using multiple neural network models, and a configuration in which artificial intelligence automatically generates a draft of a story or content, or recommends a synopsis and story flow based on user input information.

[0004] However, the former focuses on event recommendations and the completion of dramatic structures, and since AI proactively generates the story progression, it does not disclose a systematic design for the user's narrative selection. The latter adopts a structure that batch generates content centered on generative AI; while this contributes to improving the efficiency of output generation, it merely discloses a configuration where users cannot intervene step-by-step in the narrative formation process. Furthermore, existing AI-based content generation technologies are concentrated on batch generation, limiting their ability to reflect the user's creative intent and narrative structure in stages. Therefore, research and development of interaction-based content planning technologies are required to structurally design the progression from the idea stage to episode-by-episode development, enabling users to participate in the selection and modification processes. The problem to be solved

[0005] One embodiment of the present invention provides a generative AI-based interactive content planning solution system that, when a keyword is entered at a user terminal, generates a query and options for at least one element required for a planning proposal based on a Seed-to-Tree structure using the keyword as a Seed, and generates a planning proposal based on the keyword, query, and response when the user terminal responds by selecting an option, thereby enabling ordinary people who have ideas or stories but were unable to plan to plan a drama, and opening a path for writers whose ideas have run dry or whose stories are blocked to develop their stories. 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 comprises: a user terminal that inputs a keyword, responds to a query generated from the keyword, and outputs a plan generated based on the keyword, query, and response; an input unit that receives a keyword from the user terminal; a configuration unit that generates a prompt to branch out stepwise based on a Seed-to-Tree structure with the keyword as a Seed; a query generation unit that calls a pre-established Large Language Model (LM) to generate a query and options using the generated prompt; a storage unit that receives and stores a response when the user terminal selects an option; and a plan generation unit that loads a pre-established plan template and generates a plan based on the keyword, query, and response. Effects of the invention

[0007] According to any one of the means for solving the problem of the present invention described above, since items necessary for content planning can be systematically organized based on a single idea or theme to automatically generate a planning document, the difficulties that arise when a creator concretizes an idea into a planning document in document form can be effectively resolved. Furthermore, by reflecting the process of the user selecting and modifying the direction of story development in the planning stage, the problem of generating uniform planning documents that occurred in existing AI-based tools can be reduced, and a planning document reflecting the user's intention can be derived. Additionally, even if one is not a professional writer, a planning document can be written more easily and quickly, thereby simultaneously improving the efficiency and completeness of the content planning stage. Brief explanation of the drawing

[0008] FIG. 1 is a diagram illustrating a generative AI-based interactive content planning solution providing system according to one embodiment of the present invention. Figure 2 is a block diagram illustrating an integrated management server included in the system of Figure 1. FIGS. 3 and 4 are drawings for explaining an embodiment in which a generative AI-based interactive content planning solution according to an embodiment of the present invention is implemented. FIG. 5 is a flowchart illustrating a method for providing a generative AI-based interactive content planning 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 generative AI-based interactive content planning solution providing system according to an embodiment of the present invention. Referring to FIG. 1, the generative AI-based interactive content planning solution providing system (1) may include at least one user terminal (100), an integrated management server (300), and at least one tool providing server (400). However, since the generative AI-based interactive content planning solution providing system (1) 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 an integrated management server (300) through the network (200). Also, the integrated management server (300) can be connected to at least one user terminal (100) and at least one tool providing server (400) through the network (200). Additionally, at least one tool providing server (400) can be connected to the integrated management server (300) 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 user terminal that provides keywords using a web page, app page, program, or application related to a generative AI-based interactive content planning solution, receives questions and options corresponding thereto, and provides a response that selects an option. The user terminal (100) may be a terminal that outputs a planning proposal generated based on the response that selected an option.

[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 integrated management server (300) may be a server that provides a generative AI-based interactive content planning solution web page, app page, program, or application. Additionally, the integrated management server (300) may be a server that generates a prompt to branch out stepwise based on a Seed-to-Tree structure using a keyword entered from a user terminal (100) as a Seed, generates a query and a choice using the prompt, and then transmits it to the user terminal (100) to receive a response. Additionally, the integrated management server (300) may be a server that generates a plan based on the received response and transmits it to the user terminal (100).

[0023] Here, the integrated management server (300) can 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.

[0024] At least one tool providing server (400) may be a server that provides a tool called by the integrated management server (300) using a web page, app page, program, or application related to a generative AI-based interactive content planning solution, generates queries and options based on prompts and context provided by the integrated management server (300), and generates a planning proposal based on the response. Here, at least one tool 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 and a web browser.

[0025] FIG. 2 is a block diagram for explaining an integrated management server included in the system of FIG. 1, and FIG. 3 and FIG. 4 are drawings for explaining an embodiment in which a generative AI-based interactive content planning solution according to an embodiment of the present invention is implemented.

[0026] Referring to FIG. 2, the integrated management server (300) may include an input unit (310), a configuration unit (320), a query generation unit (330), a storage unit (340), a planning proposal generation unit (350), a context configuration unit (360), a mixed input unit (370), an infinite generation unit (380), a modification unit (390), and a secondary creation unit (391).

[0027] When an integrated management server (300) or another server (not shown) operating in conjunction with an embodiment of the present invention transmits a generative AI-based interactive content planning solution application, program, app page, web page, etc. to at least one user terminal (100) and at least one tool providing server (400), the at least one user terminal (100) and at least one tool providing server (400) may install or open the generative AI-based interactive content planning 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 tool 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).

[0028] Referring to FIG. 2, the input unit (310) can receive a keyword from the user terminal (100). The user terminal (100) can input a keyword. At this time, the keyword may literally be a word, but it is not excluded that it may be a sentence.

[0029] The component (320) can generate prompts to branch out and expand at least one element required for a plan step by step based on a Seed-to-Tree structure using a keyword as a Seed. A Seed-to-Tree structure refers to a structure in which a Seed representing a single idea or theme is used as an input value to expand the narrative function that the Seed must perform step by step, and as a result, the genre, characters, events, relationships, and components for creating a plan are structured and derived in a tree form. In other words, just as a Seed grows and branches out to become a Tree, it generates a query to fill in at least one element required for the plan step by step, and through the response to this query, branches are created, helping to create a plan, or a tree, overall.

[0030] At this time, the Seed-to-Tree can be configured to ask for elements of [Seed → Genre → Character Occupation → Core Event → Conflict → Crisis → Ending] at each stage, for example. For example, if a Seed is received as input as in FIG. 4a, a [Query-Option] asking for the Genre as in FIG. 4b can be generated to branch out (branch expansion). In order for the LLM to generate this, the component (320) must create a prompt, and the prompt can be structured using the Seed and which stage (Genre) among [Seed → Genre → Character Occupation → Core Event → Conflict → Crisis → Ending] it is currently at.

[0031] step Processing process tool or model 1. Receive Seed Receive a single keyword or short sentence from the user and set it as the story route. Receive a single keyword or short sentence from the user and set it as the story route. 2. Interpretation of the meaning of Seed Analyzes the emotion, theme, and narrative type of the input Seed keyword. Natural Language Semantic Analysis (NLP Embedding), Keyword Classification Models, Vector Embeddings 3. Story Stage Determination Determines the depth of the tree where the current user is located. Step State Machine, depth index management 4. Determine the expansion node type Determines the types of nodes to be expanded, such as genre, occupation, event, and crisis, based on the current depth. Rule-based routing engine, condition mapping table 5. Apply tree expansion rules Filters only logically connectable candidates by referring to the Seed and previous selection history. Graph rule engine, dependency filtering logic 6. Structuring prompts Assembles prompts with a "Seed → Selection History → Expansion Node Type" structure Prompt template engine (similar to Jinja2), JSON Schema 7. Generating Query-Options Generates one question and multiple options based on a structured prompt Large Language Model API 8. Verification of result consistency Verify that the generated result matches the current node type. Output validation logic, keyword matching, forbidden rule filter 9. Store tree nodes Stores choices and results in the form of tree nodes Tree data model, PostgreSQL, or graph DB 10. Update branch path Determines the active branch based on user selection Tree search algorithm, active path pointer 11. Prepare for the next expansion Move to the next depth and determine the expansion node type again. Loop control logic, stage transition controller

[0032] In this Seed-to-Tree structure, a Node refers to a genre, occupation, event, crisis, etc., an Edge refers to a choice (response) selected by the user, a Root refers to a Seed keyword entered by the user, and a Depth refers to a stage of story progression.

[0033] The query generation unit (330) can generate queries and options using a prompt generated by calling a pre-established LLM (Large Language Model). For example, if the Seed is [Poverty Abandoned Me] and the current stage is the stage of selecting [Genre], a prompt based on [Poverty Abandoned Me]+[Genre] is generated as follows and input into the LLM, after which queries and options can be generated.

[0034] Prompt You are a Story Design Assistant supporting drama planning. The Seed currently entered by the user is as follows: Seed: "Poverty Abandoned Me." The current stage is the 'Genre Selection' stage. Characters, worldview, events, conflicts, crises, and the ending have not yet been decided. Please strictly adhere to the following conditions: 1. Propose a genre centered on expanding the sentiment and meaning of the Seed. 2. Write a short and intuitive description of the genre. 3. Propose based on main genre candidates rather than a single genre. 4. Exclude genres that are excessively fantastical or unrelated to the realism of the Seed. 5. Create 6 to 8 genre options for the user to choose from. 6. Explain in a single sentence how the genre connects to the Seed along with the genre name. Please write the output format as follows: [Question] Write in one sentence [Option] 1. Genre Name - Brief Description 2. Genre Name - Brief Description... Now, create the query and options for genre selection based on the above conditions.

[0035] An LLM that receives such a prompt can generate query-options as shown in Table 3 below.

[0036] Q: What genre of emotions and atmosphere would be best for this story to unfold? ▶ Options 1. Human Drama – A story centered on the process of a character wounded by poverty rebuilding their life. 2. Social Drama – A story that realistically illuminates the reality individuals face amidst wealth inequality and structural disparities. 3. Coming-of-Age Drama – A story focusing on the internal transformation of a character finding their true self, using poverty as a starting point. 4. Melodrama – An emotion-driven story where love and relationships are shaken by poverty. 5. Crime Drama – A suspenseful story where poverty drives characters to dangerous choices. 6. Thriller Drama – A story dealing with the psychology and crises of characters pushed to the limit for survival. 7. Black Drama – A story that portrays the reality of poverty with a cynical and sharp perspective.

[0037] In this case, the Seed is [Poverty Abandoned Me], and emotional keywords such as loss, exclusion, survival, social structure, anger, or resignation are automatically extracted, and genres can only be suggested within this range of emotions. Therefore, romantic comedy, fantasy adventure, and teen school dramas may be excluded. Of course, since it is not impossible for the emotion of poverty to be combined with romance, fantasy, or teen dramas, recommendations can be made while keeping all such possibilities open. If the steps for determining the genre from the Seed are organized in order, they may be as shown in Table 4 below. This process can be repeated until each node (element) of [Seed → Genre → Character Occupation → Core Event → Conflict → Crisis → Ending] is filled, and once information for all nodes is filled, a project proposal can be generated based on this.

[0038] step Task details 1. Enter Seed Poverty abandoned me 2. Sentiment Analysis Deficiency, loss, sociality 3. Filtering Genre Candidates Reality-based genres first 4. Create prompt Includes genre suggestion conditions 5. LLM call Generating queries and options 6. User Selection Genre confirmed

[0039] At this time, after selecting a genre from [Seed → Genre → Character Occupation → Core Event → Conflict → Crisis → Ending], a context such as a dialogue log or choice is formed. Therefore, when generating a prompt to be inserted into the LLM, or by adding a context along with the prompt, the context can be maintained or considered to select a query-choice. This will be described later in the context configuration section (360).

[0040] The storage unit (340) can receive and store a response when the user terminal (100) selects an option. The user terminal (100) can respond to a query generated from keywords. As will be described later, the response is not limited to multiple-choice responses where an option is selected, but also includes subjective responses such as FIG. 4i, which are entered directly by the user terminal (100). All of these responses can be stored.

[0041] The planning proposal generation unit (350) can generate a planning proposal based on keywords, questions, and answers after loading a pre-established planning proposal template. The user terminal (100) can output the planning proposal generated based on keywords, questions, and answers. The planning proposal template may include a work outline, planning intention, worldview setting, main characters, character relationship diagram, and episode composition, but is not limited to those listed, and is not excluded for reasons not listed.

[0042] Seed-to-Tree output step Output data Proposal Creation Items Correspondence generation method Seed User input keywords User input keywords Sentences based on the core theme of the work 1. Genre Main Genre, Secondary Genre, and Tone Information Work Overview, Planning Intent Automatic generation of sentences explaining genre combinations 2. Character Occupation Occupations and roles of the protagonist and main characters Main characters Character Profile Template Mapping 3. Key Event The central event that triggers the story Planning Intent, Synopsis Create a case summary paragraph 4. Conflict Conflict structure between characters or values Character relationship chart, plot structure Sentences of conflict relationships 5. Crisis Episode Structure (Summary of the latter half) Episode Structure (Summary of the latter half) Summary of Tension Structure 6. Ending The direction of the story or the result of the judgment Summary of overall structure, ending direction Ending character description sentence

[0043] First, the work overview can be generated as a paragraph summarizing the theme and genre of the work using ① Seed (keyword) and genre information as input; ② the planning intent can be generated by describing the meaning of the theme and events, centering on “why this story is being told,” using the Seed, key events, and genre tone as input; ③ the world-building can be written by inferring the background environment where events might occur and formulating whether it is based on reality or fiction, using the genre, character occupations, and conditions for event occurrence as input; ④ key characters can be generated by mapping character occupations to a character template centered on occupation and role to create key character items. Additionally, ⑤ the character relationship diagram can be organized into confrontation, cooperation, and emotional relationships by structuring the subject-object relationship of the conflict using conflict as input. ⑥ the episode structure can provide an early-middle-late structure by breaking down the narrative flow using key events, conflicts, crises, and the conclusion as input.

[0044] The Seed-to-Tree output becomes "story structure data," and the proposal is generated by mapping that structure data to a document template.

[0045] Of course, in addition to the methods mentioned above, a proposal can be generated using various other methods.

[0046] In this case, the LLM that generates prompts, the LLM that generates query-choice options, the LLM that generates planning documents, and the LLM that generates the script to be described later are all broadly considered LLMs, but since their respective characteristics are all different, different LLMs can be used for each.

[0047] division Role performed Required ability Recommended model tendencies Reasons for use Prompt generation LLM Prompt assembly reflecting rules and structure Understanding of structure, logical accuracy, adherence to rules Logically stable LLMs (e.g., GPT-based reasoning, Gemini Pro-based) Since prompt errors cause the entire pipeline to collapse, accuracy is more important than degrees of freedom. Query and Option Generation LLM Generating questions and options to be presented to the user Diversity, idea expansion, branching thinking Divergent LLM (a model with strong creativity) Since the key is generating options without repetition, creative divergence ability is important. Proposal Creation LLM Structured data-based documentation Summarization ability, stylistic stability, length control Document-focused LLM (Long Text Stable Type) The proposal requires a refined document tone without structural changes. Title / Script Generation LLM Creation of Emotion, Rhythm, and Dialogue Emotional expression, dialogue rhythm, immersion LLM specializing in creative writing (a model strong in creative writing) For the title and script, emotion and immersion are more important than rules.

[0048] For example, when generating prompts, you can use the GPT-4 family / Gemini Pro, which is stable in JSON, rules, and structure; when generating query options, you can use the GPT-4 Turbo / Claude family, which has excellent divergence; when generating proposals, you can use GPT-4.1 / Gemini 1.5 Pro, which is stable in long texts; and when generating titles or scripts, you can use the Claude Opus / Creative model, which has strengths in emotion and dialogue. The model names presented here are examples, and since the important thing is the separation of tendencies, it is advisable to use a model that fits the purpose and tendency of each LLM.

[0049] Rule Engine ↓ Prompt Generation LLM (Accuracy) ↓ Query / Option Generation LLM (Divergence) ↓ Structure Data Finalization ↓ Proposal Generation LLM (Refinement) ↓ Title / Script LLM (Appraisal)

[0050] The context configuration unit (360) can configure the selection of the previous stage, the current stage, and the conversation log as a context when generating queries and options if a previous stage exists in the Seed-to-Tree stage. At this time, the context can be used together with the prompt when generating a query in the query generation unit (330). Here, a context is a structured bundle of selection results and state information that has already been determined during the Seed-to-Tree process. This context includes the following information.

[0051] Context components explanation example ① Previous Step Selection Information - Genre selected by the user - Already confirmed character occupation - Settings selected in the previous step Example: Genre = Social Drama Character Occupation = Delivery Driver ② Current Stage Information -Currently in progress Seed-to-Tree stage- Purpose of creation in this stage Example: Current Stage = Core Event Selection Goal = Deriving events that align with the character's occupation and genre ③ Seed Information - Initial entered Seed keyword Example: Seed = "Poverty abandoned me" ④ Conversation Log Summary Information - Selection tendencies shown by the user in previous queries - Selection types repeatedly excluded - History of modification requests ※ Summarized feature information, not the full original conversation Example: Users prefer realistic settings, reject repetition of fantasy elements, and prefer emotion-centered plot developments. Context ={ Seed, Previous step selection result, Current step identifier, User selection tendency summary}

[0052] This information is organized into a Context Bundle, which can be used as constraint information entered along with the prompt when generating the prompt ([[Prompt+Context]→Generate Query and Options]). Examples of contexts may be as follows.

[0053] context Seed: Poverty Abandoned Me Selected Genre: Social Drama Selected Character Occupation: Delivery Driver User Preference: Prefers Reality-Based Settings Current Stage: Key Event Context-reflecting query "What kind of incident would suit the profession of a delivery driver and the social drama genre?"

[0054] Accordingly, the context maintains consistency to avoid conflicts with previously established settings, connects steps to maintain the Seed-to-Tree flow, improves selection quality by preventing random choices, and enables real-time reflection of user intent by immediately applying changes without iterative learning. Mathematically defining the Seed-to-Tree structure, including this context, is as follows.

[0055] <Seed-to-Tree 구조>

[0056] The Seed to Tree structure is a structure that starts from a single input keyword and gradually expands story choices into a tree structure.

[0057] First, the key variables are defined as shown in Table 11 below.

[0058] sign definition S0 User input Seed keyword d Tree depth, story progression stage Dmax Maximum allowable depth Nd Set of nodes at depth d bd Branching factor at depth d nd,i The i-th node at depth d Ed Selected edge set at depth d Pd User-selected path up to depth d T Full story tree

[0059] <Tree Structure Formula Definition>

[0060] First, the root node can be defined as N0={S0}. At depth 0, there is only one Seed node, and all story expansion is performed based on that node.

[0061] The node at the next depth d+1 from the next depth d can be generated as shown in Equation 1 below (step-by-step expansion structure).

[0062]

[0063] In this case, Expand(·) is the Seed to Tree expansion function. The expansion function can be defined as shown in Equation 2 below.

[0064]

[0065] Here, bd is the number of choices generated at depth d, which can be a fixed or variable value. For example, the genre level b1=6, the job level b2=8, the event level b3=10, and the crisis level b4=6, but are not limited to these.

[0066] Each depth has a different meaning and can be as shown in mathematical formula 3 below.

[0067]

[0068] Depth d Node type 0 Seed 1 Genre 2 Character Occupation 3 Key events 4 conflict 5 crisis 6 end

[0069] In this case, extension is defined not as a simple branch, but as a path-dependent function (context-dependent extension function).

[0070]

[0071] Here, Pd is a user-selected path up to depth d, where Pd={n0,n1,...,nd}, and can use not only the current node but also the entire previous selection as a condition. That is, even for the same event node, if the previous genre and job selections are different, completely different expansion results can be generated.

[0072] At this time, the total number of nodes that can be generated is equal to the following mathematical formula 5.

[0073]

[0074] This refers to the size of the entire narrative space.

[0075] However, since it is assumed that the user selects options one by one (single path), a path such as the following mathematical formula 6 can be generated.

[0076]

[0077] In this case, D≤Dmax.

[0078] An example of how the tree expands based on the process described above is as follows.

[0079] <Example>

[0080] step Seed input "plural" This single word becomes the root node of the tree. Depth 0 (root) Node: n0= "plural" At this point, there is no story yet. Only a direction exists. Depth 1: Genre Expansion ▶ The system applies the extended rule "The first step must suggest a genre" according to the rule* User choice: Crime selection Generated Nodes n11: Crime n12: Romance n13: Fantasy n14: Thriller n15: Historical n16: School Tree Partial Revenge ├─ Crime ├─ Romance ├─ Fantasy ├─ Thriller ├─ Historical Drama └─ School Depth 2: Character Job Expansion Current Active Path Multiple → Crime ▶ Rule: If the genre is Crime, job candidates are limited to role groups suitable for crime narratives* User Selection: Former Organization Member Generated Nodes n21: Detective n22: Former Member n23: Whistleblower n24: Lawyer n25: Detective n26: Reporter Tree Partial Revenge └─ Crime ├─ Criminal ├─ Former Member ├─ Whistleblower ├─ Lawyer ├─ Detective └─ Reporter Depth 3: Key Event Expansion Current Active Path Multiple → Crime → Former Member ▶ Expansion Rule: Past connection with the organization required, includes betrayal or sacrifice incident Generated Node n31: Colleague dies from internal betrayal n32: Organization eliminates family n33: Framed and abandoned n34: Becomes a test subject for the Organization n35: Identity exposed while in hiding Tree Partial Revenge └─ Crime └─ Former Member └─ Key Case ├─ n31 ├─ n32 ├─ n33 ├─ n34 └─ n35 Depth 4: Conflict Expansion Current Active Path: Revenge → Crime → Former Member → Family Murder ▶ Expansion Rule: The conflict must take the form of a clash between two values, and the conflict must shake the protagonist's identity (Former Member → Human). Create Node n41: Infiltrates the organization again for revenge n42: Receives an offer to replace the organization boss n43: Discovers the target of revenge is the benefactor's son n44: Police pursue simultaneously Tree Partial Family Murder└─ Conflict ├─ n41 ├─ n42 ├─ n43 └─ n44 Depth 5: Crisis Expansion Current Active Path: Revenge → Crime → Former Member → Family Murder → Infiltration ▶ Expansion Rules: The crisis must be a situation where one cannot return to the state prior to the choice; the crisis must be a physical, legal, or survival risk rather than abstract anxiety; and the situation must worsen as the decision is delayed. Generated Node n51: Identity revealed, life threatened n52: Target of revenge moves first n53: Colleague betrays Tree Partial Conflict Choice └─ Crisis ├─ n51 ├─ n52 └─ n53 Depth 6: Ending Branch Current Active Path: Multiple → Crime → Former Member → Family Murder → Crisis ▶ Expansion Rules: The ending must provide an answer to the Seed question, must leave an emotional aftertaste, the final output is confirmed to have only one ending, and unselected ending nodes remain inactive. Create Node n61: Disappears after successful revenge n62: Abandons revenge and exposes the truth n63: Revenge is successful but he is also arrested Tree Partial Crisis└─ Ending ├─ n61 ├─ n62 └─ n63 Summary of overall structure Revenge → Crime → Former Member → Family Killed by the Organization → Infiltrate the Organization → Risk of Identity Exposure → Successful Revenge Seed (Revenge) ↓Genre (Crime) ↓Character (Former Member) ↓Key Event (Why the Revenge Happened) ↓Conflict (Revenge vs. Other Values) ↓Crisis (Irreversible) ↓Ending (Answer to the Question of Seed)

[0081] In this case, while one embodiment of the present invention is configured based on only one option, two or more options may be selected. For example, there may be one main genre, but there may be auxiliary genres, and there may be two or three auxiliary genres. For example, the drama "When the Camellia Blooms" is a drama that combines romantic comedy, mystery, thriller, human drama, and family drama, and does not use only one genre. In this way, while the worldview is one, multiple emotions and expressions can be controlled. However, regulations to prevent genre conflicts can be set, such as prohibiting supernatural events or the use of magic if the main genre is a thriller.

[0082] division in charge Genre Definition rule engine Genre Conflict Judgment Filtering allowed combinations Case Type Range Emotional Interpretation Artificial intelligence Dialogue, description, sentences Creative variations User Selection ↓Rule Engine ├─ Check Main Genre ├─ Determine Auxiliary Genre Allowance ├─ Remove Prohibited Combinations ├─ Determine Event Type Scope ↓AI ├─ Within Allowed Scope ├─ Create Emotions, Expressions, and Situations ↓Output

[0083] Of course, in addition to genre, multiple options can be selected for each stage. If multiple choices are allowed, an additional process may be added to require setting which one to use as the main option and verifying that there are no points of conflict between these choices.

[0084] The mixed input unit (370) can provide a hybrid interface that combines multiple-choice options and subjective options, in addition to allowing the user terminal (100) to select options when providing options to the user terminal (100) and receiving a response, by directly receiving text input from the user terminal (100). That is, in addition to the options generated by the LLM, the user can input what they want.

[0085] The infinite regeneration unit (380) can generate options continuously by looping when the user terminal (100) requests regeneration after providing options to the user terminal (100). That is, if the direction desired by the user does not appear, options can be continuously provided, and the user's intention can be identified so that options submitted by the user do not appear duplicated. Alternatively, options can be generated by receiving instructions directly from the user.

[0086] To this end, the infinite regeneration unit (380) can block duplicate generation by using a Redis-based option history store and a SHA-256 hash comparison logic to identify previously provided options, and a Sentence-BERT (SBERT) based intent similarity analysis model can infer the direction of the user's intent by vectorizing options skipped by the user and instructions entered directly. Subsequently, the prompt reconstruction engine (Jinja2 template) reassembles the prompt by reflecting the inferred intent vector and exclusion conditions, and receives the reconstructed prompt from an LLM, e.g., a GPT-4 family language model, to repeatedly generate new queries and options within a loop structure.

[0087] The modification unit (390) provides a history rollback function to allow the user terminal (100) to return to a previous step and modify the selection, and can regenerate the question based on the modified selection when the selection is modified in the previous step. For the history rollback, the state history can be stored in PostgreSQL (JSONB) or Redis (In-Memory Cache), the history can be managed using an Event Sourcing method or Step Index-based state version management, and for the rollback, State Rehydration logic or a Diff-based invalidation processing method can be used.

[0088] The secondary creation department (391) can use a generative AI that has been built based on the project proposal to recommend a title or generate a script for each episode.

[0089] Narrative Stability Assessment

[0090] In addition, one embodiment of the present invention may evaluate whether the generated tree is too chaotic, whether the emotional line is broken, or whether the genre tone is unstable using a quantitative score, and if it is below the standard, the tree expansion may be recreated. At this time, the integrated management server (300) may analyze narrative consistency by vectorizing the emotional flow, semantic similarity, and genre keyword distribution of the script generated by the Sentence-BERT (all-MiniLM-L6-v2) based sentence embedding model, and the weighted sum score calculation algorithm may calculate a narrative stability score by combining the emotional continuity score, the genre tone maintenance score, and the event connectivity.

[0091] In this case, ① the emotional continuity score is calculated by a Sentence-BERT (all-MiniLM-L6-v2) based sentence embedding model vectorizing the emotional expressions of the previous and current narrative nodes, respectively, and then comparing the semantic similarity between the two vectors. At this time, considering both the direction and intensity of emotional change, a high score is assigned when the emotional flow changes gradually, and a low score is assigned when sudden emotional discontinuity or a shift in opposite emotions occurs. Of course, continuity can be measured using various methods in addition to this one.

[0092] Here, a narrative node is a unit of information that has independent meaning in the development of the narrative, such as an event, choice, conflict, crisis, or conclusion, and in a Seed-to-Tree structure, it may be as shown in Table 15 below.

[0093] Seed └─ Genre Node └─ Character Node └─ Key Event Node ← Narrative Node └─ Conflict Node ← Narrative Node └─ Crisis Node ← Narrative Node └─ Ending Node ← Narrative Node

[0094] In other words, genre and characters can be classified as setting nodes, while events, conflicts, crises, and endings can be classified as narrative nodes. Narrative nodes possess the attributes of having a single meaning (e.g., “family murdered,” “identity revealed,” “deciding to seek revenge”), being connected to previous nodes (results of previous events, causes of subsequent actions), inducing the creation of subsequent nodes (e.g., expandable to conflicts, crises, and endings), and being able to be analyzed independently. Accordingly, narrative nodes are used as the basis when calculating each score.

[0095] ② The Genre Tone Maintenance Score can be calculated by determining the distribution of genre-related expressions included in each narrative node using a predefined genre keyword dictionary and a TF-IDF-based keyword distribution analysis technique, and then comparing how closely that distribution matches the currently active genre layer. In this case, a high score is assigned if the appearance ratio of keywords corresponding to the selected main and secondary genres is maintained above a certain level, while the score may decrease if non-genre expressions increase excessively. Of course, the Genre Tone Maintenance Score can be calculated using various methods other than this one.

[0096] ③ Event connectivity is calculated by SBERT or RoBERTa-based event embedding models vectorizing the event subjects, actions, and objects included in the previous and current narrative nodes, and comparing the semantic continuity between the two events. In this case, a high connectivity is assigned when there are consecutive actions by the same person, an event progression where causal relationships are maintained, or when the result of a preceding event acts as the cause of a subsequent event; conversely, a low connectivity may be assigned when the event subjects are disconnected or causal relationships are unclear. Of course, event connectivity can be calculated using various methods other than this one.

[0097] Narrative Overlap Detection

[0098] The integrated management server (300) calculates a similarity by comparing the event development structure of an already generated tree with the narrative path of an event development currently being generated, and operates to restrict expansion in the same development direction if the similarity exceeds a preset threshold. To this end, an event sequence vectorization model converts the event flow of each narrative node into a vector form, and a tree path similarity calculation logic can compare the structural similarity between the existing tree path and the currently active path. Subsequently, if the similarity is high, the graph rule engine performs branch control to prioritize the application of alternative expansion rules, thereby inducing narrative expansion in a new development direction rather than repetitive story generation.

[0099] This prevents the problem of repeating similar stories and allows for the maintenance of continuous freshness even in platform environments where a large amount of content accumulates. In this context, a graph rule engine refers to a rule-based processing module that manages the connection relationships between nodes generated in a Seed-to-Tree structure in a graph format, filters potential nodes to be generated next based on each node's attributes, parent-child relationships, and pre-configured expansion rules, and controls the selective expansion of only acceptable paths.

[0100] Hereinafter, the operation process according to the configuration of the integrated management 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.

[0101] Referring to FIG. 3, (a) the integrated management server (300) sets up a Seed-to-Tree structure, and when a Seed is input from a user terminal (100) as in (b), it can provide a query and options as in (c) in response. Once the user's first response is completed, the integrated management server (300) generates queries and options by considering the context to take into account the previous response as in (d). Meanwhile, the integrated management server (300) can apply a hybrid method that allows subjective responses from the user terminal (100) and provide an infinite regeneration function that regenerates options as in (b). Additionally, the integrated management server (300) can provide an opportunity to modify the selection by removing the user's selection and returning to the previous state using a history rollback function as in (c). Furthermore, the integrated management server (300) can generate a plan based on the response as in (d) and generate secondary works such as titles and scripts based on this. In addition, the script generated in this way can be converted into a video storyboard by combining AI voice and text-to-image generative AI.

[0102] As for the details regarding the method of providing a generative AI-based interactive content planning solution 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 a generative AI-based interactive content planning solution shown in FIG. 1, so further explanation will be omitted.

[0103] FIG. 5 is a diagram illustrating the process of data transmission and reception between each component included in the generative AI-based interactive content planning solution providing 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.

[0104] Referring to FIG. 5, the integrated management server receives a keyword from a user terminal (S5100) and generates a prompt to branch and expand at least one element step by step based on a Seed-to-Tree structure using the keyword as a Seed (S5200).

[0105] Additionally, the integrated management server calls the existing Large Language Model (LLM) to generate a query and options using the generated prompt (S5300), and when a user terminal responds by selecting an option, it receives and stores the response (S5400).

[0106] Then, the integrated management server loads a pre-built planning template and generates a planning proposal based on keywords, queries, and answers (S5500).

[0107] 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.

[0108] As for the details regarding the method of providing a generative AI-based interactive content planning solution of Fig. 5 that are not described, they are identical to or can be easily inferred from the details described above regarding the method of providing a generative AI-based interactive content planning solution through Figs. 1 to 4, so further explanation will be omitted.

[0109] A method for providing a generative AI-based interactive content planning 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, as well as removable and inseparable media. Additionally, a computer-readable medium may include all computer storage media. Computer storage media include both volatile and non-volatile, removable and inseparable media implemented by any method or technique for storing information such as computer-readable instructions, data structures, program modules, or other data.

[0110] The method for providing a generative AI-based interactive content planning solution according to one embodiment of the present invention described above may be executed by an application that is basically installed on a terminal (which may include a program included in a platform or operating system, etc., that is basically installed on the terminal), or it may be executed by an application (i.e., a program) that a user directly installs 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 generative AI-based interactive content planning 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.

[0111] 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.

[0112] 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. Industrial applicability

[0113] The solution according to one embodiment of the present invention is not limited to drama planning but can be applied to various fields characterized by deriving structured results by progressively expanding a single idea. For example, it can be used in planning story-based content such as webtoons, web novels, and animations; designing game scenarios and quests; configuring educational scenario content; planning brand storytelling and marketing campaigns; designing interactive content; writing simulation scenarios; and designing customer response scenarios or consultation flows. In particular, since complex development structures can be systematically designed through step-by-step selection and rule-based expansion, it can be widely used in various industrial sectors requiring stories, procedures, and scenarios.

Claims

Claim 1 A user terminal that inputs keywords, responds to queries generated from the keywords, and outputs a plan generated based on the keywords, queries, and responses; and an input unit that receives keywords from the user terminal; a configuration unit that generates prompts to branch out stepwise based on a Seed-to-Tree structure using the keywords as a Seed, thereby expanding genres, characters, events, relationships, and components required for the plan; a query generation unit that generates the query and options using the generated prompts by calling a pre-established LLM (Large Language Model); a storage unit that receives and stores the response when the user terminal provides a response selecting an option; a plan generation unit that generates a plan based on the keywords, query, and response after loading a pre-established plan template; a context configuration unit that configures the selection of the previous stage, the current stage, and the conversation log as a context when generating the query and options, if a previous stage exists in the stages of the Seed-to-Tree; a mixed input unit that provides a hybrid interface that combines subjective input, which receives text directly from the user terminal in addition to allowing multiple-choice selection when providing options to the user terminal and receiving a response; and a user terminal that requests regeneration after providing options to the user terminal. In the case of an integrated management server comprising: an infinite regeneration unit that continuously generates the above-mentioned options while looping; a history rollback function that allows the user terminal to return to a previous step and modify the selection, and a modification unit that regenerates the question based on the modified option if the option is modified after returning to the previous step; and a secondary creation unit that recommends a title and generates a script for each episode using a generative AI built in advance based on the above-mentioned plan;It includes, and the above-mentioned planning template comprises: a work overview generated as a work overview paragraph by summarizing what theme the work deals with and in what genre using the above-mentioned Seed and genre information as input; a planning intent generated by describing the meaning of the theme and event, focusing on why this story is being told, using the above-mentioned Seed, key event, and genre tone as input; a worldview setting written by inferring the background environment where an event can occur using the genre, character occupation, and event occurrence conditions as input, and formulating whether it is based on reality or virtuality; key characters, where a main character item is generated by mapping the character occupation to an occupation or role-centered character template using the character occupation as input; a character relationship diagram organized into confrontation, cooperation, and emotional relationships by structuring the theme and object relationship of the conflict using the conflict as input; and an episode composition that provides an beginning, middle, and end structure by decomposing the flow of development using the key event, conflict, crisis, and conclusion as input; and the above-mentioned context is used together with the above-mentioned prompt when generating the above-mentioned query in the above-mentioned query generation unit, and includes previous stage selection information including the genre selected by the user, already confirmed character occupation, and settings selected in the previous stage, the currently ongoing Seed-to-Tree stage, and the generation of this stage. A generative AI-based interactive content planning solution providing system comprising current stage information including a purpose, seed information including an initially entered seed keyword, and conversation log summary information including selection tendencies shown by the user in previous queries, selection types repeatedly excluded, and modification request history, wherein the infinite regeneration unit is characterized by identifying the user's intent and continuously providing options so that options passed by the user do not appear duplicated. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete Claim 6 delete Claim 7 delete

Images