A creative content generation method, system, medium and program product based on AIGC

Abstract

The present application discloses a creative content generation method, system, medium and program product based on AIGC, which relates to the field of artificial intelligence technology. The method includes determining the theme and style of the creative content, and performing preliminary content creation in combination with target audience information to generate a preliminary creative content framework; performing emotional enrichment and creativity enhancement processing on the preliminary creative content framework, and obtaining the processed creative content framework and the corresponding creative element set; based on the historical combination data of different creative elements in the creative element set, in an optimized content generation model guided by creative effects, determining the optimal creative content generation strategy according to the innovation factor correlation coefficient of the creative elements and the corresponding creative optimization strategy; based on the optimal creative content generation strategy, using the optimized content generation model to intelligently generate the final creative content; while ensuring the efficiency of content generation, the present application improves the personalization level of the generated content to meet the diverse needs of different user groups.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a creative content generation method, system, medium and program product based on AIGC. Background Art

[0002] ‌AIGC (Artificial Intelligence Generated Content)‌ refers to the method of automatically generating content using artificial intelligence technology, including text, images, audio, video and other forms; through deep learning technology, computers can better understand the meaning and structure of language, thereby realizing automated content generation; current artificial intelligence content generation technology has been widely used in text, pictures, videos and other fields, greatly improving the efficiency of content production.

[0003] However, with the growing number of Internet users and the continuous improvement of personalized needs, the existing general content generation methods are difficult to meet users' pursuit of content personalization. Therefore, how to improve the personalization level of generated content while ensuring content production efficiency and meet the diverse needs of different user groups, thereby enhancing the attractiveness and dissemination of content, is the focus of the current AIGC-based creative content generation technology that urgently needs to be improved and researched. Summary of the invention

[0004] In order to ensure the efficiency of content generation while improving the personalization level of generated content and meeting the diverse needs of different user groups, the present application provides a creative content generation method, system, medium and program product based on AIGC.

[0005] In the first aspect, the invention objective of the present application is achieved by adopting the following technical solutions:

[0006] A creative content generation method based on AIGC, comprising:

[0007] Obtain information on creative needs, creative style preferences and target audiences;

[0008] Analyze the creative demand information and creative style preferences, determine the theme and style of the creative content, and conduct preliminary content creation based on the target audience information to generate a preliminary creative content framework;

[0009] Performing emotional enrichment and creativity enhancement processing on the preliminary creative content framework, and obtaining the processed creative content framework and a corresponding creative element set; the creative element set includes a plurality of creative elements and innovation factor correlation coefficients between the creative elements;

[0010] Based on the historical combination data of different creative elements in the creative element set, creative effect-oriented training optimization is performed in a preset content creation model to obtain an optimized content generation model;

[0011] In the optimized content generation model, according to the innovation factor correlation coefficient of the creative elements and the corresponding creative optimization strategy, the best creative effect index of each creative element at different use stages is determined, and the optimal creative content generation strategy is formulated;

[0012] According to the optimal creative content generation strategy, the final creative content is intelligently generated using an optimized content generation model.

[0013] By adopting the above technical solutions, in order to improve the personalization level of generated content and meet the diverse needs of different user groups, the user's creative demand information, creative style preferences and target audience information are first collected to ensure that the generated creative content can accurately meet the user's specific needs and expectations. After determining the theme and creative style, a preliminary creative content framework is generated to help eliminate inappropriate creative directions in the early stages and reduce the workload of adjusting the later creative content. At the same time, the sentiment analysis model and creative stimulation module are used to enrich the sentiment and enhance the creativity of the preliminary creative content framework to enhance the emotional expression and innovation of the creative content. In addition, the obtained creative element set and its innovation factor correlation coefficient provide a quantitative basis for subsequent optimization. The historical combination data of creative elements is trained and optimized to improve the adaptability and flexibility of the content generation model, so that it can better cope with diverse creative tasks. The optimal creative content generation strategy is specified through the best creative effect indicator to facilitate the evaluation of AIGC-based creative content and the improvement of the creative process. The creative effect of the processed creative content framework is quantitatively evaluated and iteratively optimized and adjusted, and the optimized content generation model is used to intelligently generate the final creative content, thereby achieving efficient and high-quality content output, reducing the cost of manual intervention, and improving overall creative efficiency. The application of the technical solution of this application is conducive to improving the personalization level of generated content while ensuring the content generation efficiency, so as to meet the diverse needs of different user groups.

[0014] In a preferred example of the present application, the emotional enrichment and creativity enhancement processing of the preliminary creative content framework specifically includes:

[0015] Applying a preset sentiment analysis model to analyze the sentiment components in the preliminary creation content framework, and identifying positive sentiment words, negative sentiment words, and neutral sentiment words;

[0016] Using the creative stimulation module to add or replace specific words to the preliminary creative content framework to introduce novel concepts or expressions, and to evaluate the emotional coherence and rhythm between the various creative content parts to obtain preliminary creative evaluation results;

[0017] Based on the preliminary creative evaluation results and in combination with the collected user historical behavior data, the tone intensity and emotional tendency in the preliminary creative content framework are adjusted, and the creative elements used and their innovation factor correlation coefficients are recorded.

[0018] By adopting the above technical solution, the emotional components in the preliminary creative content framework are first analyzed, so that the preliminary creative content framework can accurately convey the expected emotional color, which is conducive to enhancing the emotional resonance of the work; by utilizing the creative addition module to introduce novel concepts or expressions, and evaluate the emotional coherence and rhythm between the various creative parts, the innovation and attractiveness of the work are effectively improved; at the same time, the preliminary creative content framework is emotionally enriched and creatively enhanced, which reduces the need for manual intervention and speeds up the creative process; at the same time, the tone intensity and emotional tendency are adjusted according to the user's historical behavior data to achieve personalized customization and further improve the creation efficiency.

[0019] In a preferred example of the present application, the creation effect-oriented training optimization is performed in the preset content creation model to obtain an optimized content generation model, specifically including:

[0020] Calculate the usage frequency of each creative element in historical creations and the probability of occurrence of its combination pattern based on the historical combination data of different creative elements in the creative element set;

[0021] Based on the occurrence probability and creative effect feedback information, the preset content creation model is optimized using a deep reinforcement learning algorithm; the optimal creation path is explored by simulating the reward mechanism of different creative element combinations in a real creative environment and combining a generative adversarial network;

[0022] Based on the acquired external evaluation information and a predefined set of quality indicators, the Bayesian optimization algorithm is used to dynamically adjust the model parameters of the content creation model to obtain an optimized content generation model.

[0023] By adopting the above-mentioned technical solution, the usage frequency and occurrence probability in the historical combination data of creative elements are calculated, and combined with the creative effect feedback information, the deep reinforcement learning algorithm is used to optimize the content creation model, ensuring the optimal combination mode between different creative elements and improving the overall creation quality; the Bayesian optimization algorithm is combined with the adversarial generative network to explore the optimal creation path, which not only considers the internal logical structure, but also takes into account external evaluation indicators (such as click-through rate, number of shares, etc.), comprehensively guarantees the quality and dissemination effect of the work, and adjusts the tone intensity and emotional tendency of the creative content according to the user's historical behavior data, so that the generated content is more in line with the preferences of the target audience, enhancing the pertinence and applicability of the work.

[0024] In a preferred example of the present application, the obtaining of creation demand information, creation style preference and target audience information specifically includes:

[0025] Collect creative demand information in the form of text, pictures or videos submitted by users based on multimodal data collection tools;

[0026] Utilize natural language processing technology and sentiment analysis algorithms to analyze the user's text input information, identify the specific subject tendency and emotional color corresponding to the text input information, and determine the creative style preference;

[0027] Analyze the personal and social attributes of the target audience, form a target user portrait, and obtain target audience information.

[0028] By adopting the above technical solutions, based on collecting creative demand information in the form of text, pictures or videos submitted by users, not only the information source is enriched, but also a comprehensive understanding of user intentions is ensured, which is conducive to capturing demand information in more dimensions; through detailed analysis of the personal attributes (such as age, gender, etc.) and social attributes (such as interests, hobbies, socioeconomic status, etc.) of the target audience, an accurate target user portrait is formed, which helps to better understand the audience's needs; based on the target user portrait, the target audience information is obtained, so that the creative content is more in line with the characteristics and preferences of specific groups, enhancing the personalization and attractiveness of the work. At the same time, the application of multimodal data collection tools and natural language processing technology realizes the automated processing of creative demand information, reduces the need for manual intervention, and speeds up the acquisition and analysis of information.

[0029] In a preferred example of the present application, the analysis of the creation demand information and creation style preference, determining the theme and style of the creation content, and performing preliminary content creation in combination with the target audience information to generate a preliminary creation content framework specifically includes:

[0030] Applying a semantic understanding module to classify and weight key words in the creation demand information to establish a core concept map;

[0031] Based on the creative style preference, the pre-trained art style transfer model is called to adjust the expression of the initial creative content;

[0032] Based on the target audience portrait, select a content structure and plot development path that conforms to the predefined plot development context to generate a preliminary creative content framework.

[0033] By adopting the above technical solution, the semantic understanding module is applied to classify and weight the key words in the creative demand information, and a core concept map is established, which is conducive to ensuring the coherence and consistency of the creative content in the theme; combined with the creative style preference, the pre-trained art style transfer model is called to adjust the initial creative content expression method, which is conducive to making the work more unified in style, and can also flexibly adjust the expression method according to different creative styles. Through the carefully planned plot development path, the attractiveness and interactivity of the story are enhanced, making it easier for readers or audiences to immerse themselves in it, and improving the user's sense of participation and experience satisfaction.

[0034] In a preferred example of the present application: the method further includes:

[0035] Obtaining a preliminary creative content framework and its corresponding creative element set;

[0036] According to the creative element set and the preset content quality standard, the preliminary creative content framework is segmented, and a directional quality analysis is performed on the local creative sections with unqualified content quality to obtain corresponding quality assessment parameters;

[0037] The content optimization intensity of the local creative paragraph is adjusted according to the quality assessment parameters, and the optimization process parameters of the local creative paragraph are generated and sent to a preset content optimization module according to the adjusted optimization intensity, so that the preset content optimization module adjusts the content optimization process of the local creative paragraph according to the optimization process parameters.

[0038] By adopting the above technical solution, the preliminary creative content framework and its corresponding creative element set are obtained, and segmented processing is performed according to the preset content quality standards, ensuring that each local creative paragraph can receive a systematic quality assessment, thereby improving the accuracy of the quality assessment; targeted quality analysis is performed on local creative paragraphs with unqualified content quality, and corresponding quality assessment parameters are obtained, making optimization measures more targeted and able to accurately solve specific problems; the content optimization intensity of the local creative paragraphs is adjusted according to the quality assessment parameters, ensuring the flexibility and adaptability of the optimization process, and by adopting a dynamic adjustment mechanism, it is possible to flexibly respond to different creative needs according to actual conditions, thereby improving the optimization efficiency.

[0039] In a preferred example of the present application, the preliminary creation content framework is segmented, and a directional quality analysis is performed on the local creation sections with unqualified content quality to obtain corresponding quality assessment parameters, specifically including:

[0040] According to the creative element set, obtaining the content quality difference between adjacent creative paragraphs;

[0041] Segmenting the preliminary creation content framework according to the poor content quality to generate local creation paragraphs with similar content features;

[0042] Marking unqualified local creative paragraphs whose content quality is lower than a preset quality threshold, and performing quality analysis on the marked unqualified creative paragraphs to obtain content feature parameters of the unqualified creative paragraphs;

[0043] According to the content feature parameters, the content optimization requirements corresponding to the unqualified creative paragraphs and the directional optimization process corresponding to each creative element are analyzed to obtain the quality evaluation parameters of the unqualified creative paragraphs.

[0044] By adopting the above technical solution, the content quality difference between adjacent creative paragraphs is obtained according to the creative element set, which provides a scientific basis for segmentation processing. Through quantitative calculation, the accuracy and rationality of segmentation processing are guaranteed; then, local creative paragraphs with similar content characteristics are generated, which is convenient for concentrating resources for targeted optimization and simplifies the workload of analysis and adjustment. Then, local creative paragraphs with content quality lower than the preset quality threshold are marked as unqualified, and a detailed quality analysis is performed on them to ensure that all potential problems can be identified and handled. The content optimization requirements corresponding to the unqualified creative paragraphs and the directional optimization process corresponding to each creative element are analyzed according to the content feature parameters, and personalized optimization strategies are formulated.

[0045] In the second aspect, the invention objective of the present application is achieved by adopting the following technical solutions:

[0046] A creative content generation system based on AIGC is applied to a creative content generation method based on AIGC as described above, and the system comprises:

[0047] Demand information acquisition module, used to obtain creative demand information, creative style preferences and target audience information;

[0048] A preliminary content creation module is used to analyze the creation demand information and creation style preference, determine the theme and style of the creation content, and perform preliminary content creation in combination with the target audience information to generate a preliminary creation content framework;

[0049] An emotional creativity enhancement module, used to perform emotional enrichment and creativity enhancement processing on the preliminary creative content framework, and obtain the processed creative content framework and a corresponding creative element set; the creative element set includes multiple creative elements and innovation factor correlation coefficients between each creative element;

[0050] A model training optimization module, used to perform creation effect-oriented training optimization in a preset content creation model based on historical combination data of different creation elements in the creation element set, so as to obtain an optimized content generation model;

[0051] A strategy formulation module is used to determine the best creative effect index of each creative element at different use stages and formulate the optimal creative content generation strategy in the optimized content generation model according to the innovation factor correlation coefficient of the creative element and the corresponding creative optimization strategy;

[0052] The final content generation module is used to intelligently generate the final creative content using the optimized content generation model according to the optimal creative content generation strategy.

[0053] In the third aspect, the invention objective of the present application is achieved by adopting the following technical solutions:

[0054] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned method for generating creative content based on AIGC are implemented.

[0055] In a fourth aspect, the invention objective of the present application is achieved by adopting the following technical solutions:

[0056] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned AIGC-based creative content generation method.

[0057] In summary, the present application includes at least one of the following beneficial technical effects:

[0058] 1. Through systematic creative process design, intelligent information processing methods and efficient model optimization strategies, the quality, efficiency and personalization of creative content have been significantly improved;

[0059] 2. The quality of creative content and user experience have been significantly improved, while creative efficiency has been improved, from the aspects of comprehensive and accurate acquisition of creative demand information, precise portrayal of the target audience, the theme consistency of the creative content, optimization of expression methods, and rationality of the plot development path. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1is a flow chart of a creative content generation method based on AIGC in one embodiment of the present application;

[0061] Figure 2 This is another flowchart of a creative content generation method based on AIGC in one embodiment of the present application. DETAILED DESCRIPTION

[0062] The present application is further described in detail below in conjunction with the accompanying drawings.

[0063] In one embodiment, if Figure 1 As shown, the present application discloses a creative content generation method based on AIGC, which specifically includes the following steps:

[0064] S1: Obtain information on creative needs, creative style preferences and target audiences.

[0065] In this embodiment, step S1 includes:

[0066] S11: Collect creative demand information in the form of text, pictures or videos submitted by users based on multimodal data collection tools.

[0067] Specifically, a multimodal data acquisition tool refers to a data acquisition system that can process multiple types of input (such as text, pictures, audio, video, etc.). The multimodal data acquisition tool receives multimedia content from users through an interface (such as an API interface that supports input in multiple formats).

[0068] S12: Use natural language processing technology and sentiment analysis algorithms to parse the user's text input information, identify the specific subject tendencies and emotional colors corresponding to the text input information, and determine the creative style preferences.

[0069] Specifically, natural language processing (NLP) technology is used to parse and understand human language (such as BERT, RoBERTa), including word segmentation, part-of-speech tagging, named entity recognition, semantic analysis and other functions; sentiment analysis algorithm refers to the use of machine learning models to identify emotional components in text (such as positive, negative, and neutral emotions) and evaluate the emotional color of the text, such as sentiment vocabularies, sentiment classifiers (such as naive Bayes, support vector machines, neural networks), sentiment polarity scores, etc.; based on the identified themes and emotional tendencies, combined with historical data analysis, determine the user's creative style preferences (such as humor, seriousness, lyricism, etc.).

[0070] S13: Analyze the personal and social attributes of the target audience, form a target user portrait, and obtain target audience information.

[0071] Specifically, personal attribute characteristics include basic information such as age, gender, occupation, and geographic location; social attribute characteristics involve a wide range of social behavior characteristics such as interests, consumption habits, and social network activities; social attribute characteristics include users' interest tags, consumption records, social media interaction data, online behavior trajectories, etc.; target user portraits include users' behavior patterns, preferences, demand forecasts, etc. Through detailed target user portraits, we can better understand audience needs and ensure that the content created is more in line with the characteristics and preferences of specific groups.

[0072] S2: Analyze creative demand information and creative style preferences, determine the theme and style of the creative content, and conduct preliminary content creation based on the target audience information to generate a preliminary creative content framework.

[0073] In this embodiment, step S2 specifically includes:

[0074] S21: Apply the semantic understanding module to classify and weight the key words in the creation demand information and establish a core concept map.

[0075] Specifically, the core concept map organizes the classified keywords into a structured map according to semantic relationships to show the associations between keywords; the semantic understanding module is based on natural language processing (NLP) technology, which is used to parse the semantic relationships in the text and identify key words and their contextual meanings; keyword classification refers to dividing keywords into different categories according to their role and meaning in the text, such as subject words, descriptive words, sentiment words, etc., which can be automatically classified using predefined classification rules or machine learning models; weighted sorting refers to calculating the importance of keywords and assigning a weight value to each keyword to reflect its importance in the text. Weighted sorting can use TF-IDF algorithm, PageRank algorithm, deep learning model, etc.; in the constructed core concept map, the map nodes represent keywords and the edges represent the semantic relationships between keywords. In actual applications, visualization tools can also be used to help creators intuitively view and adjust the core concept map.

[0076] S22: Based on the creative style preference, call the pre-trained art style transfer model to adjust the content expression of the initial creation.

[0077] Specifically, creative style preference refers to the specific style that users want their works to present; the art style transfer model is a deep learning model that can convert one style into another, and is commonly used in the fields of images and text; content expression refers to the language style, sentence structure, vocabulary selection and other aspects of the creative content.

[0078] S23: Based on the target audience portrait, select the content structure and plot development path that conforms to the predefined plot development context, and generate a preliminary creative content framework.

[0079] In this embodiment, the content structure refers to the overall layout and components of the creative content, such as chapter division, paragraph arrangement, etc.; the plot development context refers to the main clues and structure of the story or content development, including the starting point, turning point, climax and ending, etc., involving plot types (such as linear, non-linear), key events, character relationships, etc.

[0080] Specifically, according to the portrait of the target audience, select the plot development template that best suits its characteristics; based on the selected plot development template, plan the overall structure of the creative content, including chapter division, paragraph arrangement, etc.; determine the specific plot development path, that is, how the story or content develops from the starting point to the end point; based on the planned structure and path, generate a preliminary creative content framework as the basis for subsequent creation; further, the application provides a dynamic adjustment mechanism that combines user feedback and data analysis to dynamically adjust the plot development context and content structure to ensure adaptation to changing audience needs, and considers different application scenarios (such as literary creation, script writing, advertising copywriting, etc.) to provide a variety of plot development templates for selection.

[0081] S3: The preliminary creative content framework is processed with emotional enrichment and creativity enhancement, and the processed creative content framework and the corresponding creative element set are obtained; the creative element set includes multiple creative elements and the innovation factor correlation coefficients between each creative element.

[0082] In this embodiment, the preliminary creation content framework is processed for emotional enrichment and creativity enhancement, specifically including:

[0083] S31: Apply the preset sentiment analysis model to analyze the sentiment components in the preliminary creative content framework, and identify positive sentiment words, negative sentiment words and neutral sentiment words.

[0084] Specifically, a sentiment analysis model is a machine learning or deep learning model used to identify sentiment components in a text; sentiment components refer to the emotional tendencies expressed in the text, such as positive sentiment words (such as "happy", "satisfied"), negative sentiment words (such as "sad", "angry"), and neutral sentiment words (such as "facts", "data"); the preset sentiment analysis model can generate a sentiment component distribution map based on the identified sentiment components in the text, intuitively display the sentiment tendency of the text, reduce the need for manual analysis of sentiment sentence by sentence, speed up the identification of sentiment components, and improve the efficiency of the creation preparation stage.

[0085] S32: Use the creative stimulation module to add or replace specific words to the preliminary creative content framework to introduce novel concepts or expressions, and evaluate the emotional coherence and rhythm between the various creative content parts to obtain preliminary creative evaluation results.

[0086] Specifically, the creative stimulation module includes a vocabulary replacement library, a concept generation algorithm, and semantic similarity calculation (applying semantic similarity algorithms (such as cosine similarity and Jaccard similarity), which is a tool based on natural language processing and generation technology to introduce novel concepts or expressions; emotional coherence refers to the consistency and fluency of emotional expression between different parts of the text, including: emotional transition smoothness, emotional consistency score, etc.; rhythm refers to the rhythm changes in the expression of the text, such as sentence length, tone strength, etc., involving sentence structure analysis, tone strength score, etc.

[0087] S33: Based on the preliminary creative evaluation results and in combination with the collected user historical behavior data, adjust the tone intensity and emotional tendency in the preliminary creative content framework, and record the creative elements used and their innovation factor correlation coefficients.

[0088] Specifically, the preliminary creative evaluation results include emotional consistency scores, rhythm scores, suggested improvement points, etc.; user historical behavior data refers to the behavioral patterns and preference information displayed by users in past creations or interactions; tone intensity refers to the strength of the tone expressed in the text, such as mild, strong, etc., and a tone intensity scoring system can be introduced; the correlation coefficient between creative elements and their innovation factors refers to the quantitative indicators of the various creative elements used in the text (such as vocabulary, sentence patterns, rhetorical techniques) and their degree of innovation; the specific value of the innovation factor correlation coefficient can be quantitatively evaluated in the form of a percentage, or it can be evaluated in the form of an interval of 0-1.

[0089] For example, the platform is promoting a new type of smart watch with functions such as health monitoring, sports tracking, and fashionable appearance. There are two main types of users on the platform - young office workers and middle-aged housewives. The goal is to generate targeted advertising copy for these two types of users to increase their willingness to buy. The preliminary creative evaluation results analysis for the target users of young office workers shows that the emotional coherence and rhythm of the existing copy are good, but the tone is a bit bland and lacks appeal. The preliminary creative evaluation results for middle-aged housewives are analyzed as follows: the existing copy overemphasizes technical parameters, lacks emotional expression, and fails to resonate.

[0090] According to the user's historical behavior data, young office workers pay more attention to the technology and convenience of the product, and like concise and powerful expressions, while middle-aged housewives tend to care more about the health function and practicality of the product, and prefer warm and caring emotional expressions. At this time, the tone intensity and emotional tendency of the advertising copy are adjusted respectively. For example, the tone adjustment content of young office workers is: enhance the sense of urgency and technology of the copy, use more impactful words such as "experience now" and "leading technology", and maintain a positive emotional color, emphasizing the convenience and efficiency of the product; the sample copy is: "Experience the latest smart watch now and master your every moment! Leading technology, easily manage health and life."; For middle-aged housewives, the tone adjustment of the advertising copy adopts a gentle and friendly tone, and the emotion tends to increase the emotional expression of warmth and care, emphasizing the practicality and thoughtful design of the product. The sample copy is: "Caring for your health starts with this smart watch. Real-time monitoring of heart rate protects you and your family every day." At the same time, the correlation coefficient of the creative elements and their innovation factors is recorded as follows:

[0091] Young office workers:

[0092] Creative elements: The use of words such as "immediate experience" and "leading technology" enhances the sense of urgency and technology.

[0093] Innovation factor correlation coefficient: The innovation factor score of these words in similar advertisements is 0.85 (high), indicating that they are highly attractive and novel.

[0094] Middle-aged housewife:

[0095] Creative elements: The use of words such as "care for your health" and "protect you and your family" adds emotional expression of warmth and care.

[0096] Innovation Factor Correlation Coefficient: These words have an Innovation Factor score of 0.78 (high) among similar ads, indicating that they are effective in triggering emotional resonance.

[0097] S4: Based on the historical combination data of different creative elements in the creative element set, creative effect-oriented training optimization is performed in the preset content creation model to obtain an optimized content generation model.

[0098] In this embodiment, step S4 specifically includes:

[0099] S41: Based on the historical combination data of different creative elements in the creative element set, calculate the usage frequency of each creative element in the historical creation and the probability of occurrence of its combination pattern.

[0100] Specifically, the creative element set refers to a collection of multiple creative elements (such as vocabulary, sentence patterns, rhetorical devices, etc.) and their innovation factor correlation coefficients; historical combination data refers to the historical records of creative elements used in the past creative process and their combination methods, including specific creative projects, used creative elements, combination methods, etc.; usage frequency refers to the number of times each creative element is used in historical creations; the probability of occurrence of a combination pattern refers to the frequency of occurrence of a specific creative element combination in historical creations.

[0101] S42: Based on the occurrence probability and creative effect feedback information, the preset content creation model is optimized using the deep reinforcement learning algorithm; by simulating the reward mechanism of different creative element combinations in the real creative environment and combining the adversarial generative network, the optimal creative path is explored.

[0102] In this embodiment, the creative effect feedback information refers to the effect evaluation information of the creative content by users or reviewers, including user ratings, comments, click-through rates, number of shares, etc.; the generative adversarial network (GAN) is a generative adversarial network used to generate realistic content and explore the optimal creative path, including a generator, a discriminator, adversarial training, etc.; deep learning algorithms include Q-learning and policy gradient algorithms.

[0103] Specifically, a variety of reward mechanisms are designed to cover different creative effect indicators (such as click-through rate, number of shares, user ratings, etc.) to comprehensively evaluate the quality of creation; based on simulation results and feedback information, the parameters of the content creation model are continuously adjusted and optimized; through multiple rounds of iterative optimization, the performance of the model is gradually improved until it reaches the optimal state; by simulating the reward mechanism of different creative element combinations in a real creative environment, the optimal creative path is explored, and the quality and attractiveness of the work are improved; combined with a variety of reward mechanisms, the model can flexibly respond to different creative needs and scenarios, enhancing the flexibility and adaptability of creation.

[0104] S43: Based on the acquired external evaluation information and the predefined quality indicator set, the model parameters of the content creation model are dynamically adjusted using the Bayesian optimization algorithm to obtain an optimized content generation model.

[0105] Specifically, external evaluation information refers to quality evaluation information about creative content from third-party reviews or user feedback, including expert review opinions, user ratings, and social media interaction data; the quality indicator set refers to a pre-defined series of standards and indicators for evaluating the quality of creative content; including logical consistency, emotional expression, innovation, communication effect, etc.; the Bayesian optimization algorithm is an optimization method based on Bayesian inference, which finds the optimal solution by constructing an agent model.

[0106] S5: In the optimized content generation model, based on the innovation factor correlation coefficient of the creative elements and the corresponding creative optimization strategy, the optimal creative effect index of each creative element at different usage stages is determined, and the optimal creative content generation strategy is formulated.

[0107] In this embodiment, the effects of each creative element at different stages of use are analyzed to determine the optimal creative effect index; and factors such as the creation cost-effectiveness ratio and the communication influence are comprehensively considered to allocate a reasonable resource investment ratio to each creative element; based on the synergy between the creative elements, the optimal combination sequence and interaction mode are planned to ensure the overall coordination and uniqueness of the final work.

[0108] S6: Based on the optimal creative content generation strategy, the optimized content generation model is used to intelligently generate the final creative content.

[0109] In this embodiment, the final creative content is intelligently generated using an optimized content generation model based on the optimal creative content generation strategy; a multi-round iterative review mechanism is implemented, with an AI-assisted manual editing team jointly reviewing the quality of the generated content and promptly correcting parts that do not meet the standards; all qualified content fragments are integrated, and after final polishing, a complete, high-quality final creative content product is output.

[0110] In one embodiment, if Figure 2 As shown, a creative content generation method based on AIGC also includes:

[0111] S10: Obtain a preliminary creative content framework and its corresponding creative element set.

[0112] In this embodiment, the preliminary creative content framework refers to the content structure formed after preliminary creation, including basic components such as paragraphs and chapters, including text framework, plot development, main time and characters, etc.; the creative element set refers to the various creative elements used in the creative process (such as vocabulary, sentence patterns, rhetoric, etc.) and their innovation factor correlation coefficients.

[0113] S20: According to the creative element set and the preset content quality standard, the preliminary creative content framework is segmented, and a targeted quality analysis is performed on the local creative sections with unqualified content quality to obtain corresponding quality assessment parameters.

[0114] In this embodiment, the preliminary creative work is divided into multiple independent paragraphs or chapters according to the preset paragraph division rules and chapter structure; the content of the targeted quality analysis involves logical coherence scoring, emotional consistency scoring, and innovation scoring (the specific scoring system and scoring values ​​can be defined by yourself); the quality assessment parameters refer to the specific indicator values ​​obtained during the evaluation process, which are used to quantify the content quality.

[0115] S30: Adjust the content optimization strength of the local creative paragraph according to the quality evaluation parameters, and generate optimization process parameters for the local creative paragraph according to the adjusted optimization strength and send them to the preset content optimization module, so that the preset content optimization module adjusts the content optimization process of the local creative paragraph according to the optimization process parameters.

[0116] In this embodiment, content optimization intensity refers to the strength taken when optimizing a specific paragraph or chapter, which can be a slight adjustment or a complete rewrite, and the intensity is evaluated based on the degree of vocabulary replacement and the amplitude of sentence structure adjustment; optimization process parameters refer to the specific operating parameters that need to be followed during the optimization process, which are used to guide the work of the optimization module, including optimization goals, optimization strategies and optimization steps; the preset content optimization module is a tool or system based on an optimization algorithm, which is specifically used to optimize creative content.

[0117] In one embodiment, in step S20, the preliminary creation content framework is segmented, and a directional quality analysis is performed on the local creation sections with unqualified content quality to obtain corresponding quality assessment parameters, specifically including:

[0118] S201: Obtain the content quality difference between adjacent creative paragraphs according to the creative element set.

[0119] In this embodiment, poor content quality refers to the difference in quality between adjacent creative paragraphs, including gaps in logical consistency, emotional expression, innovation, etc.

[0120] Specifically, the text features, sentiment features and style features of each creative paragraph are first extracted and converted into numerical representations; the feature similarity algorithm (such as cosine similarity or Jaccard similarity) is used to compare the feature vectors of adjacent creative paragraphs and calculate the difference between them; then, through a preset weight matrix, the above feature differences are integrated into an overall content quality difference.

[0121] For example, the AIGC-based creative content generation platform first generates a preliminary suspense novel framework based on Xiao Zhang's demand information, and uses a pre-trained multimodal content quality assessment model to evaluate the preliminary creative content and calculate the content quality difference between adjacent creative paragraphs; first use natural language processing (NLP) technology to extract key words, sentence length, complexity and other text features of each paragraph. Assume that the first paragraph contains the keywords "crime scene" and "evidence", and the average sentence length is 15 words; apply the sentiment analysis model to identify that the emotional tendency of the first paragraph is tense (the proportion of negative emotions is high), and the style feature extraction result is that the first paragraph uses third-person narration, less dialogue, and uses more suspense setting rhetoric. The above features are converted into numerical form, such as using the TF-IDF algorithm to convert text features into vectors, emotional features use sentiment polarity scores, and style features use specific encoding methods (such as 0 for narration and 1 for dialogue).

[0122] For each paragraph, a comprehensive feature vector is constructed, including the numerical representation of text features, sentiment features and style features. For example, the feature vector of the first paragraph may be [0.7, 0.3, 0.9] (representing the numerical representation of text features, sentiment features and style features respectively); the feature vector similarity obtained using the cosine similarity algorithm is 0.85 for the first and second paragraphs, indicating that the two have a high similarity in features (the difference score is obtained by subtracting the similarity score from 1, for example, the difference between the first and second paragraphs is 0.15 (1-0.85)); the importance weights of different features are set according to the type of creation and the target audience. For example, for suspense novels, text features and sentiment features may be more important, while style features are relatively less important.

[0123] Example: Set the weight matrix to [0.5, 0.3, 0.2] (representing the weights of text features, sentiment features, and style features, respectively), multiply the difference of each feature by its corresponding weight, and then sum them up to get the final content quality difference. Example: Assume that the text feature difference between the first and second paragraphs is 0.15, the sentiment feature difference is 0.2, and the style feature difference is 0.1, then the comprehensive content quality difference is 0.15×0.5+0.2×0.3+0.1×0.2=0.125.

[0124] S202: Segment the preliminary creation content framework according to the difference in content quality, and generate local creation segments with similar content features.

[0125] In this embodiment, segmentation processing refers to dividing the preliminary creative content framework into multiple independent paragraphs or chapters so that they can be evaluated and optimized one by one; according to the continuity of poor content quality, the preliminary creative content framework is divided into multiple local creative paragraphs; similar content features refer to paragraphs with similar features in terms of logic, emotion, innovation, etc.

[0126] S203: Mark the local creative paragraphs whose content quality is lower than a preset quality threshold as unqualified, and perform quality analysis on the marked unqualified creative paragraphs to obtain content feature parameters of the unqualified creative paragraphs.

[0127] In this embodiment, the preset quality threshold refers to a pre-set content quality standard used to determine whether a paragraph is qualified, involving a logical consistency threshold, an emotional expression threshold, and an innovation threshold; the content feature parameters refer to various parameters that describe specific problems of unqualified paragraphs, such as logical incoherence, emotional blandness, and lack of innovation.

[0128] Specifically, predefined quality assessment indicators (such as fluency, consistency, attractiveness, etc.) are first applied to score each local creation paragraph; for paragraphs with scores below the preset threshold, their specific content features (such as vocabulary selection, sentence structure, emotional expression, etc.) are recorded as content feature parameters.

[0129] S204: Analyze the content optimization requirements corresponding to the unqualified creative paragraphs and the directional optimization process corresponding to each creative element according to the content feature parameters to obtain the quality assessment parameters of the unqualified creative paragraphs.

[0130] In this embodiment, content optimization requirements refer to specific optimization requirements for unqualified paragraphs, such as enhancing logic, increasing emotional expression, introducing novel concepts, etc.; targeted optimization procedures refer to specific optimization operations for specific creative elements, such as vocabulary replacement, sentence structure adjustment, etc., and quality assessment parameters refer to the specific indicator values ​​obtained during the assessment process, which are used to quantify content quality.

[0131] It should be understood that the serial numbers of the steps in the above embodiments do not imply a sequence of execution. The execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0132] In one embodiment, a creative content generation system based on AIGC is provided, and the creative content generation system based on AIGC corresponds to the creative content generation method based on AIGC in the above embodiment.

[0133] A creative content generation system based on AIGC includes a demand information acquisition module, a preliminary content creation module, an emotional creativity enhancement module, a model training optimization module, a strategy formulation module and a final content generation module. The detailed description of each functional module is as follows:

[0134] Demand information acquisition module, used to obtain creative demand information, creative style preferences and target audience information;

[0135] The preliminary content creation module is used to analyze the creation demand information and creation style preferences, determine the theme and style of the creation content, and conduct preliminary content creation based on the target audience information to generate a preliminary creation content framework;

[0136] The emotional creativity enhancement module is used to enrich the initial creative content framework with emotions and enhance its creativity, and obtain the processed creative content framework and the corresponding creative element set; the creative element set includes multiple creative elements and the innovation factor correlation coefficients between the creative elements;

[0137] A model training optimization module is used to perform creation effect-oriented training optimization in a preset content creation model based on historical combination data of different creation elements in the creation element set, so as to obtain an optimized content generation model;

[0138] A strategy formulation module is used to determine the best creative effect index of each creative element at different use stages and formulate the optimal creative content generation strategy in the optimized content generation model according to the innovation factor correlation coefficient of the creative element and the corresponding creative optimization strategy;

[0139] The final content generation module is used to intelligently generate the final creative content based on the optimal creative content generation strategy using the optimized content generation model.

[0140] For the specific limitations of a creative content generation system based on AIGC, please refer to the limitations of a creative content generation method based on AIGC above, which will not be repeated here; each module in the above-mentioned creative content generation system based on AIGC can be implemented in whole or in part through software, hardware and a combination thereof; each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0141] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0142] S1: Obtain information on creative needs, creative style preferences and target audiences;

[0143] S2: Analyze creative demand information and creative style preferences, determine the theme and style of the creative content, and conduct preliminary content creation based on the target audience information to generate a preliminary creative content framework;

[0144] S3: enrich the initial creative content framework with emotions and enhance its creativity, and obtain the processed creative content framework and the corresponding creative element set; the creative element set includes multiple creative elements and the innovation factor correlation coefficients between the creative elements;

[0145] S4: Based on the historical combination data of different creative elements in the creative element set, creative effect-oriented training optimization is performed in the preset content creation model to obtain an optimized content generation model.

[0146] S5: In the optimized content generation model, according to the innovation factor correlation coefficient of the creative elements and the corresponding creative optimization strategy, the optimal creative effect index of each creative element at different use stages is determined, and the optimal creative content generation strategy is formulated;

[0147] S6: Based on the optimal creative content generation strategy, the optimized content generation model is used to intelligently generate the final creative content.

[0148] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0149] In one embodiment, in particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing a creative content generation method based on AIGC as described above. In such an embodiment, the computer program can be downloaded and installed from a network through a communication module, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), various functions defined in the present invention are performed.

[0150] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0151] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments may still be modified, or some of the features thereof may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A creative content generation method based on AIGC, characterized in that: include: Obtain information on creative needs, creative style preferences and target audiences; Analyze the creative demand information and creative style preferences, determine the theme and style of the creative content, and conduct preliminary content creation based on the target audience information to generate a preliminary creative content framework; Performing emotional enrichment and creativity enhancement processing on the preliminary creative content framework, and obtaining the processed creative content framework and a corresponding creative element set; the creative element set includes a plurality of creative elements and innovation factor correlation coefficients between the creative elements; Based on the historical combination data of different creative elements in the creative element set, creative effect-oriented training optimization is performed in a preset content creation model to obtain an optimized content generation model; In the optimized content generation model, according to the innovation factor correlation coefficient of the creative elements and the corresponding creative optimization strategy, the best creative effect index of each creative element at different use stages is determined, and the optimal creative content generation strategy is formulated; According to the optimal creative content generation strategy, the final creative content is intelligently generated using the optimized content generation model; The emotional enrichment and creativity enhancement processing of the preliminary creative content framework specifically includes: Applying a preset sentiment analysis model to analyze the sentiment components in the preliminary creation content framework, and identifying positive sentiment words, negative sentiment words, and neutral sentiment words; Using the creative stimulation module to add or replace specific words to the preliminary creative content framework to introduce novel concepts or expressions, and to evaluate the emotional coherence and rhythm between the various creative content parts to obtain preliminary creative evaluation results; Based on the preliminary creative evaluation results and in combination with the collected user historical behavior data, the tone intensity and emotional tendency in the preliminary creative content framework are adjusted, and the creative elements used and their innovation factor correlation coefficients are recorded.

2. According to the AIGC-based creative content generation method of claim 1, it is characterized in that: The method of performing creation effect-oriented training optimization in a preset content creation model to obtain an optimized content generation model specifically includes: Calculate the usage frequency of each creative element in historical creations and the probability of occurrence of its combination pattern based on the historical combination data of different creative elements in the creative element set; Based on the occurrence probability and creative effect feedback information, the preset content creation model is optimized using a deep reinforcement learning algorithm; the optimal creative path is explored by simulating the reward mechanism of different creative element combinations in a real creative environment and combining a generative adversarial network; Based on the acquired external evaluation information and a predefined set of quality indicators, the Bayesian optimization algorithm is used to dynamically adjust the model parameters of the content creation model to obtain an optimized content generation model.

3. The method for generating creative content based on AIGC according to claim 1, characterized in that: The obtaining of creation demand information, creation style preference and target audience information specifically includes: Collect creative demand information in the form of text, pictures or videos submitted by users based on multimodal data collection tools; Utilize natural language processing technology and sentiment analysis algorithms to analyze the user's text input information, identify the specific subject tendency and emotional color corresponding to the text input information, and determine the creative style preference; Analyze the personal and social attributes of the target audience, form a target user portrait, and obtain target audience information.

4. The method for generating creative content based on AIGC according to claim 3, characterized in that: The analysis of the creation demand information and creation style preference, determining the theme and style of the creation content, and performing preliminary content creation in combination with the target audience information to generate a preliminary creation content framework specifically includes: Applying a semantic understanding module to classify and weight key words in the creation demand information to establish a core concept map; Based on the creative style preference, the pre-trained art style transfer model is called to adjust the expression of the initial creative content; Based on the target audience portrait, select a content structure and plot development path that conforms to the predefined plot development context to generate a preliminary creative content framework.

5. The method for generating creative content based on AIGC according to claim 3, characterized in that: The method further comprises: Obtaining a preliminary creative content framework and its corresponding creative element set; According to the creative element set and the preset content quality standard, the preliminary creative content framework is segmented, and a directional quality analysis is performed on the local creative sections with unqualified content quality to obtain corresponding quality assessment parameters; The content optimization intensity of the local creative paragraph is adjusted according to the quality assessment parameters, and the optimization process parameters of the local creative paragraph are generated and sent to a preset content optimization module according to the adjusted optimization intensity, so that the preset content optimization module adjusts the content optimization process of the local creative paragraph according to the optimization process parameters.

6. The method for generating creative content based on AIGC according to claim 3, characterized in that: The preliminary creation content framework is segmented, and a directional quality analysis is performed on the local creation sections with unqualified content quality to obtain corresponding quality assessment parameters, specifically including: According to the creative element set, obtaining the content quality difference between adjacent creative paragraphs; Segmenting the preliminary creation content framework according to the poor content quality to generate local creation paragraphs with similar content features; Marking unqualified local creative paragraphs whose content quality is lower than a preset quality threshold, and performing quality analysis on the marked unqualified creative paragraphs to obtain content feature parameters of the unqualified creative paragraphs; According to the content feature parameters, the content optimization requirements corresponding to the unqualified creative paragraphs and the directional optimization process corresponding to each creative element are analyzed to obtain the quality evaluation parameters of the unqualified creative paragraphs.

7. A creative content generation system based on AIGC, characterized in that: A method for generating creative content based on AIGC as described in any one of claims 1 to 6, wherein the system comprises: Demand information acquisition module, used to obtain creative demand information, creative style preferences and target audience information; A preliminary content creation module is used to analyze the creation demand information and creation style preference, determine the theme and style of the creation content, and perform preliminary content creation in combination with the target audience information to generate a preliminary creation content framework; An emotional creativity enhancement module, used to perform emotional enrichment and creativity enhancement processing on the preliminary creative content framework, and obtain the processed creative content framework and a corresponding creative element set; the creative element set includes multiple creative elements and innovation factor correlation coefficients between each creative element; A model training optimization module, used to perform creation effect-oriented training optimization in a preset content creation model based on historical combination data of different creation elements in the creation element set, so as to obtain an optimized content generation model; A strategy formulation module is used to determine the best creative effect index of each creative element at different use stages and formulate the optimal creative content generation strategy in the optimized content generation model according to the innovation factor correlation coefficient of the creative element and the corresponding creative optimization strategy; A final content generation module, used to intelligently generate final creative content using an optimized content generation model according to the optimal creative content generation strategy; According to the optimal creative content generation strategy, the final creative content is intelligently generated using the optimized content generation model; The emotional enrichment and creativity enhancement processing of the preliminary creative content framework specifically includes: Applying a preset sentiment analysis model to analyze the sentiment components in the preliminary creation content framework, and identifying positive sentiment words, negative sentiment words, and neutral sentiment words; Using the creative stimulation module to add or replace specific words to the preliminary creative content framework to introduce novel concepts or expressions, and to evaluate the emotional coherence and rhythm between the various creative content parts to obtain preliminary creative evaluation results; Based on the preliminary creative evaluation results and in combination with the collected user historical behavior data, the tone intensity and emotional tendency in the preliminary creative content framework are adjusted, and the creative elements used and their innovation factor correlation coefficients are recorded.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for generating creative content based on AIGC as described in any one of claims 1 to 6 are implemented.

9. A computer program product, characterized in that When the computer program product runs on an intelligent system, the intelligent system executes a creative content generation method based on AIGC as described in any one of claims 1 to 6.

Images