Internet new media data feedback generation method and system

Abstract

The present application relates to the field of data feedback analysis, and more particularly to an internet new media data feedback generation method and system. The method comprises the following steps: obtaining new media video in production and internet user feedback big data; performing multi-period decomposition and individual video segment deep semantic analysis on the new media video to construct multiple content label video segments; performing social communication effect mining and video communication effect conversion rate calculation according to the internet user feedback big data to generate the social communication conversion rate of each video segment; performing positive and negative communication effect inference according to the social communication conversion rate, and performing communication conversion value evaluation on the multiple content label video segments to generate the communication conversion value of each video segment. The present application dynamically optimizes new media data based on user demand and interest feedback, and extracts the quality and communication effect of video content.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data feedback analysis, and in particular to an internet new media data feedback generation method and system. BACKGROUND

[0002] With the rapid development of the Internet and the continuous progress of information technology, new media has become an important carrier of information dissemination in modern society. Whether in social platforms, video sharing websites, or in mobile applications, the generation and dissemination of new media content have profoundly affected people's daily life, work and entertainment. Especially with the promotion of technologies such as big data, artificial intelligence and machine learning, the generation and optimization of new media content have become more intelligent and precise. The key to new media content generation and optimization is how to effectively capture the needs of the audience, analyze the behavior of the audience, and then create high-quality content that meets the interests and needs of the audience, thereby improving the dissemination effect of the content and the audience participation.

[0003] However, with the proliferation of new media content and the increasing diversity of audience needs, content creators face a huge challenge - how to stand out in the sea of information and how to ensure that their created content can be accepted and loved by the audience. Traditional new media content generation methods often rely on the experience and intuition of creators, lacking real-time analysis and feedback on audience behavior. This method not only has low efficiency, but also easily produces mismatch between content and audience needs, and cannot achieve fine-grained personalized recommendation and optimization.

[0004] Currently, the optimization method of new media content mainly relies on some conventional data analysis tools, such as user viewing time, like number, comment number and other basic indicators, but these traditional indicators often cannot deeply mine the emotional needs and personalized preferences of users. At the same time, existing content generation and optimization methods rely on manual intervention or regular analysis and inspection, which is difficult to capture and respond to the dynamic changes of audience needs in real time, and cannot make real-time prediction and optimization on the dissemination effect of content. Therefore, traditional methods are not up to the task when faced with complex user behavior, personalized needs and changing audience preferences. SUMMARY

[0005] To solve the above technical problems, the present application provides an internet new media data feedback generation method and system to solve at least one of the above technical problems.

[0006] To achieve the above purpose, the present application provides an internet new media data feedback generation method, comprising the following steps:

[0007] Step S1: Obtain new media video in production and Internet user feedback big data; perform multi-period decomposition and deep semantic analysis on each video segment of the new media video to construct a plurality of content label video segments;

[0008] Step S2: According to the Internet user feedback big data, social communication effect mining and video communication effect conversion rate calculation are performed to generate the social communication conversion rate of each video segment;

[0009] Step S3: According to the social communication conversion rate, positive and negative communication effect inference is performed, and the communication conversion value of the plurality of content label video segments is evaluated to generate the communication conversion value of each video segment;

[0010] Step S4: According to the Internet user feedback big data, user interest degree quantification analysis and emotional color analysis are performed to generate user personalized emotional feedback information of each video segment;

[0011] Step S5: Based on the communication conversion value and the emotional feedback information, deep content demand mining is performed, and the to-be-optimized element is identified to generate the to-be-optimized video element of each video segment;

[0012] Step S6: According to the to-be-optimized video element, dynamic content optimization is performed, and global coordination fitting is performed to construct a feedback optimized new media video.

[0013] The present application can ensure in-depth analysis of key content elements of each video segment by decomposing the video into multiple time periods and performing semantic analysis one by one. Emotional turning points, important dialogues, and plot climaxes of the video can be accurately identified and extracted. This process not only labels the video content with basic tags (such as theme, character, location, etc.), but also includes more detailed tags such as emotional tone and key plot. The multi-tag system provides a foundation for subsequent personalized recommendation and content optimization. Through deep semantic analysis and tagging, customized data support can be provided for each video segment, facilitating subsequent audience behavior analysis and transmission effect evaluation, so as to ensure that the video content can accurately meet the needs of different audiences. Through analysis of user feedback big data, the social transmission effect of each video segment can be quantified, including transmission breadth, sharing frequency, discussion heat, etc., which provides specific transmission effect data for video producers and helps to evaluate which content has more transmission potential. By calculating the transmission conversion rate of each video segment, it can be identified which video segments have produced high transmission effect among the audience and which have not effectively aroused resonance, which helps content creators understand audience preferences and provide references for subsequent content creation. Through the calculation of transmission effect conversion rate, producers can adjust their content release strategy or promotion strategy in real time, focusing on promoting video segments with good transmission effect, and improving the overall transmission effect of the content. Through analysis of the social transmission conversion rate, it can be inferred which video segments have produced positive social effects (such as increasing interaction and improving brand recognition), and which video segments may have produced negative effects (such as causing controversy or dissatisfaction). The distinction between positive and negative effects helps content optimization and adjustment. The transmission conversion value evaluation of each video segment can evaluate the transmission effect and potential influence of different segments, which can help content creators or advertisers accurately select which video segments are worth promoting or optimizing, thereby improving the overall transmission value. Through the analysis of transmission conversion value, data support can be provided for content creators to help them understand which content elements have higher audience appeal and transmission potential, thereby providing decision-making basis for optimizing video content. Through the quantification of user interest and emotional color analysis, the emotional needs and viewing preferences of the audience can be converted into operational data, which enables content creators to generate personalized content for different audience groups and improve the viewing experience and participation of the audience. Emotional color analysis of each video segment can identify different emotional responses of the audience during the viewing process. A certain paragraph may cause the audience to feel happy, while another paragraph may make the audience feel anxious. Through these emotional feedback, creators can adjust the emotional atmosphere of the video content to better meet the emotional needs of the audience. Through the analysis of user interest and emotion, the platform can provide personalized recommendations. If a certain audience prefers a certain emotional tone or topic, relevant content can be pushed through precise data analysis to enhance user stickiness and satisfaction.By combining the propagation conversion value and emotional feedback information, the deep needs of users can be deeply mined, such as some users may prefer a certain type of plot development or specific emotional fluctuations. Through these insights, creators can incorporate more elements that meet user needs into the content. Through the demand mining of each video segment, elements that need to be optimized (such as rhythm, emotional tone, story line, etc.) can be clearly identified, which helps to refine optimization goals and ensure that optimized content can accurately meet the emotional and interest needs of the audience. By deeply mining the content needs, each video segment can be more refined and accurate in terms of emotion and plot, thereby improving the audience's attraction and participation. According to the elements to be optimized, the content is dynamically optimized to better adapt to the needs and interests of users, and this dynamic adjustment can respond to audience feedback in real time to ensure continuous optimization of video content. While adjusting the video elements, global coordination and logic should also be considered to avoid inconsistencies or inconsistencies in content details, and this global optimization can ensure the coherence and integrity of the video content, improving the user's viewing experience. Through global optimization and feedback adjustment, a feedback-optimized video that has been optimized and adjusted multiple times is finally generated, which can meet user needs while improving the overall quality and propagation effect of the content, and this optimization method improves the market competitiveness and propagation effect of video content.

[0014] In the present specification, an internet new media data feedback generation system is provided for performing the internet new media data feedback generation method as described above, comprising:

[0015] A semantic analysis module is configured to obtain new media videos under production and internet user feedback big data, perform multi-period decomposition and deep semantic analysis of each video segment on the new media videos, and construct a plurality of content tag video segments.

[0016] A propagation conversion rate module is configured to perform social propagation effect mining and video propagation effect conversion rate calculation based on the internet user feedback big data to generate a social propagation conversion rate of each video segment.

[0017] A propagation value evaluation module is configured to perform positive and negative propagation effect inference based on the social propagation conversion rate, and evaluate the propagation conversion value of the plurality of content tag video segments to generate a propagation conversion value of each video segment.

[0018] An emotional feedback analysis module is configured to perform user interest degree quantitative analysis and emotional color analysis based on the internet user feedback big data to generate user personalized emotional feedback information of each video segment.

[0019] a content demand mining module for deep content demand mining based on the propagation conversion value and the emotional feedback information, and for identifying elements to be optimized to generate video elements to be optimized for each video segment;

[0020] a dynamic content optimization module for dynamic content optimization according to the video elements to be optimized, and for global coordination fitting to build a feedback-optimized new media video.

[0021] The present application can finely label the content of each video segment by decomposing the video into multiple video segments and performing deep semantic analysis, ensuring that subsequent analysis and optimization can be based on accurate data. Through multi-dimensional data labeling, the platform can better perform personalized recommendation and precise content pushing, improving user viewing satisfaction. The subsequent modules are provided with reliable basic data (such as the emotion, theme, and key plot of the video content), ensuring the data accuracy and consistency of the entire optimization process. By calculating the propagation effect conversion rate of each video segment, the creator can be provided with quantitative data of the video segment propagation effect, which enables the content creator to accurately assess which paragraphs have high propagation potential and which need further optimization. According to the propagation conversion rate, the creator can selectively optimize the paragraphs with poor propagation effect, thereby improving the overall content propagation effect. Data support is provided for creators and marketers, enabling them to decide on advertising investment, content promotion, and other strategies based on the propagation effect, further enhancing the social communication influence of the content. By evaluating the propagation conversion value of each video segment, the creator can determine which paragraphs are popular and have a positive effect, and which paragraphs fail to attract the audience. This provides guidance for subsequent content optimization. The high propagation value of positive propagation effect can provide feedback to the creator, indicating which plots or elements are successful, and vice versa, the negative propagation effect can prompt the creator which content may have problems and needs to be adjusted or optimized.

[0022] By disseminating the evaluation of conversion value, creators can more accurately adjust the structure of the content to maximize the dissemination value of the content and audience engagement. Through emotional color analysis and interest quantification, creators can more clearly understand the emotional response of the audience to different segments of the video, which can help creators identify which content elicits a strong response from the audience and which fails to produce the expected emotional resonance. Based on the emotional feedback of different audience groups, creators can adjust the video content to better meet the emotional needs of users. Some users may prefer more light-hearted content, while others may prefer more challenging or in-depth content. By better understanding user emotional responses, creators can produce content that is more appealing to users, thereby increasing audience engagement, interaction frequency, and content sharing rates. Through deep demand mining, creators can identify which content elements are most important to the audience and focus on optimizing these elements. By identifying elements to optimize, creators can apply improvement measures to specific video segments, ensuring that optimization directly addresses the core needs of the audience and improves the watchability and audience satisfaction of the content. Deep demand analysis helps content creators understand market trends and audience preferences, thereby improving the market adaptability and competitiveness of the content. Dynamic content tuning can respond to user feedback and data analysis in real time, ensuring that creators can make necessary optimizations before and after video release, maintaining high-quality video content at multiple stages. Through global coordination fitting, the video ensures coordination in emotional tone, rhythm, and other aspects, thereby improving the viewing experience of the audience. Precise adjustment of content not only improves the watchability of the video but also increases the emotional resonance of users, thereby promoting the word-of-mouth and social communication effects of the video. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 A step flowchart of an internet new media data feedback generation method of the present application is shown in the figure.

[0024] Figure 2 A detailed implementation step flowchart of step S1 is shown in the figure.

[0025] Figure 3 A detailed implementation step flowchart of step S2 is shown in the figure.

[0026] Figure 4 A detailed implementation step flowchart of step S3 is shown in the figure. DETAILED DESCRIPTION

[0027] It should be understood that the specific embodiments described herein are intended to be illustrative only and not limiting of the present application.

[0028] The application example provides an internet new media data feedback generation method and system. The execution subject of the internet new media data feedback generation method and system includes but is not limited to the following: mechanical equipment, data processing platform, cloud server node, network upload device, etc. which can be regarded as general computing nodes of the application. The data processing platform includes but is not limited to the following: audio image management system, information management system, cloud data management system, at least one of which.

[0029] Please refer to Figures 1 to 4 The application provides an internet new media data feedback generation method, which includes the following steps:

[0030] Step S1: obtaining new media video in production and internet user feedback big data; performing multi-period decomposition and depth semantic analysis on the new media video, and constructing multiple content label video segments;

[0031] Step S2: according to the internet user feedback big data, performing social communication effect mining and video communication effect conversion rate calculation to generate the social communication conversion rate of each video segment;

[0032] Step S3: according to the social communication conversion rate, performing positive and negative communication effect inference, and performing communication conversion value evaluation on the multiple content label video segments to generate the communication conversion value of each video segment;

[0033] Step S4: according to the internet user feedback big data, performing user interest degree quantitative analysis and emotional color analysis to generate the user personalized emotional feedback information of each video segment;

[0034] Step S5: based on the communication conversion value and the emotional feedback information, performing deep content demand mining and identifying the to-be-optimized elements to generate the to-be-optimized video elements of each video segment;

[0035] Step S6: according to the to-be-optimized video elements, performing dynamic content optimization and global coordination fitting to construct the feedback optimized new media video.

[0036] The present application can ensure in-depth analysis of key content elements of each video segment by decomposing the video into multiple time periods and performing semantic analysis one by one. Emotional turning points, important dialogues, and plot climaxes of the video can be accurately identified and extracted. This process not only labels the video content with basic tags (such as theme, character, location, etc.), but also includes more detailed tags such as emotional tone and key plot. The multi-tag system provides a foundation for subsequent personalized recommendation and content optimization. Through deep semantic analysis and tagging, customized data support can be provided for each video segment, facilitating subsequent audience behavior analysis and transmission effect evaluation, so as to ensure that the video content can accurately meet the needs of different audiences. Through analysis of user feedback big data, the social transmission effect of each video segment can be quantified, including transmission breadth, sharing frequency, discussion heat, etc., which provides specific transmission effect data for video producers and helps to evaluate which content has more transmission potential. By calculating the transmission conversion rate of each video segment, it can be identified which video segments have produced high transmission effect among the audience and which have not effectively aroused resonance, which helps content creators understand audience preferences and provide references for subsequent content creation. Through the calculation of transmission effect conversion rate, producers can adjust their content release strategy or promotion strategy in real time, focusing on promoting video segments with good transmission effect, and improving the overall transmission effect of the content. Through analysis of the social transmission conversion rate, it can be inferred which video segments have produced positive social effects (such as increasing interaction and improving brand recognition), and which video segments may have produced negative effects (such as causing controversy or dissatisfaction). The distinction between positive and negative effects helps content optimization and adjustment. Transmission conversion value evaluation of each video segment can evaluate the transmission effect and potential influence of different segments, which can help content creators or advertisers accurately select which video segments are worth promoting or optimizing, thereby improving the overall transmission value. Through the analysis of transmission conversion value, data support can be provided for content creators to help them understand which content elements have higher audience appeal and transmission potential, thereby providing decision-making basis for optimizing video content. Through the quantification of user interest and emotional color analysis, the emotional needs and viewing preferences of the audience can be converted into operational data, which enables content creators to generate personalized content for different audience groups and improve the viewing experience and participation of the audience. Emotional color analysis of each video segment can identify different emotional responses of the audience during the viewing process. A certain paragraph may cause the audience to feel happy, while another paragraph may make the audience feel anxious. Through these emotional feedback, creators can adjust the emotional atmosphere of the video content to better meet the emotional needs of the audience. Through the analysis of user interest and emotion, the platform can make personalized recommendations. If a certain audience prefers a certain emotional tone or topic, relevant content can be pushed through precise data analysis to enhance user stickiness and satisfaction.By combining the propagation transformation value and emotional feedback information, the deep needs of users can be deeply excavated, for example, some users may prefer a certain type of plot development or specific emotional fluctuations. Through these insights, creators can incorporate more elements that meet user needs into the content. Through the demand excavation of each video segment, elements that need to be optimized (such as rhythm, emotional tone, story line, etc.) can be clearly identified, which helps to refine the optimization target and ensure that the optimized content can accurately meet the emotional and interest needs of the audience. By deeply excavating the content demand, each video segment can be more refined and accurate in terms of emotion and plot, thereby improving the audience's attraction and participation. According to the elements to be optimized, the content is dynamically optimized to better adapt to the needs and interests of users, and this dynamic adjustment can respond to audience feedback in real time to ensure continuous optimization of video content. While adjusting the video elements, global coordination and logic should also be considered to avoid disharmony or incoherence in the content details, and this global optimization can ensure the coherence and integrity of the video content and improve the user's viewing experience. Through global optimization and feedback adjustment, a feedback-optimized video that has been optimized and adjusted multiple times is finally generated, which can meet user needs while improving the overall quality and propagation effect of the content, and this optimization method improves the market competitiveness and propagation effect of the video content.

[0037] In the embodiments of the present application, referring to Figure 1 The present application is a method for generating internet new media data feedback, and the steps of the method are as follows:

[0038] Step S1: Obtain new media video in production and internet user feedback big data; perform multi-period decomposition and depth semantic analysis of each video segment on the new media video to construct multiple content label video segments;

[0039] In this example, after obtaining authorization from the user and related platforms, the new media video sources that need to be analyzed are clearly defined, which can be self-owned platforms, social media, video sharing websites, etc. The theme and target audience of the video are determined to facilitate subsequent analysis. If the goal is to analyze introduction videos about technology products, relevant videos can be selected from platforms such as YouTube and Bilibili. Use appropriate tools to collect videos, ensuring that the quality and completeness of the collected videos are high. Collection can be done through API interface, crawler technology or direct download tools. Assuming that 10 videos are selected, each with a duration of 5 to 10 minutes, ensuring that the collected content is representative and covers different time periods and themes. Organize the obtained videos, ensuring that the basic information of each video (such as title, duration, uploader, etc.) is recorded. In addition, video files need to be stored uniformly in designated folders for subsequent processing. An Excel table can be created to record the title, duration and download link of each video for easy retrieval later. Determine which channels to collect user feedback data from. Feedback data can come from comment sections, social media sharing and discussions, like and dislike data, etc. These feedbacks help understand the audience's views on video content. Analyze YouTube video comments or obtain relevant user discussions through social media platform APIs. Use data collection tools (such as crawlers or API interfaces) to extract user feedback data. Ensure that each user feedback contains relevant information such as user ID, feedback content, timestamp, likes and replies, etc. Assuming that 200 user comments are extracted from each video, record these data and store them in a database or Excel table, ensuring their structure for subsequent analysis. Clean the collected user feedback data, removing invalid comments (such as advertisements, spam or comments without content), ensuring the quality and reliability of the remaining data. If 400 invalid comments are removed from 2000 comments, the remaining 1600 valid comments will be used for subsequent analysis. Divide each video into multiple time periods and determine appropriate segmentation points. Segmentation can be based on natural changes in video content (such as scene changes, topic shifts or emotional fluctuations). For an 8-minute video, it can be divided into 4 periods, each 2 minutes long, ensuring that each segment can independently convey information. Use video analysis tools (such as Adobe Premiere Pro or FFmpeg) to segment the video. Position the timeline to ensure the start and end times of each segment are accurate. Assuming the segmentation of a certain video segment is as follows: segment 1 (0:00-2:00), segment 2 (2:00-4:00), segment 3 (4:00-6:00), segment 4 (6:00-8:00), ensuring that these segments cover different content highlights. Record the information of each segment, including the duration, theme and content summary of each segment, which will provide a basis for subsequent deep semantic analysis.The theme of section 1 is "product introduction", section 2 is "user feedback", section 3 is "market comparison", and section 4 is "summary and prospect". Selecting appropriate deep semantic analysis tools and technologies, natural language processing (NLP) models such as BERT or GPT series can be used to analyze the text content and extract the key themes, emotions and related semantic features in the video segments. Through NLP model analysis of user comments, the main emotional tendency and keywords are extracted. For each video segment, deep semantic analysis is performed to extract content-related theme words, emotional color and potential user needs, which can be achieved through keyword extraction, sentiment analysis and other methods. Analyze the text content of section 1 and identify keywords such as "innovation", "user experience" and "market demand", with a positive emotional tendency score of 8 / 10. Organize the deep semantic analysis results of each video segment into structured data for subsequent analysis, and this dataset will contain the theme, emotional score and key semantic features of each paragraph. The semantic features of section 1 are "theme: innovation, emotional score: 8 / 10, keywords: user experience, market demand".

[0040] Step S2: According to the Internet user feedback big data, the social communication effect is mined and the video communication effect conversion rate is calculated to generate the social communication conversion rate of each video segment;

[0041] In this embodiment, the data collected from user feedback, such as comments, likes, shares and forwards, are integrated. Ensure that the relevant social communication data of each video segment can match the corresponding user feedback information, so as to provide a comprehensive perspective for subsequent analysis. Assuming that feedback from 1000 users is collected for each video segment, including 500 comments, 300 likes and 200 shares. Record these data in a structured database for subsequent processing. Define and calculate the relevant indicators of social communication effect, including the number of comments, likes and shares. Through these indicators, the communication effect of each video segment on social platforms can be evaluated. If the number of comments for segment 1 is 150, the number of likes is 200, and the number of shares is 100, the social communication effect indicators of this segment can be calculated to evaluate its popularity among users. Use social network analysis (SNA) method to evaluate the pattern and communication path of user interaction. By analyzing the interaction between users, identify the key nodes and influential users of content communication. Use network visualization tools (such as Gephi or Cytoscape) to display the relationship network of user comments and shares, so as to identify the users and content with the most significant communication effect. Organize the results of social communication effect analysis into a report to ensure that the communication effect indicators of each video segment are recorded and interpreted, which will provide an important basis for subsequent communication conversion rate calculation. Record the communication effect of segment 1 as "comments: 150, likes: 200, shares: 100", and analyze the positivity of audience response. Determine the calculation method of video communication effect conversion rate. Generally, conversion rate can be defined as the ratio of effective viewing times generated by social sharing and user interaction to total viewing times. If the total viewing time of a certain video segment is 2000 times, and the viewing time caused by sharing and interaction is 500 times, the conversion rate formula is: conversion rate = (induced viewing time / total viewing time) x 100%. According to the integrated social communication data and viewing times, calculate the conversion rate of each video segment one by one. Ensure that accurate statistical data is used in calculation and possible errors are considered. If the total viewing time of segment 1 is 2000, the viewing time caused by sharing is 300, and the viewing time caused by likes is 200, the conversion rate of segment 1 is: conversion rate = (500 / 2000) x 100% = 25%. Record the communication effect conversion rate of each video segment calculated in the database and visualize it. You can use bar charts or pie charts and other methods to intuitively understand the conversion effect of each segment. Through data visualization tools, draw a chart of the conversion rate of each segment to show the differences in communication effect of different video segments, so as to facilitate subsequent optimization decisions.

[0042] Step S3: According to the social communication conversion rate, the positive and negative communication effect is inferred, and the communication conversion value of multiple content label video segments is evaluated to generate the communication conversion value of each video segment.

[0043] In this example, the evaluation criteria for determining positive and negative propagation effects are established. Positive propagation effects are typically associated with positive user feedback such as likes, positive comments, and shares, while negative propagation effects are associated with negative user feedback such as dislikes, negative comments. By analyzing user feedback data, a comprehensive evaluation model is established. If the number of likes for a video segment is significantly higher than the number of dislikes, and the comments contain a large number of positive sentiment words, it can be determined that the segment has a positive propagation effect. Perform sentiment analysis on user comments for each video segment using natural language processing techniques to identify sentiment trends. Sentiment analysis tools such as VADER or TextBlob can be used to score comments and assess their emotional color. For example, if 80% of the comments for a video segment are positive, 10% are neutral, and 10% are negative, it can be inferred that the segment has a significant positive propagation effect. Combine the results of social propagation conversion rate and sentiment analysis to comprehensively evaluate each video segment. Use a weighted model to combine conversion rate and positive and negative propagation effects to obtain a comprehensive effect score. If the conversion rate of segment 1 is 25% and the positive sentiment proportion is 80%, the comprehensive propagation effect score can be calculated as follows: Comprehensive score = conversion rate × positive sentiment proportion = 25% × 0.8 = 20%. Organize the results of the positive and negative propagation effect inference for each video segment into a report to ensure a clear description of the propagation effect of different video segments. Record the positive and negative effects and their corresponding comprehensive scores for each segment. Define the calculation method of propagation conversion value, which can usually combine positive and negative propagation effects and social propagation conversion rate to form a comprehensive propagation conversion value evaluation model. The propagation conversion value can be defined as: propagation conversion value = (positive propagation effect - negative propagation effect) × social propagation conversion rate. Based on the positive and negative propagation effects inferred in the previous steps and the calculated social propagation conversion rate, calculate the propagation conversion value of each video segment one by one. Ensure that accurate indicators and data are used during calculation. Assuming that the positive effect score of segment 1 is 15%, the negative effect score is 5%, and the social conversion rate is 25%, the propagation conversion value = (15% - 5%) × 25% = 2.5%. Record the calculated propagation conversion value of each video segment in the database and visualize it. Bar charts or pie charts can be used to facilitate intuitive understanding of the propagation conversion value of different video segments. Use data visualization tools to plot the propagation conversion value of each segment into a chart to show the differences in propagation effect of different video segments to facilitate subsequent optimization decisions. Analyze the propagation conversion value of each video segment to identify segments with good and poor performance and explore possible reasons. This analysis will provide guidance for content optimization and help creators understand which content is more likely to trigger user interaction and sharing.If the propagation conversion value of section 1 is 2.5, and the conversion value of section 3 is only 0.5, the content and propagation strategy of section 1 can be analyzed to extract the successful factors, and the corresponding adjustments can be made in section 3.

[0044] Step S4: According to the big data of Internet user feedback, the user interest degree quantification analysis and emotional color analysis are carried out to generate user personalized emotional feedback information of each video section;

[0045] In this embodiment, the relevant data collected from user feedback is integrated, including comment content, like number, share number and viewing time, etc. These data will provide the basis for quantifying user interest degree. Assuming that the number of user comments of a certain video section is 300, the like number is 150, the share number is 50, and the total viewing time is 2000 seconds, these data will help us understand the user's interest degree of the content. Determine the indicators for quantifying user interest degree. The number of comments, likes, shares and viewing time can be combined to form a comprehensive interest score model. Each indicator can be given different weights according to its importance. Assuming that we define the weights as follows: comment number accounts for 30%, like number accounts for 40%, share number accounts for 20%, and viewing time accounts for 10%. According to these weights, the contribution of each indicator can be calculated. According to the defined weights, the user interest score of each video section is calculated one by one. After multiplying each indicator by its weight and summing up, the comprehensive interest score is obtained. If the number of comments of a certain video section is 300, the like number is 150, the share number is 50, and the viewing time is 2000 seconds, the interest score can be calculated as:

[0046] Comment contribution = 300 x 0.3 = 90

[0047] Like contribution = 150 x 0.4 = 60

[0048] Share contribution = 50 x 0.2 = 10

[0049] Viewing time contribution = (2000 / 2000) x 10 = 10

[0050] Comprehensive interest score = 90 + 60 + 10 + 10 = 170.

[0051] Selecting the appropriate sentiment analysis tool and technology, usually using natural language processing (NLP) models to analyze the sentiment of user comments. Commonly used tools include VADER, TextBlob, etc., which can effectively identify the emotional color in the comments. If a video comment uses positive words such as "love", "like", "great", etc., sentiment analysis tools will be able to identify these emotional tendencies. Each user comment is scored for sentiment, analyzing the proportion of positive, negative, and neutral sentiment in the comments. The number of comments in each sentiment category can be recorded, and the distribution of sentiment color can be calculated. Assuming that a video with 300 comments has 60% positive sentiment, 10% negative sentiment, and 30% neutral sentiment, the sentiment color feature of this segment can be derived as "positive: 60%, negative: 10%, neutral: 30%". According to the results of sentiment analysis, generate sentiment color features for each video segment, which can be done by calculating the proportion of different sentiment types for subsequent analysis and personalized feedback. Record the sentiment features of segment 1 as "positive sentiment 60%, negative sentiment 10%, neutral sentiment 30%", which will support the subsequent personalized emotional feedback information. Organize the sentiment color analysis results of each video segment into a report to ensure that the content is clear and easy to understand, which will provide the basis for subsequent user personalized emotional feedback information generation. Record the sentiment features of segment 1 and store them together with the user interest score to provide decision-making basis for content optimization. Combine the user interest score and sentiment color features to generate user personalized emotional feedback information for each video segment, which will provide specific user preferences and emotional needs for content creators. If the interest score of segment 1 is 170 and the sentiment feature is "positive sentiment 60%", the feedback information can be generated as "the user of segment 1 is interested in the content, the positive sentiment proportion is high, and the current content style should be maintained". Record the generated user personalized emotional feedback information in the database to ensure the structure and traceability of the information, which will provide the basis for subsequent content optimization and strategy adjustment. Record the personalized feedback information of segment 1 for reference in subsequent content production to ensure that creators can adjust according to user needs. Finally, integrate the user personalized emotional feedback information into a report to ensure that content creators can intuitively understand user needs, which will promote content optimization and improvement, and improve user satisfaction and participation.

[0052] Step S5: Based on the propagation conversion value and the emotional feedback information, deep content demand mining is carried out, and the to-be-optimized elements are identified to generate the to-be-optimized video elements of each video segment;

[0053] In this embodiment, the propagation conversion value is integrated with the emotional feedback information. Ensure that the propagation conversion value and user emotional feedback information of each video segment can be accurately corresponded, in order to carry on the in-depth analysis, this data integration will provide solid foundation for the identification of the to-be-optimized elements. Suppose the propagation conversion value of segment 1 is 2.5, and the emotional feedback information shows that the proportion of positive emotions of users to the content is 60%. Record these information in a structured data table for subsequent analysis. Use the propagation conversion value and emotional feedback information to analyze the content demand characteristics of each video segment. Through statistical methods such as cluster analysis or factor analysis, the user's preference and demand for different content types can be identified. Users may show higher positive emotions and participation in some paragraphs, indicating that these content types are more popular. Suppose it is found in the analysis that the positive emotional score of users to the content type of "product use case" is generally high, and its propagation conversion value is also outstanding, then it can be determined that this content type is a high demand characteristic. Map the identified deep content demand characteristics to ensure that the demand characteristics of each video segment can be clearly recorded. This process will help content creators understand the real needs of users and provide guidance for subsequent optimization. Record the demand characteristics of segment 1 as "high demand for product use cases, suggest adding relevant examples in subsequent content". Determine the identification criteria of the to-be-optimized elements, which can include the problems mentioned in user feedback, the negative emotion ratio in emotional feedback, and the content segments with low propagation conversion value. Clearly defining these criteria will help more systematically identify to-be-optimized elements. If the negative emotion ratio of a certain video segment exceeds 20% and the propagation conversion value is less than 1, it can be determined that this paragraph is a to-be-optimized element. According to the defined criteria, identify the to-be-optimized elements of each video segment. Through text analysis of user feedback, specific content problems such as "information is not clear enough" and "too slow pace" can be identified. Suppose there are multiple "information is not clear enough" feedback in the user comments of segment 1, then it can be marked as a to-be-optimized element. Organize the identified to-be-optimized elements to ensure that the to-be-optimized parts of each video segment can be clearly recorded. These elements may include the clarity of the voiceover, the rhythm of the picture, the logical coherence of the content, etc. Record the to-be-optimized elements of segment 1 as "information is not clear enough, suggest adding examples to explain".

[0054] Step S6: dynamically optimize the content according to the to-be-optimized video elements, and globally coordinate and fit to build a feedback-optimized new media video.

[0055] In this embodiment, the identified video elements for optimization in each video segment are analyzed in detail. These elements may include unclear information transmission, improper rhythm, inconsistent emotional expression, etc. The analysis should consider audience feedback and emotional color to ensure that the optimization decision can effectively solve the problems raised by users. If the user feedback for a certain video segment generally mentions "unclear information transmission", the voiceover content, illustrations and information display method of that paragraph should be focused on. According to the specific situation of the elements to be optimized, the corresponding optimization strategy is formulated. For the problem of unclear information transmission, you can consider adding examples, simplifying language or using charts to assist in explanation; for the problem of improper rhythm, you can adjust the rhythm of the editing points or background music. Suppose the voiceover content of segment 1 is too complex, the creator can decide to simplify it, use more direct language and add relevant examples to help the audience better understand the content. Implement the optimization strategy in the video editing software. According to the formulated scheme, the video segment is edited and adjusted accordingly. Through frame-by-frame analysis, it is ensured that each adjustment can achieve the expected effect and consider the viewing experience of the audience. For segment 1, after modifying the voiceover text, play it back several times to ensure that the information transmission is clearer, and adjust it immediately according to the feedback. If you find that the length of a certain example explanation is too long, you can shorten it appropriately to maintain the audience's attention. After completing the optimization of each video segment, evaluate the overall content to ensure the logical coherence and emotional consistency between paragraphs. Through overall playback, observe the smoothness of the video and the effectiveness of information transmission. Play the entire video and observe whether the transitions between different paragraphs are natural and whether the emotional expression is consistent. If you find that the emotional expression of a certain paragraph does not match the overall style, further adjustments are needed. According to the results of global evaluation, formulate a coordinated adjustment strategy. It may be necessary to fine-tune the emotional tone, rhythm and voiceover style of certain paragraphs to ensure the unity and appeal of the overall content. Suppose the emotional expression of segment 3 is too flat, while the previous and subsequent paragraphs are full of vitality, you can consider adding more emotionally charged language or background music to segment 3 to enhance its emotional performance. During the coordination adjustment process, collect the feedback of the audience in time and make necessary corrections. You can understand their feelings about the optimized content through small-scale audience testing and make further optimizations based on feedback. Invite a small group of target audience to watch the optimized video and collect their opinions. If the audience generally reflects that the emotional expression of the paragraph has improved, you can continue to maintain this adjustment strategy; if there is still feedback that points out problems, you need to make further corrections. After completing all the optimization and coordination, compile the final version of the optimized new media video to ensure that all modifications have been implemented and the quality of the video has reached the expected target. You can use video editing software to make detailed adjustments to the final version, such as editing, sound effects and subtitle optimization. Ensure that the volume of each paragraph is consistent and the display time of the subtitles is synchronized with the voiceover to improve the viewing experience of the audience.

[0056] In this embodiment,Figure 2 For the detailed implementation step flowchart of step S1, in this embodiment, the detailed implementation step of step S1 includes:

[0057] Obtaining new media videos in production and Internet user feedback big data;

[0058] Performing multi-period decomposition and deep semantic analysis on each video segment of the new media videos to generate deep semantic features of each video segment;

[0059] Performing content label classification according to the deep semantic features to generate accurate content labels of each video segment;

[0060] Performing time sequence positioning mark fitting on the new media videos according to the accurate content labels to construct a plurality of content label video segments.

[0061] In this embodiment, after obtaining authorization from the user and the relevant platform, the storage location and access channel of the new media video are determined, including social media platforms, video sharing websites, etc. User feedback data such as access volume, like count, and comment count of the relevant video are obtained for comprehensive analysis. If a certain video is obtained from YouTube, the record format is "video ID, title, publication date, access volume, like count, comment count", such as "video_001, 'New Media Content Analysis', 2023-01-01, 10000, 500, 200". Relevant user comments and feedback are extracted to analyze user sentiment and opinions, which will be used as the basis for subsequent content optimization. The format of user comments is "user ID, comment content, timestamp", such as "user_001, 'Content is interesting!', 2023-01-0208:00". The collected data is cleaned to remove invalid or duplicate information, ensuring data accuracy and integrity. The processed data is stored in the database for subsequent analysis. If a comment is found to be an invalid link, it is marked and deleted, recorded as "data cleaning completed, deleted invalid comments". The new media video is divided into segments according to time periods, usually every 30 seconds or every minute, to facilitate subsequent analysis. Each video segment should be labeled with its start and end time. If the total length of the video is 6 minutes, it is divided into "segment 1: 0:00-0:30, segment 2: 0:30-1:00", and so on. Deep semantic analysis is performed on each video segment using natural language processing (NLP) technology to analyze the dialogues, voiceovers, and subtitles in the video. The core themes, emotions, and keywords in the video segments are extracted. The dialogue content of segment 1 is processed using sentiment analysis tools, recorded as "segment 1 deep semantic features: theme 'user experience', sentiment positive". The deep semantic features of each video segment are recorded in the database for subsequent label classification and analysis. The record format is "segment number, theme, sentiment direction, keyword", such as "segment 1, user experience, positive, 'fun, interaction'". According to the deep semantic features, a content label classification model is constructed. Machine learning algorithms such as support vector machines or random forests can be used to classify video segments to ensure that the model can accurately identify different types of content. Set the label categories to "education, entertainment, technology, emotion" and train according to the features. Use the constructed label classification model to classify each video segment and output the accurate content label of each segment. Ensure the accuracy and effectiveness of the classification results. If segment 1 is classified as "education", record it as "segment 1 accurate content label: education". Verify the classification results using cross-validation and other methods to ensure the accuracy of the model. If there are inaccurate labels, adjust the model parameters. After verification, it is found that the accuracy of the education label is 90%, recorded as "label classification accuracy verification completed". According to the accurate content label of each video segment, time sequence positioning is marked to ensure that each label matches the corresponding time period.Create a precise timing label framework. The record format is "segment number, start time, end time, label", such as "segment 1, 0:00, 0:30, education". Integrate each video segment with labels to generate multiple content-labeled video segments that can be used for subsequent content recommendation or post-processing. The generated records are "segment 1: education, segment 2: entertainment, segment 3: technology", ensuring that each segment has a clear theme. Record all content-labeled video segments and timing labels in the database and generate reports for subsequent analysis and optimization of content strategy. The record format is "video ID, segment information", such as "video_001, segment 1: education, segment 2: entertainment".

[0062] In this embodiment, the specific steps for multi-period decomposition and deep semantic analysis of each video segment of the new media video to generate deep semantic features of each video segment are as follows:

[0063] Identify the full video content rhythm and scene transition turning points of the new media video;

[0064] Based on the content rhythm and scene transition turning points, calculate the optimal segmentation time points and extract multiple video segmentation time points;

[0065] Based on the multiple video segmentation time points, perform multi-period decomposition to obtain multiple time window video segments;

[0066] Extract key frames from multiple time window video segments to obtain key frames of each video segment;

[0067] Extract key semantic fields and key content elements from the key frames;

[0068] Perform latent association analysis on the key semantic fields and key content elements to obtain latent semantic subject feature vectors;

[0069] Based on the latent semantic subject feature vectors, perform deep semantic analysis of each video segment to generate deep semantic features of each video segment.

[0070] In this embodiment, the first step is to analyze the audio and visual content in the video to identify its rhythm. Using audio signal processing techniques, the audio signal in the video is extracted, focusing on the beat of the music and the pace of the dialogue. By analyzing the audio waveform and spectrum, the rhythm changes of the audio can be determined and synchronized with the changes in the video frames. Using tools such as audio analysis software, the audio is segmented, and different beats are marked, such as where the rhythm is faster, there may be more frequent scene changes. In this way, the speed of the video rhythm can be identified, and the basis for subsequent scene transformation analysis is provided. For scene transformation in the video, visual content analysis techniques are used to detect changes in the picture. By comparing the color histogram between adjacent frames or using edge detection algorithms, significant change points in the picture can be identified. When the picture tone, brightness, or content changes significantly, it is considered a turning point for scene switching. By analyzing each frame of the video, these turning points are marked, and the reasons for their occurrence are understood, such as whether due to plot development, emotional change, or theme shift, etc. After completing the identification of rhythm and turning points, these information is organized into a dataset for subsequent analysis, which will contain the timestamp, type, and possible impact of each rhythm point and scene transformation, ensuring that subsequent analysis can accurately reflect the overall structure of the video. Based on the identified content rhythm and scene transformation points, a reasonable segmentation strategy is designed. The goal is to maintain the coherence of each video segment in terms of content while avoiding cutting at important information or emotional high points. By analyzing the content density and emotional changes of different paragraphs, the best segmentation location is determined. By analyzing the distribution of rhythm points and scene turning points, multiple suitable time segmentation points are calculated. Considering the length of each paragraph and the importance of its content, it is ensured that these time points not only reflect natural content changes but also conform to the viewing habits of the audience. If the rhythm speeds up and the scene changes frequently at a certain time, a segmentation may be set at that place. According to the previously calculated segmentation time points, the video content is decomposed. Each segment should maintain consistency in terms of visual and narrative, ensuring that the audience can understand the theme and emotion of the paragraph when watching. If a paragraph discusses product features, the paragraph should contain relevant visual information and commentary. Key frames are extracted from each decomposed paragraph, usually using uniform sampling or content change-based selection methods to ensure that representative frames of each paragraph are selected. These key frames should be able to reflect the core information and emotions of the paragraph. Every few frames in a paragraph are extracted to form a visual summary of the paragraph. In-depth analysis is performed on each key frame, mainly using image recognition technology and optical character recognition (OCR) technology to extract content. For key frames containing text information, OCR technology is used to obtain the text content, and objects in the image are identified to extract key semantic fields and important content elements. In the key frames, information such as "product demonstration" or "user feedback" is identified, which will serve as the basis for subsequent analysis.Perform in-depth correlation analysis on the extracted key semantic fields and content elements to uncover potential relationships between them. Use clustering analysis or association rule learning methods to identify which elements frequently appear together and which fields have semantic connections. Discover that there is a strong association between "user feedback" and "product features," indicating that these two often appear in the same context. Based on the results of the latent association analysis, generate feature vectors for each video segment that will reflect the thematic and sentiment characteristics of the passage, facilitating subsequent deep semantic parsing. Map each key semantic field to a vector and weight it according to its importance in the passage to form a comprehensive feature representation. Use the generated latent semantic subject feature vectors to perform deep semantic parsing on each video segment. By analyzing the key content and sentiment of each passage, extract rich semantic information, including themes, sentiment orientation, and important concepts. Analyze the sentiment of the passage content using sentiment analysis techniques to determine its sentiment orientation as positive, negative, or neutral, and combine it with the theme for a comprehensive description. Integrate the parsed results to generate deep semantic features for each video segment, which will include information such as themes, sentiment, and keywords, forming a comprehensive understanding of the passage. Report that the theme of a certain passage is "product introduction," the sentiment orientation is "positive," and extract relevant keywords. Organize the deep semantic features of each video segment into a document to ensure that the analysis results can be effectively utilized, providing important references for subsequent content optimization and evaluation.

[0071] In this embodiment, referring to Figure 3 For the detailed implementation step flowchart of step S2, in this embodiment, the detailed implementation steps of step S2 include:

[0072] Calculate the user viewing time, video click volume, and play volume of each video segment of the internet user feedback big data;

[0073] Perform user behavior analysis on the user viewing time, video click volume, and play volume to obtain user viewing behavior characteristics;

[0074] Calculate the sharing volume and forwarding rate of each video segment of the internet user feedback big data;

[0075] Perform user sharing behavior analysis on the sharing volume and forwarding rate to generate user sharing behavior characteristics;

[0076] Based on the user sharing behavior characteristics and user viewing behavior characteristics, perform social communication effect mining to generate social communication effect characteristics;

[0077] Based on the social communication effect characteristics, calculate the video communication effect conversion rate of each content label video segment to generate the social communication conversion rate of each video segment.

[0078] In this embodiment, the user viewing time, click volume and play volume of each video segment are obtained from the Internet user feedback big data. These data can usually be obtained through the analysis tool of the video platform, and record the user's behavior when watching the video, including the start and end time stamp of the video. If the play volume of a certain video segment is 2000 times, and the total viewing time of the user is 3000 seconds, the detailed viewing time and play of each paragraph can be recorded. For each video segment, the average viewing time of the user is calculated, which can be achieved by dividing the total viewing time of the user by the number of users watching the segment. This data can help us understand the audience's investment in the video segment. For example, if the viewing number of a certain video segment is 1000 and the total viewing time is 3000 seconds, the average viewing time is 3 seconds. This indicator can reflect the attractiveness of the content of the segment. Click volume refers to the number of times the user clicks to watch the video, while play volume is the actual number of times the video is played. By comparing these two data, the attractiveness and viewing conversion rate of the video can be evaluated. If the click volume is 2500 times and the play volume is 2000 times, the viewing conversion rate of the video can be calculated as 80%. According to the calculated user viewing time, click volume and play volume, the user's viewing behavior is extracted. Analyze the user's viewing habits in different video segments and identify the paragraphs with longer viewing time or higher viewing rate. If it is found that the average viewing time of a certain video segment is 5 seconds, while the viewing time of other segments is only 2 seconds, it can be considered that the content of this segment is more popular.

[0079] Using statistical methods such as clustering analysis, the user's viewing behavior is pattern-recognized. Users can be divided into high-interactive users and low-interactive users, and their viewing habits in different video segments are analyzed. By analyzing the user's viewing frequency and interaction level (such as comments and likes), the user's preferences and habits can be further understood. From social media platforms or video sharing websites, the sharing volume and forwarding rate of each video segment are obtained. These data usually include the number of times a user shares a video on social networks and the number of times a video is watched through a shared link. If a video segment has a sharing volume of 500 and a playback volume of 1000 through sharing, the sharing forwarding rate can be calculated. The method of calculating the forwarding rate is to divide the sharing volume by the playback volume, then multiply by 100 to get the percentage. The forwarding rate is an important indicator of the spread of content on social media. For example, if the sharing volume is 500 and the playback volume is 2000, the forwarding rate is 25%, which reflects the social spread of the video. According to the data of sharing volume and forwarding rate, the sharing behavior of users is characterized. Identify which video segments have a high sharing rate and analyze the reasons, which may be related to the interestingness of the content, the amount of information or emotional resonance. If a video segment has a sharing rate higher than 30%, it may be because the content has strong emotional resonance or social topics. According to the sharing behavior characteristics, user portraits are constructed, which can be achieved by analyzing the user's social media behavior, sharing frequency and preferred content types. If a user often shares technology videos, it can be considered that he has a high interest in technology content. The analysis results are organized into a report to show the characteristics and patterns of user sharing behavior, providing feasible optimization suggestions for content creators. Such a report will help content creators understand which content is easy to share by users, so as to produce relevant content targetedly. Based on the characteristics of user sharing behavior and user viewing behavior, the social spread effect of video content is analyzed, which includes evaluating the spread breadth, depth and influence of the content. The spread network analysis method can be used to understand which video segments have the best spread effect in the social network and which content has triggered extensive discussion and sharing by users. For each video segment, a social spread effect feature vector is generated, which will include sharing volume, forwarding rate, viewing time and other multi-dimensional features. Through these vectors, the social spread effect of each video segment can be quantified. If a video segment has a high sharing volume and long viewing time, these dimensions will be given higher weights in the feature vector. Based on the social spread effect features, the conversion rate of the spread effect of each content-labeled video segment is calculated. The conversion rate can be obtained by comparing the effective interactions (such as comments and likes) generated through sharing and viewing with the total viewing volume. If a video segment has a viewing volume of 1000 and an effective interaction of 200 through sharing, the spread conversion rate is 20%. By analyzing the conversion rates of multiple content-labeled video segments, it can be identified which content is more likely to trigger user interaction and sharing, which will directly affect the future direction and strategy of content production.If it is found that the conversion rate of the spread of certain topics is high, the content creators can consider increasing the proportion of such topics in future creation.

[0080] In this embodiment, referring to Figure 4 For the detailed implementation step flowchart of step S3, in this embodiment, the detailed implementation steps of step S3 include:

[0081] Based on the social spread conversion rate, the conversion rate difference is calculated to obtain the conversion rate difference value of different video segments;

[0082] The user feedback value of the social spread effect feature is evaluated to obtain the user feedback value of each video segment;

[0083] Based on the user feedback value, the positive and negative spread effects of the social spread conversion rate are inferred to obtain the positive and negative spread effects of each video segment;

[0084] Based on the positive and negative spread effects, the spread conversion value is evaluated to generate the spread conversion value of each video segment.

[0085] In this embodiment, the conversion rate difference value is calculated, usually using basic statistical analysis methods. It can be obtained by calculating the difference between the conversion rate of each video segment and the average conversion rate of all video segments. If the average conversion rate of the three video segments is 21.67%, the difference of segment 1 is 20%-21.67%=-1.67%, the difference of segment 2 is 30%-21.67%=8.33%, and the difference of segment 3 is 15%-21.67%=-6.67%. These difference values can reflect the relative advantages or disadvantages of each video segment in social communication. The calculated conversion rate difference values of each video segment are arranged into a report for subsequent analysis. In addition, these difference values can be visualized to facilitate a more intuitive understanding of the performance of each video segment. Collect user feedback data related to each video segment, including comments, likes, shares, and watch time, which are direct reflections of user reactions to video content and can provide a basis for subsequent value assessment. If a video segment has 150 comments, 500 likes, and a watch time of 2000 seconds, these data will be important basis for evaluating user feedback value. Based on the collected user feedback data, a feedback value evaluation model can be developed. This model can consider multiple factors, such as the emotional tendency of comments, the proportion of likes, and the watch time. A weighted scoring system can be used to assign different weights to each indicator to calculate the total feedback value. Assuming that the weight of comments is 0.5, the weight of likes is 0.3, and the weight of watch time is 0.2, the total feedback value can be calculated as: feedback value = 0.5 × comments + 0.3 × likes + 0.2 × watch time (unit adjusted to seconds), which will generate a comprehensive feedback value score. When conducting positive and negative propagation effect inference, it is necessary to define what constitutes positive and negative propagation effect. Positive effect is usually related to positive user feedback, sharing and interaction, while negative effect may be related to negative comments or lower watch time. If a video segment has a large number of positive comments and high sharing rate, it can be inferred that it has a positive propagation effect; otherwise, it may have a negative propagation effect. Based on user feedback value and social communication conversion rate, a positive and negative propagation effect inference model is constructed. This model can judge the propagation effect of each video segment by analyzing the emotional tendency (using sentiment analysis tools) and interaction data in user feedback. If 80% of the feedback of a video segment is positive and its conversion rate is higher than the average, it can be determined that the segment has a significant positive propagation effect. Based on positive and negative propagation effect, the propagation conversion value of each video segment is evaluated. Propagation conversion value can be calculated by combining user feedback value and propagation effect using a weighted model. If the weight of positive propagation effect is 0.7 and the weight of negative effect is 0.3, the propagation conversion value can be represented as: propagation conversion value = positive effect × 0.7 - negative effect × 0.3.After calculating the propagation conversion value of each video segment, it can be identified which segments have higher conversion potential in social propagation. This analysis will help creators optimize content strategies and focus on high conversion value content. If the propagation conversion value of a video segment is high and positive feedback is the majority, the creator can consider making follow-up content on related topics.

[0086] In this embodiment, step S4 includes the following steps:

[0087] According to the big data of Internet user feedback, the user barrage content of multiple video segments is extracted;

[0088] Malicious barrage identification is performed on the user barrage content, and malicious barrage is marked;

[0089] High-frequency same barrage of the user barrage content is calculated;

[0090] Abnormal barrage detection is performed on the high-frequency same barrage, and abnormal high-frequency barrage is extracted;

[0091] Barrage cleaning processing is performed on the malicious barrage and abnormal high-frequency barrage, and effective barrage content is obtained;

[0092] User emotional color analysis is performed on the effective barrage content on a video segment by video segment basis, and user emotional color features of each video segment are generated;

[0093] User interest degree quantification analysis is performed according to the effective barrage content, to obtain user interest degree quantification values;

[0094] Personalized emotional information analysis is performed on the user interest degree quantification values and the user emotional color features, to generate user personalized emotional feedback information of each video segment.

[0095] In this example, the user's barrage content is extracted from the video platform, which can usually be obtained through API interface or data crawling technology. Ensure that the data includes the barrage content of each video segment, as well as additional information such as sending time, user ID, etc. Assuming that 1000 barrage contents are extracted from a certain video segment, these data will provide the basis for subsequent analysis. Organize the extracted barrage content and clean up invalid information (such as repeated, empty or format error barrage). Associate each barrage with its corresponding timestamp and user ID to form structured data. Remove 50 invalid barrages from the 1000 barrages, ensuring that the remaining 950 valid data can be used for subsequent analysis. Use natural language processing (NLP) technology to build a malicious barrage identification model. Supervised learning method can be used to train the model with labeled malicious barrage, so that it can identify offensive, insulting or spam information. Use word vector model (such as Word2Vec) to convert barrage content into vector, and train classifier (such as SVM or random forest) to identify malicious barrage. Mark the identified malicious barrage to ensure that valid barrage and malicious barrage can be distinguished in subsequent processing. For each barrage, mark it as "malicious" or "normal" and record its identification result. If 20 barrages are identified as malicious, mark these barrages as "malicious" and annotate them in the data set. Perform frequency statistics on the extracted barrage content to identify high-frequency identical barrages, which can be done by building a word frequency matrix or using a hash table to record the number of times each barrage appears. In 950 barrages, a barrage appears 150 times, which can be marked as high-frequency barrage. Record high-frequency identical barrage and provide data support for subsequent anomaly detection. Ensure that the recorded content includes barrage text and its frequency. Record "barrage content: 'good to see', frequency: 150". Perform anomaly detection on high-frequency barrage to identify those with abnormally high frequency, which can be identified by setting a threshold (such as twice the average number of barrages). If the average number of barrage appearances is 10, barrages with more than 20 appearances are considered abnormal. Extract the identified high-frequency abnormal barrage to ensure that these barrages can be further analyzed. Record the content and frequency of each abnormal barrage. If a barrage appears 300 times while the average is 10, record it as "abnormal barrage: 'too good to see', frequency: 300". Clean up the identified malicious barrage and high-frequency abnormal barrage to remove inappropriate content and ensure that the remaining barrage is valid. Organize the cleaning rules, such as deleting all barrages marked as "malicious" and high-frequency abnormal barrages. After cleaning, organize the valid barrage content and store it as the basis data for downstream analysis. Ensure that each valid barrage reflects the user's real feedback. After cleaning, there are 800 valid barrages left, which will be used for sentiment analysis.For effective BGM content, sentiment analysis techniques such as sentiment lexicon method or deep learning models are used for sentiment color analysis. Identify the sentiment conveyed by each BGM, such as positive, negative or neutral. If the analysis result shows that 500 BGMs are positive, 200 are negative, and 100 are neutral, the sentiment distribution can be clearly recorded. Generate sentiment color features for each video segment, which can usually be represented as the proportion of each sentiment. This feature will provide data support for subsequent personalized emotional feedback information. The sentiment feature of a certain video segment is "positive: 62.5%, negative: 25%, neutral: 12.5%". Based on the content of effective BGM and the degree of user interaction, the degree of user interest in each video segment is quantified. The number of BGMs, sentiment orientation, and user interaction (such as likes and shares) can be comprehensively evaluated. If a video segment has a large number of BGMs and most of them are positive, the user interest level quantification value of that segment can be set to high. Calculate a user interest level quantification value for each video segment, which can usually be represented in the range of 0 to 100, with higher values indicating higher user interest levels. If the quantification value of a certain video segment is 80, it means that the user is very interested in the content. Combine the user interest level quantification value with the user sentiment color feature to generate personalized emotional feedback information for each video segment. By associating sentiment features with user interest levels, more in-depth user feedback perspectives can be provided. If a video segment has a positive sentiment feature and an interest level value of 80, the personalized emotional feedback information for that segment can be generated as "user emotional feedback is positive and interest is strong". Organize the personalized emotional feedback information for each video segment into a report to ensure that the content is clear and easy to read. This report will provide practical insights for content creators to make targeted adjustments in future content production. The report lists the user emotional feedback for each video segment to help creators understand the real feelings and preferences of the audience.

[0096] In this embodiment, step S5 includes the following steps:

[0097] Based on the user personalized emotional feedback information and the propagation conversion value of each video segment, deep content demand mining is performed to obtain user deep content demand features of different video segments;

[0098] Based on the user deep content demand features, sentiment tone and rhythm optimization analysis is performed to obtain sentiment tone and rhythm optimization data for each video segment;

[0099] According to the user deep content demand features, an adaptive expression optimization method is analyzed;

[0100] The adaptive expression optimization method, sentiment tone and rhythm optimization data are identified to generate the video elements to be optimized for each video segment.

[0101] In this embodiment, user personalized emotional feedback information is integrated with the propagation conversion value of each video segment. User feedback information includes emotional inclination, interest level, and interaction data, while the propagation conversion value reflects the propagation effect of the content. Assuming that the user feedback of a certain video segment indicates that the audience's emotional score for the segment is 8 / 10, and its propagation conversion value is 0.3, the integrated data set facilitates subsequent analysis. Deep content demand feature extraction is performed on the integrated data. By analyzing the relationship between user emotional feedback and propagation conversion value, the audience's deep-level demand for content is identified. Some paragraphs may exhibit high emotional scores and high propagation conversion values, indicating that users have strong demand for these contents. Using clustering analysis methods, video segments are divided into different categories, and the common demand features of each category are identified, which may include preferences for specific themes, intensity of emotional resonance, etc. Based on user deep-level content demand features, an emotional tone and rhythm optimization analysis model is constructed. Regression analysis or machine learning algorithms can be used to analyze the relationship between user feedback and video segment content to identify the best emotional tone and rhythm pattern. By analyzing user emotional feedback data, it is identified that under a certain rhythm (such as fast or slow rhythm), users' emotional response is more positive, thus providing data support for rhythm optimization. The emotional tone of user feedback is analyzed to identify users' preferences for different video segment emotional tones. Emotional analysis tools can be used to quantify user feedback and extract emotional tones (such as warm, cold, positive, negative, etc.). If it is found that users prefer the emotional tone of a certain video segment to be "positive" and the emotional score is generally higher than that of other paragraphs, it can be determined that the emotional tone of this segment needs to be maintained or enhanced. Based on the results of emotional tone analysis and user demand features, rhythm optimization data is generated for each video segment, which will include suggested rhythm changes and possible editing points to ensure better user viewing experience. For user-preferred "fast-paced" segments, it is suggested to switch quickly at certain key points to attract audience attention. Based on user deep-level content demand features, suitable adaptive expression methods are analyzed, which may include adjusting the tone of the voiceover, picture style, background music, etc. in the video to better match users' emotional needs. If the emotional tone of user feedback in a certain video segment is generally "positive", it is suggested to use a more cheerful tone and bright pictures when expressing. By evaluating feedback on different expression methods, the most popular expression method is identified. A / B testing methods can be used to analyze user feedback differences under different expression methods. If two different voiceover styles are used in the same paragraph, user viewing time and emotional feedback are analyzed to determine which style is more popular. Based on adaptive expression optimization methods, emotional tone and rhythm optimization data, the elements to be optimized in the video segment are identified, which can be determined by analyzing specific problems mentioned in user feedback, such as "too slow rhythm" or "not strong enough emotion", etc.If users generally feedback that the rhythm of a paragraph is slow and the emotional tone is not positive enough, then that paragraph is an element to be optimized. The identified elements to be optimized are systematically organized to ensure that the optimized parts of each video segment are clearly recorded. These elements may include narration, editing rhythm, picture style, etc. The identified elements to be optimized in a certain video segment are recorded as "narration tone needs to be more lively, editing rhythm needs to be accelerated". The elements to be optimized are organized into a report to facilitate subsequent content production and optimization. Ensure that this information can provide effective improvement suggestions for content creators to improve the overall quality of the video and audience satisfaction. The report will list the video segments to be optimized and their specific problems to help creators make targeted adjustments in subsequent production.

[0102] In this embodiment, step S6 includes the following steps:

[0103] Video position positioning is performed on the video elements to be optimized, and the video positions of each video element to be optimized are extracted;

[0104] Dynamic content tuning is performed based on the video positions to generate multi-video segment tuning content;

[0105] Global content detail rationality analysis is performed on the multi-video segment tuning content to extract the detail rationality;

[0106] Abnormal logic recognition is performed according to the detail rationality, and abnormal logic content is marked;

[0107] Content detail local optimization is performed on the abnormal logic content, and global coordination fitting is performed to construct a feedback optimized new media video.

[0108] In this example, for the video segment to be optimized, a frame-by-frame analysis of the video content is performed. Using computer vision techniques, the specific locations of the elements to be optimized in the video are identified, which may include voiceover segments, specific pictures, or subtitles. If there is a voiceover in a certain video that needs to be optimized, the start and end timestamps of each voiceover will be recorded for subsequent positioning. Video frame extraction is performed using video editing and analysis tools such as FFmpeg or OpenCV. By extracting each frame, combined with timestamp information, the specific location of the element to be optimized in the time axis and picture is determined. From a 10-minute video, 100 frames are extracted, and if it is found that the voiceover to be optimized starts at 2 minutes and ends at 2 minutes and 30 seconds, its position is recorded as "2:00-2:30". The position of each element to be optimized is organized into structured data to ensure that the time and position of each element can be accurately referenced, and this record will provide the necessary basis for subsequent dynamic content optimization. For the voiceover to be optimized, it is recorded as "Voiceover Position: 2:00-2:30" to quickly locate it during subsequent adjustments. According to the position of the element to be optimized, a dynamic content optimization strategy is designed. The goal of optimization is to adjust the key elements in the video in real time according to user feedback and content needs. For parts of the voiceover found to be too slow, consider shortening the voiceover time or speeding up the voiceover speed. In the video editing software, based on the recorded positions, dynamic adjustments are made to the elements to be optimized. Optimization can be achieved through time axis scaling, speed adjustment, or voiceover tone change. In the 2:00 to 2:30 time period, the voiceover speed is increased by 20%, and the rhythm of the background music is appropriately increased to make the overall content more compact and engaging. Record the changes in the content of each optimized video segment for subsequent effect analysis. Ensure that each adjustment reflects changes in user needs. The adjusted video feedback shows that the user viewing time has increased from an average of 3 minutes to 4 minutes, indicating that the optimization was successful. Global content detail rationality analysis is performed on the content of the optimized video segments. By comparing the original video with the optimized video, the logical coherence, emotional consistency, and rhythm are evaluated. A combination of expert review and user feedback can be used to collect opinions from all aspects to ensure comprehensive analysis. During the analysis process, focus on the transition effects between video segments, emotional expression, and the effectiveness of information transmission. Record the performance of each paragraph, especially its relevance to audience emotional feedback. If the emotional expression of a paragraph is found to be inconsistent with the previous and subsequent paragraphs, mark it as an unreasonable detail. Organize the results of the detail rationality analysis into a report to ensure that the content is clear and easy to understand, which will provide data support for subsequent abnormal logic identification. The report states that "paragraph 3 and paragraph 4 have inconsistent emotions and need to be adjusted." Based on the results of the detail rationality analysis, use logic analysis tools to identify abnormal logic in the video content. Natural language processing techniques can be used to analyze the content of voiceovers and subtitles to check if they conform to logical reasoning.If the event mentioned in a paragraph contradicts the description in the previous paragraph, it is marked as abnormal logic. Mark the identified abnormal logic to quickly locate the problem in subsequent optimization. These markers will help content creators identify the parts that need to be addressed. If it is found that "the character mentioned in paragraph 2 is inconsistent with paragraph 3", it is marked as "abnormal logic: character inconsistency". Compile the results of abnormal logic identification into a document to ensure that each logical problem can be effectively tracked. This document will provide a reference for subsequent content detail optimization. Record "the character inconsistency problem between paragraph 2 and paragraph 3". Based on the identified abnormal logic, develop a local optimization strategy to ensure that the logic is reasonably presented in the video. Optimization may include modifying the voiceover content, adjusting the video clips, or re-producing certain scenes. To address the "character inconsistency" issue, you may need to re-record the voiceover in paragraph 3 to match the description in paragraph 2. Perform local optimization in video editing software to ensure that the adjusted content is coherent and logical. You can analyze each adjustment frame by frame to ensure it meets the desired effect. In the voiceover of paragraph 3, add the character traits mentioned in the previous paragraph to make the logic more coherent. After completing the local optimization, perform effect verification. You can collect user feedback and expert reviews to ensure that the optimized content effectively addresses the previously identified issues. Collect user feedback again, and if users have improved understanding of the paragraph and increased viewing time, it indicates that the optimization is effective. Coordinate local optimization with global content to ensure consistency in the overall structure and emotional expression of the video. You can ensure smooth transitions between paragraphs by playing back and reviewing the entire video. Adjust the transition music and visual effects between paragraphs to make the overall viewing experience smoother. Evaluate the optimized new media video as a whole, collect audience feedback and ratings, and ensure that the new video meets user expectations and needs.

[0109] In this embodiment, an internet new media data feedback generation system is provided for executing the internet new media data feedback generation method as described above, comprising:

[0110] A semantic analysis module is configured to obtain new media videos under production and internet user feedback big data, perform multi-period decomposition and in-depth semantic analysis of each video segment on the new media videos, and construct a plurality of content tag video segments.

[0111] A propagation conversion rate module is configured to perform social propagation effect mining and video propagation effect conversion rate calculation based on the internet user feedback big data to generate a social propagation conversion rate of each video segment.

[0112] A propagation value evaluation module is configured to perform positive and negative propagation effect inference based on the social propagation conversion rate, and evaluate the propagation conversion value of the plurality of content tag video segments to generate a propagation conversion value of each video segment.

[0113] an emotional feedback analysis module for quantitatively analyzing the interest degree of users and analyzing the emotional color according to the big data of Internet user feedback, to generate user personalized emotional feedback information of each video segment;

[0114] a content demand mining module for deep content demand mining based on the propagation conversion value and the emotional feedback information, and identifying the to-be-optimized elements, to generate to-be-optimized video elements of each video segment;

[0115] a dynamic content optimization module for dynamic content optimization according to the to-be-optimized video elements, and global coordination fitting, to construct a feedback-optimized new media video.

[0116] The video is decomposed into multiple video segments and deep semantic analysis is performed, so that the content of each video segment can be carefully labeled, and subsequent analysis and optimization can be based on accurate data. Through multi-dimensional data labeling, the platform can better perform personalized recommendation and precise content pushing, improving user viewing satisfaction. Reliable basic data (such as the emotion, theme, and key plot of the video content) is provided for subsequent modules, ensuring the data accuracy and consistency of the entire optimization process. The propagation effect conversion rate of each video segment is calculated, providing quantitative data for the video segment propagation effect for the creator, which enables the content creator to accurately assess which paragraphs have high propagation potential and which need further optimization. According to the propagation conversion rate, the creator can selectively optimize the paragraphs with poor propagation effect, thereby improving the overall content propagation effect. Data support is provided for creators and marketers, enabling them to decide on advertising investment, content promotion, and other strategies based on the propagation effect, further enhancing the social communication influence of the content. By evaluating the propagation conversion value of each video segment, the creator can determine which paragraphs are popular and have positive effects, and which paragraphs fail to attract the audience. This provides guidance for subsequent content optimization. The high propagation value of positive propagation effect can provide feedback to the creator, indicating which plots or elements are successful, and vice versa. Negative propagation effect can prompt the creator to adjust or optimize the content that may have problems.

[0117] By disseminating the evaluation of conversion value, creators can more accurately adjust the content structure to maximize the dissemination value of the content and audience engagement. Through emotional color analysis and interest quantification, creators can more clearly understand the emotional response of the audience to different segments of the video, which can help creators identify which content causes strong audience reactions and which fails to produce the expected emotional resonance. According to the emotional feedback of different audience groups, creators can adjust the video content to better meet the emotional needs of users. Some users may prefer more light-hearted content, while others may prefer more challenging or in-depth content. By better understanding user emotional responses, creators can produce more attractive content to users, thereby improving audience engagement, interaction frequency, and content sharing rate. Through deep demand mining, creators can identify which content elements are most important to the audience and focus on optimizing these elements. By identifying elements to be optimized, creators can apply improvement measures to specific video segments to ensure that optimization can hit the core needs of the audience and improve the watchability and audience satisfaction of the content. Deep demand analysis helps content creators understand market trends and audience preferences, thereby improving the market adaptability and competitiveness of the content. Dynamic content optimization can respond to user feedback and data analysis in real time, ensuring that creators can make necessary optimizations before and after video release, so that the video content maintains high quality at multiple stages. Through global coordination fitting, the video is coordinated in terms of emotional tone, rhythm, etc., thereby improving the viewing experience of the audience. Precise adjustment of content not only improves the watchability of the video, but also improves the emotional resonance of users, thereby promoting the word-of-mouth spread and social spread effect of the video.

[0118] Therefore, the embodiments should be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the description given above, and all changes which come within the meaning and range of equivalency of the claims are intended to be embraced therein.

[0119] The foregoing merely illustrates the principles of the application and applies only to the particular cases described and illustrated herein. Numerous modifications and adaptations will be apparent to those skilled in the art without departing from the spirit and scope of the present application. Therefore, the scope of the present application is not intended to be limited to the particular embodiments described and illustrated herein, but is intended to cover all modifications and adaptations that come within the scope of the following claims.

Claims

1. An Internet new media data feedback generation method, characterized in that, The method comprises the following steps: Step S1: obtaining new media videos in production and Internet user feedback big data; performing multi-period decomposition and deep semantic analysis of each video segment on the new media videos to construct multiple content label video segments; Step S2: performing social communication effect mining and video communication effect conversion rate calculation according to the Internet user feedback big data to generate a social communication conversion rate of each video segment; Step S3: performing positive and negative communication effect inference according to the social communication conversion rate, and performing communication conversion value evaluation on the multiple content label video segments to generate a communication conversion value of each video segment; Step S4: performing user interest degree quantification analysis and emotional color analysis according to the Internet user feedback big data to generate user personalized emotional feedback information of each video segment; Step S5: performing deep content demand mining based on the communication conversion value and the emotional feedback information, and identifying to-be-optimized elements to generate to-be-optimized video elements of each video segment; Step S6: performing dynamic content optimization according to the to-be-optimized video elements, and performing global coordination fitting to construct feedback-optimized new media videos. The specific steps of step S3 are: performing conversion rate difference calculation based on the social communication conversion rate to obtain conversion rate difference values of different video segments; performing user feedback value evaluation on social communication effect characteristics to obtain a user feedback value of each video segment; performing positive and negative communication effect inference on the social communication conversion rate based on the user feedback value to obtain a positive and negative communication effect of each video segment; performing communication conversion value evaluation based on the positive and negative communication effects to generate a communication conversion value of each video segment. 2.The method of claim 1, wherein, The specific steps of step S1 are: obtaining new media videos in production and Internet user feedback big data; performing multi-period decomposition and deep semantic analysis of each video segment on the new media videos to generate deep semantic characteristics of each video segment; performing content label classification according to the deep semantic characteristics to generate accurate content labels of each video segment; performing time sequence positioning marker fitting according to the accurate content labels to construct multiple content label video segments. 3.The method of claim 2, wherein, The specific steps of performing multi-period decomposition and deep semantic analysis of each video segment on the new media videos to generate deep semantic characteristics of each video segment are: identifying the full video content rhythm and scene transformation turning points of the new media videos; performing optimal segmentation time point calculation based on the content rhythm and scene transformation turning points to extract multiple video segmentation time points; performing multi-period decomposition based on the multiple video segmentation time points to obtain video segments of multiple time windows; performing key frame extraction on the video segments of multiple time windows to obtain key frames of each video segment; extracting key semantic fields and key content elements of the key frames; performing latent association analysis on the key semantic fields and key content elements to obtain latent semantic subject feature vectors; performing deep semantic analysis of each video segment based on the latent semantic subject feature vectors to generate deep semantic characteristics of each video segment. 4.The method of claim 1, wherein, The specific steps of step S2 are: Calculate the user viewing time, video click volume and play volume of each video segment of the internet user feedback big data; Analyze the user behavior based on the user viewing time, video click volume and play volume to obtain user viewing behavior characteristics; Calculate the sharing volume and forwarding rate of each video segment of the internet user feedback big data; Analyze the user sharing behavior based on the sharing volume and forwarding rate to generate user sharing behavior characteristics; Based on the user sharing behavior characteristics and the user viewing behavior characteristics, mine the social communication effect to generate social communication effect characteristics; Based on the social communication effect characteristics, calculate the social communication conversion rate of each video segment to generate the social communication conversion rate of each video segment.

5. The Internet new media data feedback generation method of claim 1, wherein, The specific steps of step S4 are: Extract the user barrage content of multiple video segments from the internet user feedback big data; Identify malicious barrages based on the user barrage content, and mark the malicious barrages; Calculate the high-frequency same barrage of the user barrage content; Detect abnormal barrages based on the high-frequency same barrage, and extract abnormal high-frequency barrages; Clean the malicious barrages and abnormal high-frequency barrages to obtain effective barrage content; Analyze the user emotional color of each video segment based on the effective barrage content to generate user emotional color characteristics of each video segment; Quantitatively analyze the user interest degree based on the effective barrage content to obtain user interest degree quantitative values; Analyze the user interest degree quantitative values and the user emotional color characteristics to generate user individualized emotional feedback information of each video segment. 6.The method of claim 1, wherein, The specific steps of step S5 are: Based on the user individualized emotional feedback information and the communication conversion value of each video segment, mine the deep content demand to obtain user deep content demand characteristics of different video segments; Based on the user deep content demand characteristics, analyze the emotional tone and rhythm optimization to obtain emotional tone and rhythm optimization data of each video segment; Analyze the adaptive expression optimization mode based on the user deep content demand characteristics; Identify the to-be-optimized elements based on the adaptive expression optimization mode, emotional tone and rhythm optimization data to generate to-be-optimized video elements of each video segment.

7. The Internet new media data feedback generation method of claim 1, wherein, The specific steps of step S6 are: Position the to-be-optimized video elements to extract the video position of each to-be-optimized video element; Based on the video position, dynamically optimize the content to generate multiple video segment optimized content; Analyze the global content details rationality of the multiple video segment optimized content to extract the details rationality; Identify abnormal logic based on the details rationality, and mark the abnormal logic content; Optimize the content details of the abnormal logic content locally, and perform global coordination fitting to construct a feedback optimized new media video.

8. An Internet new media data feedback generation system, characterized in that, The internet new media data feedback generation method comprises the following steps: A semantic analysis module is configured to obtain a new media video under production and internet user feedback big data, decompose the new media video into multiple time periods, and perform depth semantic analysis on each video segment to construct multiple content label video segments; The propagation conversion rate module is configured to perform social propagation effect mining and video propagation effect conversion rate calculation according to the Internet user feedback big data, so as to generate a social propagation conversion rate of each video segment. The propagation value evaluation module is configured to perform positive and negative propagation effect inference according to the social propagation conversion rate, and evaluate the propagation conversion value of the plurality of content label video segments, so as to generate a propagation conversion value of each video segment. The emotional feedback analysis module is configured to perform user interest degree quantitative analysis and emotional color analysis according to the Internet user feedback big data, so as to generate user individualized emotional feedback information of each video segment. The content demand mining module is configured to perform deep-level content demand mining based on the propagation conversion value and the emotional feedback information, and identify to-be-optimized elements, so as to generate to-be-optimized video elements of each video segment. The dynamic content tuning module is configured to perform dynamic content tuning according to the to-be-optimized video elements, and perform global coordination fitting, so as to construct a feedback-optimized new media video.

Images