Automatic film content rating method and system based on multimode data prediction, terminal and storage medium

Abstract

The invention discloses an automatic film content rating method and system based on multimode data prediction, a terminal and a storage medium. The method comprises the steps of receiving film related information, input by a user, of a to-be-evaluated film, obtaining multi-dimensional data, and integrating the multi-dimensional data and the film related information to obtain a film data set; according to the movie data set, analyzing movie data of the to-be-evaluated movie to obtain a plurality of movie analysis results; based on the film related information, performing preliminary rating prediction on the to-be-evaluated film, and generating a preliminary rating prediction result; and constructing a multi-factor weighting model, and performing content rating on the to-be-evaluated film according to the film data set, the plurality of film analysis results and the preliminary rating prediction result to obtain a final rating result. According to the method, the information of the film is considered from multiple perspectives, the film content is comprehensively, accurately, efficiently and objectively rated, a large number of films can be rated in a short time, and the accuracy and reliability of rating are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet film and television content evaluation, and particularly to an automated film content rating method, system, terminal, and computer-readable storage medium based on multi-modal data prediction. Background Art

[0002] Traditional film content rating mainly relies on professional reviewers to rate according to their own experience and established criteria after manually watching the film.

[0003] With the rapid development of the film and television industry, the number of films has increased sharply. Manual rating requires a large amount of time to watch films one by one, and it is difficult to meet the rapidly growing film rating demand. For example, the number of newly added films on a large video platform may reach thousands per day. If manual rating is adopted, it cannot be processed in time, which will lead to delays in film release and affect the operation efficiency of the platform.

[0004] There are differences in the understanding and judgment criteria of film content among different reviewers. For example, for some controversial plots or expression techniques in a film, different reviewers may give different rating results due to their personal cognitions and values, resulting in the lack of consistency and reliability in ratings.

[0005] In addition, subjective factors such as personal preferences and cultural backgrounds of reviewers will also have a greater impact on the rating results. For example, reviewers who prefer a certain type of film (such as art films) may give unobjective evaluations to other types of films (such as action films), and cannot truly reflect the actual content quality and suitable viewing population of the film.

[0006] At present, there are some simple automated auxiliary tools, such as only scanning keywords in the text script of the film to judge the content type. However, rating based on a single data type cannot comprehensively reflect the actual content of the film. A film is a comprehensive art form composed of multiple elements, including pictures, sound effects, performances, etc. For example, the script of some films may be relatively plain, but through means such as picture processing and sound effect rendering during shooting, it may present strong sensory stimuli or implicit bad content, and the rating method based only on script keyword scanning cannot accurately evaluate these contents. In addition, the analysis of single data is easily affected by the limitations of the data itself, resulting in inaccurate rating results. For example, when judging the emotional atmosphere in a film, it is difficult to consider the emotional intensity and potential impact conveyed by actors through expressions, intonations, body languages, etc. only by analyzing the text in the script, so the actual emotional tendency and content suitability of the film cannot be accurately evaluated.

[0007] Therefore, the existing technology still needs to be improved and developed. Summary of the Invention

[0008] The main purpose of the present invention is to provide an automated film content rating method, system, terminal and computer-readable storage medium based on multimodal data prediction, aiming to solve the problem in the prior art that film content cannot be rated comprehensively, accurately, efficiently and objectively.

[0009] To achieve the above object, the present invention provides an automatic film content rating method based on multimodal data prediction, and the automatic film content rating method based on multimodal data prediction comprises the following steps:

[0010] Receive the film-related information of the film to be evaluated input by the user, connect to the upstream new film storage system, obtain the multi-dimensional data of the film to be evaluated, and integrate the multi-dimensional data with the film-related information through data fusion and completion algorithms to obtain a film data set;

[0011] According to the film data set, analyzing the film data of the film to be evaluated at different preset angles to obtain multiple film analysis results;

[0012] Based on the film-related information, using natural language processing technology and machine learning algorithms to make a preliminary rating prediction for the film to be evaluated, and generate a preliminary rating prediction result;

[0013] A multi-factor weighted model is constructed to perform content rating on the to-be-evaluated film based on the film data set, the plurality of film analysis results and the preliminary rating prediction results, and based on the rating criteria to obtain a final rating result.

[0014] Optionally, in the automated film content rating method based on multimodal data prediction, the multi-dimensional data includes: basic information of the film, basic information of the filmmakers, peripheral information of the film, historical data of the same subject matter, historical data of the main staff and public opinion data of the entire network.

[0015] Optionally, the method for automatic film content rating based on multimodal data prediction, wherein the step of analyzing the film data of the film to be evaluated at different preset angles according to the film data set to obtain multiple film analysis results, specifically includes:

[0016] Collecting user comment data on the film to be evaluated by using web crawler technology, analyzing the comment data by using natural language processing technology, and obtaining comment analysis results;

[0017] Setting keyword filtering and event monitoring algorithms to monitor negative public opinion events related to the film to be evaluated in real time;

[0018] Conduct data mining and analysis on the market performance and ratings of films of the same type as the film to be evaluated in historical periods, and calculate statistical indicators;

[0019] Analyze the popularity trends and hot topics of current films of the same type as the film to be evaluated, and obtain the impact on the film to be evaluated based on the popularity trends and hot topics of current films of the same type;

[0020] Through industry reports and market data, statistically analyze the average commercial revenue and profit models of films with the same theme, and evaluate the potential and rating impact of the film to be evaluated in terms of commercial value;

[0021] Track the recent creative styles and public opinion images of specific creators of the film to be evaluated, and evaluate the potential impact of specific creators on the film to be evaluated.

[0022] Optionally, in the automated film content rating method based on multi-modal data prediction, the rating criteria include: level definition, sensitive content quantification criteria, weighting or de-weighting mechanisms, rating dimension definitions, rating scoring criteria, and rating confidence criteria.

[0023] Optionally, in the automated film content rating method based on multi-modal data prediction, in the construction of the multi-factor weighted model, according to multiple film analysis results and the preliminary rating prediction results, conduct content rating on the film to be evaluated to obtain the final rating result. After that, it further includes:

[0024] Output the final rating result in an intuitive manner, including the final rating conclusion and detailed rating reasons of the film to be evaluated.

[0025] Optionally, in the automated film content rating method based on multi-modal data prediction, in the construction of the multi-factor weighted model, according to multiple film analysis results and the preliminary rating prediction results, conduct content rating on the film to be evaluated to obtain the final rating result. After that, it further includes:

[0026] When there is a dispute over the final rating result, provide a manual review mechanism. Professional reviewers conduct a re-review of the film to be evaluated and adjust the final rating result according to the actual situation. At the same time, record the process and results of the manual review for optimizing and improving the automatic rating algorithm.

[0027] Optionally, in the automated film content rating method based on multi-modal data prediction, the automated film content rating method based on multi-modal data prediction further includes:

[0028] Regularly update various types of data in the system, and optimize the multi-factor weighted model according to the new data;

[0029] Analyze the reasons for the deviation of the system rating by combining the feedback and adjustment results of reviewers during the manual review process, and improve the feature selection and weight assignment mechanisms of the multi-factor weighted model accordingly.

[0030] In addition, to achieve the above object, the present invention also provides an automated movie content rating system based on multi-modal data prediction. Among them, the automated movie content rating system based on multi-modal data prediction includes:

[0031] An information input and completion module, configured to receive movie-related information of the movie to be evaluated input by a user, connect to an upstream new movie warehousing system, obtain multi-dimensional data of the movie to be evaluated, and integrate the multi-dimensional data with the movie-related information through a data fusion and completion algorithm to obtain a movie data set;

[0032] A multi-modal data perception module, configured to analyze the movie data of the movie to be evaluated from different preset angles according to the movie data set to obtain a plurality of movie analysis results;

[0033] A rating intelligent prediction module, configured to perform a preliminary rating prediction on the movie to be evaluated based on the movie-related information by using natural language processing technology and machine learning algorithms to generate a preliminary rating prediction result;

[0034] A content rating engine module, configured to construct a multi-factor weighted model, and perform content rating on the movie to be evaluated according to the movie data set, a plurality of the movie analysis results, and the preliminary rating prediction result, and based on a rating standard to obtain a final rating result.

[0035] In addition, to achieve the above object, the present invention also provides a terminal. Among them, the terminal includes: a memory, a processor, and an automated movie content rating program based on multi-modal data prediction stored on the memory and executable on the processor. When the automated movie content rating program based on multi-modal data prediction is executed by the processor, the steps of the above-mentioned automated movie content rating method based on multi-modal data prediction are implemented.

[0036] In addition, to achieve the above object, the present invention also provides a computer-readable storage medium. Among them, the computer-readable storage medium stores an automated movie content rating program based on multi-modal data prediction. When the automated movie content rating program based on multi-modal data prediction is executed by a processor, the steps of the above-mentioned automated movie content rating method based on multi-modal data prediction are implemented.

[0037] In the present invention, film-related information of a film to be evaluated input by a user is received, and it is connected to an upstream new film warehousing system to obtain multi-dimensional data of the film to be evaluated. Through a data fusion and completion algorithm, the multi-dimensional data is integrated with the film-related information to obtain a film data set. According to the film data set, the film data of the film to be evaluated is analyzed from different preset angles to obtain multiple film analysis results. Based on the film-related information, the film to be evaluated is initially rated and predicted by using natural language processing technology and machine learning algorithms to generate an initial rating prediction result. A multi-factor weighted model is constructed, and according to the film data set, multiple film analysis results and the initial rating prediction result, and based on a rating standard, the content rating of the film to be evaluated is performed to obtain a final rating result. The present invention considers the information of the film from multiple angles, comprehensively, accurately, efficiently and objectively rates the film content, can rate a large number of films in a short time, and improves the accuracy and reliability of the rating. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flowchart of a preferred embodiment of the automated film content rating method based on multi-modal data prediction of the present invention;

[0039] Figure 2 is a schematic diagram of the entire process of rating the film content in a preferred embodiment of the automated film content rating method based on multi-modal data prediction of the present invention;

[0040] Figure 3 is a structural diagram of a preferred embodiment of the automated film content rating system based on multi-modal data prediction of the present invention;

[0041] Figure 4 is a structural diagram of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] To make the objectives, technical solutions and advantages of the present invention clearer and more definite, the following further describes the present invention in detail with reference to the accompanying drawings and by way of examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0043] The automated film content rating method based on multi-modal data prediction according to a preferred embodiment of the present invention, as Figure 1 and Figure 2 shown, the automated film content rating method based on multi-modal data prediction includes the following steps:

[0044] Step S10: receiving the film-related information of the film to be evaluated input by the user, connecting with the upstream new film storage system, obtaining the multi-dimensional data of the film to be evaluated, integrating the multi-dimensional data with the film-related information through data fusion and completion algorithms, and obtaining a film data set.

[0045] Specifically, the system receives film-related information input by the user, including the film name and script information, etc. The user can conveniently input this information through the system interface, and the input script information will serve as one of the important data sources for subsequent content analysis. For example, the user enters the name of the new movie "XXX" in the operation interface and uploads its script file (supports common text formats such as .txt, .doc, etc.). The system parses the uploaded script file, extracts text information such as characters, plots, dialogues, etc., and structures this information for subsequent analysis.

[0046] like Figure 2 As shown in the figure, the data flow and connection relationship between each step are presented in detail, which intuitively reflects the operation mechanism of the entire rating system. It can clearly understand the whole process of the system from user input to the final rating conclusion output, as well as the role of each part in it; it is connected to the upstream new film storage system to obtain multi-dimensional data such as basic information of the film, basic information of the filmmakers, information around the film, historical data of the same theme, historical data of the main staff, and public opinion data of the whole network. Through data fusion and completion algorithms, these multi-dimensional data are integrated with the film-related information input by the user to form a comprehensive and rich film data set. For example, by associating the information of the characters and filmmakers in the film, as well as the film type and historical data of the same theme, a complete data foundation is provided for subsequent multi-modal data perception and rating.

[0047] Obtain multi-dimensional data of the film from the upstream new film storage system, including:

[0048] Basic information of the film: such as genre (drama, action, science fiction, etc.), duration, production cost, filming location, etc. This information can be obtained through the metadata of the film, and the system imports this data into the local database through the data interface.

[0049] Basic information of filmmakers: including the resumes (previous works, awards, etc.) and reputations (industry evaluation, audience reputation, etc.) of major creative personnel such as directors and leading actors. This information can be obtained from professional film and television databases or related websites, and associated with the current film through data crawling and integration technology.

[0050] Film peripheral information: such as trailers, posters, promotional materials, etc. The system can obtain this information from the film's official website, social media accounts and other channels, and analyze it to extract factors that may affect the rating, such as the style of the trailer, the visual elements of the poster, etc.

[0051] Obtain multi-dimensional data of the film through the aggregation data completion engine, including:

[0052] Historical data of the same genre: including the average rating, market share, box office revenue, audience rating distribution, etc. of the same type of films in the past certain period of time (e.g., the past 1-3 years). These data can be obtained from industry statistical reports, professional data institutions, etc., and through data mining and analysis technology, provide historical reference for the rating of the current film.

[0053] Historical data of main staff: the ratings, reputation, and commercial performance of previous works of major creative staff such as directors and leading actors. By analyzing their creative trajectories, their potential impact on the current film can be evaluated. This data can be obtained from film and television databases and related film review websites, and then integrated and analyzed.

[0054] Public opinion data on the entire network: Collect public opinion data such as the popularity of topics related to the film, the proportion of positive or negative reviews, hot events, etc. from social media platforms (such as Weibo, WeChat, Douyin, etc.), news media, and film review websites. Through web crawler technology and natural language processing technology, these data are monitored and analyzed in real time to capture public opinion trends that may affect the film rating in a timely manner.

[0055] When performing data fusion, technical means such as data cleaning, entity alignment, and feature fusion are used to fuse multimodal data from different data sources. For example, for the character data in a film, the character features extracted from the script information are fused with the actor image features obtained from the basic information of the filmmakers. By constructing a unified character feature vector, the system can fully understand the character image and potential impact in the film. The specific data fusion process includes preprocessing the data to remove noise and redundant information, then matching and aligning related entities in different data sets through entity recognition technology, and finally using feature engineering methods to fuse the aligned entity features to form a comprehensive feature representation.

[0056] Using data fusion technology, integrate the data obtained from the upstream new film warehousing with the script data input by users. For example, through entity recognition and relationship extraction technology, match the characters in the script with the actors in the basic information of film personalities, and at the same time associate the genre information of the film with the historical data of the same theme to construct a complete film information graph. During the process of constructing the graph, use graph database technology, take entities such as films, characters, and events as nodes, and the relationships between them as edges, and store, query, and analyze the film-related information through graph algorithms to provide a rich data basis for subsequent multi-modal data perception and rating.

[0057] Step S20: Analyze the film data of the film to be evaluated from different preset perspectives according to the film data set, and obtain multiple film analysis results.

[0058] Specifically, collect the comment data of users on the film to be evaluated from major social media platforms (such as Weibo, Douban, etc.), film review websites (such as Mtime, etc.) through web crawler technology. Analyze the comments using natural language processing technology (including lexical analysis, syntactic analysis, and semantic analysis), including sentiment analysis, theme mining, etc. For example, count the proportions of positive, negative, and neutral comments, and mine the main content themes concerned by users, such as evaluations in aspects such as plot, performance, and special effects, to provide data support for film rating in terms of user feedback. For example, perform part-of-speech tagging on the words in the comments, analyze the sentence structure, and extract key semantic information. Then use sentiment analysis algorithms, such as deep learning-based sentiment classifiers, to classify user comments into positive, negative, and neutral categories, and further mine the themes and key opinions in the comments. For example, if a large number of user comments mention that a certain inappropriate scene in the film is too serious, the system will record this negative feedback and use it as a factor affecting the film rating. At the same time, through the mining of comment themes, understand the specific evaluations of users on various aspects of the film (such as plot, performance, special effects, etc.) to provide detailed data for comprehensively evaluating the user word-of-mouth of the film.

[0059] Real-time monitor negative public opinion events related to the film to be evaluated, including negative news during the film production process, inappropriate behaviors of actors, etc. Through setting keyword filtering and event monitoring algorithms, conduct real-time scanning of channels such as news media and social media. For example, once negative news about the lead actor of the film appears, promptly capture and evaluate the potential impact of this event on the overall image and rating of the film. Through setting keyword filtering and event monitoring algorithms, conduct real-time monitoring of channels such as news media and social media. For example, establish indexes for keywords related to the film (such as the film name, names of main actors, etc.). When these keywords appear in news headlines or popular topics on social media, the system automatically grabs the relevant content for analysis. If negative news about the lead actor of the film appears, the system will promptly capture this public opinion and evaluate the degree of impact of this event on the film rating according to a preset risk assessment model. This risk assessment model will comprehensively consider factors such as the spread range, severity of the negative news, and its relevance to the film itself.

[0060] Conduct data mining and analysis on the market performance and rating situations of films of the same type over a certain period in the past (such as the past 1 - 3 years). Through data mining techniques, extract relevant data of films of the same theme from industry databases, including box office revenues, audience ratings, rating distributions, etc. For example, use clustering analysis methods to classify films of the same theme according to their box office and rating characteristics, calculate statistical indicators such as the average audience rating of each category of films and the proportion of each level of ratings, and then analyze the overall development trend of films of the same theme. These data will provide historical reference data for the rating of the current film and help determine the approximate position of the current film among films of the same type.

[0061] Analyze the popular trends and hot topics of the current films of the same type as the film to be evaluated. By monitoring social media hot topics, discussion focuses in industry forums, etc., understand the current creative trends and audience preferences of films of the same theme. For example, use text mining techniques to analyze topics related to films of the same theme on social media, extract popular topic tags and keywords, and identify the popular plot patterns (such as reverse plots, hero growth, etc.), style characteristics (such as retro style, cyberpunk style, etc.) or presentation techniques (such as first-person perspective shooting, non-linear narrative, etc.) of current films of the same type. This information on popular trends will be used as one of the reference factors for the rating of the current film. If the current film to be rated conforms to the popular trends, it may have a positive impact on its rating.

[0062] Collect and analyze the commercial revenue of films of the same theme, including box office revenue, derivative product revenue, etc. Through industry reports and market data, statistically analyze the average commercial revenue and profit models of films of the same theme. For example, analyze the ratio of the box office revenue to the production cost of films of the same theme to understand the profitability of this type of film. At the same time, pay attention to the market performance of derivative products (such as toys, games, novels, etc.) of films of the same theme and evaluate their commercial value. These commercial revenue data will help evaluate the potential of the current film in terms of commercial value and thus affect its rating. That is, evaluate the potential of the current film in terms of commercial value and the possible rating impact.

[0063] Track the recent creative styles and public opinions of specific creative personnel of the film (i.e., the main creative personnel, such as the director, lead actor, etc.). By analyzing the style characteristics of the recent works of these creative personnel and their reputations in the media and among the audience, evaluate their potential impact on the current film. For example, if the recent works of the director have a good reputation, his current film may be positively affected to a certain extent during the rating process. That is, by analyzing the film reviews of the director's recent works, summarize the evolution of his creative style (such as from art films to commercial films) and the acceptance level of the audience. If the director's recent works are known for their high quality and are deeply loved by the audience, then this will help improve the rating of the current film; conversely, if the director's recent works have a poor reputation, it may have a negative impact on the rating of the current film.

[0064] Step S30: Based on the film-related information, use natural language processing technology and machine learning algorithms to make a preliminary rating prediction for the film to be evaluated, and generate a preliminary rating prediction result.

[0065] Specifically, based on the film-related information input by the user (mainly the script information), use natural language processing technology and machine learning algorithms to make a preliminary rating prediction for the film to be evaluated. For example, by analyzing the vocabulary, sentence structure, plot clues, etc. in the script, identify the key factors that may affect the rating, analyze the potential degree of relevant content, and generate a preliminary rating prediction result accordingly.

[0066] The system inputs the script information input by the user into a pre-trained intelligent prediction model. This model is based on deep learning algorithms, such as the Transformer architecture, and performs semantic analysis on the script sentence by sentence and paragraph by paragraph. For example, the model identifies plot clues that may involve sensitive content by analyzing the scene descriptions, character actions, and dialogues in the script, and predicts the preliminary rating range of the film based on these clues (such as it may be rated A or B). During the analysis process, the model uses its self-attention mechanism to effectively capture the long-distance semantic dependencies in the text, thereby more accurately judging the development of the plot and potential content risks.

[0067] When making preliminary predictions, based on deep learning architectures such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) and their variants (such as LSTMs, GRUs), a deep semantic analysis is performed on the movie script. By pre-training and fine-tuning on a large-scale movie script dataset, the algorithm can accurately capture the content clues in the script and predict the possible rating range of the movie. For example, during training, the words in the script are vectorized and input into the neural network. Through a multi-layer network structure, the semantics of the script are abstracted and represented layer by layer, and finally the preliminary rating prediction result of the movie is output. At the same time, in order to improve the generalization ability of the algorithm, technical means such as data augmentation and regularization are adopted to prevent overfitting.

[0068] Step S40: Construct a multi-factor weighted model, and based on the movie dataset, multiple movie analysis results, and the preliminary rating prediction result, and based on the rating criteria, perform content rating on the movie to be evaluated to obtain the final rating result.

[0069] Specifically, multiple data (i.e., the movie dataset, multiple movie analysis results, and the preliminary rating prediction result) are integrated. Based on the rating criteria, a complex rating algorithm is used to perform content rating on the movie to be evaluated. This algorithm comprehensively considers multiple factors, including the quantification degree of sensitive content in the movie (such as the proportion and performance intensity of relevant content), the comprehensive user review score, the reference weight of historical data of the same theme, the public opinion risk assessment, etc. For example, by constructing a multi-factor weighted model, corresponding weights are assigned to each factor according to its importance to the rating, and the final rating result of the movie to be evaluated is calculated.

[0070] The present invention defines the standards and specifications of the entire rating system, that is, the rating criteria include:

[0071] (1) Level definition:

[0072] The specific meanings and boundaries of different levels (such as S, A, B levels, etc.) are clearly defined. For example, the S level indicates that the movie content quality is extremely high, with profound ideological connotations, excellent production techniques, and no bad content; the A level indicates that the movie content quality is relatively high, with a small amount of mild controversial content, but overall conforms to the mainstream values; the B level indicates that the movie content quality is average, with more defects or content that needs to be treated with caution.

[0073] (2) Sensitive content quantification standard:

[0074] Establish standards for how to accurately measure the degree of relevant content in the movie. For example, through image recognition and analysis of inappropriate scenes in the video, the types, frequencies, and degrees of inappropriate behaviors are counted and converted into quantitative indicators for rating calculation.

[0075] (3) Weighting or de - weighting mechanism:

[0076] Assign weights or de - weight the ratings according to the importance of different factors. For example, for user word - of - mouth, if the proportion of positive reviews is relatively high, a higher weight is given; if there are serious negative public opinions, the movie rating is de - weighted.

[0077] (4) Definition of rating dimensions:

[0078] Specific methods for rating from multiple dimensions such as production quality (including picture quality, sound quality, editing level, etc.), content theme (including depth and innovation, value orientation, etc.), and social impact (including user word - of - mouth, public opinion impact, etc.). For example, when evaluating picture quality, factors such as resolution, color restoration, and camera use are considered.

[0079] (5) Rating scoring criteria:

[0080] Methods for quantitatively scoring each rating dimension. For example, in the picture quality dimension, specific scoring rules are formulated according to factors such as the level of resolution and the accuracy of color restoration, so as to comprehensively calculate the score of the movie in this dimension.

[0081] (6) Rating confidence criteria:

[0082] Methods for evaluating the reliability and confidence of rating results. For example, by analyzing factors such as data integrity and algorithm stability, determine the confidence interval of the rating results. When the confidence level is low, it may indicate that further review or other auxiliary means are needed to confirm the rating.

[0083] Comprehensively use machine - learning classification algorithms (such as support vector machines, decision trees, etc.) and rule - based logical reasoning. On the one hand, train classification algorithms on a labeled movie sample dataset so that they can perform rating classification according to the input multi - modal data features. For example, use the collected rated movie data as training samples, extract the multi - modal data features of the movie (such as inappropriate content features, user word - of - mouth features, etc.) as input, and the actual rating of the movie as output, and train a support vector machine model. On the other hand, according to pre - set industry rules and ethical norms, make compliance judgments on movie content through logical reasoning. For example, when judging whether the inappropriate content in a movie exceeds the standard, both refer to the inappropriate degree score given by the machine - learning model based on data such as pictures and sounds, and make logical judgments based on industry regulations on the display of inappropriate scenes to ensure that the rating results comply with industry norms and ethical standards.

[0084] The content rating engine synthesizes all data and the film information that has been fused and complemented, and conducts content rating on the film according to the pre-set rating algorithm. For example, assuming that in the rating algorithm, it is stipulated that the proportion of inappropriate content in the film exceeding 10% will have a greater impact on the rating. The system analyzes the script and relevant video data, and calculates that the actual proportion of inappropriate content in the film is 8%. When calculating the proportion of inappropriate content, the system uses image recognition technology to identify and count inappropriate scenes (such as fighting, weapon use, etc.) in the film video sample, and combines the description of inappropriate plots in the script to comprehensively obtain the quantitative data of inappropriate content. At the same time, combining factors such as user word-of-mouth (70% positive comments) and the historical average rating of the same theme (Grade A), the final rating of the film is calculated according to a certain weight (such as the weight of the proportion of inappropriate content is 0.3, the weight of user word-of-mouth is 0.4, and the weight of historical data of the same theme is 0.3).

[0085] Furthermore, the final rating result of the film to be evaluated is output in an intuitive way, including the final rating conclusion of the film to be evaluated and the detailed rating reasons. For example, the output result may be "The film is rated Grade A, and the reason is: There are few inappropriate scenes in the film and they are handled properly, the user word-of-mouth is good, the historical data of the same theme shows that the average rating of this type of film is relatively high, and there is no relevant negative public opinion impact." For example, the specific output result may be "The film is rated Grade A. The rating reasons are as follows: The proportion of inappropriate content in the film is 8%, which does not exceed the specified threshold; the user word-of-mouth is good, and positive comments account for 70%; the historical data of the same theme shows that the average rating of this type of film is Grade A. Considering the above factors comprehensively, the film is given a Grade A rating." This output result not only provides the user with the final rating of the film, but also makes the rating result interpretable through detailed reason explanations, which is convenient for users to understand the basis and process of the film rating.

[0086] Furthermore, for some controversial films, the system provides an artificial review mechanism. Professional reviewers conduct a re-review of the film and adjust the rating result according to the actual situation. For example, when there is a large difference between the system rating result and the artificial experience judgment, professional reviewers can view the multi-modal data and analysis process relied on by the system, and combine their professional judgment to finally determine the rating of the film. The artificial review mechanism can further ensure the accuracy and reliability of the rating result. During the artificial review process, reviewers can use the film playback function provided by the system and combine the multi-modal data report to conduct a comprehensive evaluation of the film. At the same time, the system will record the process and results of the artificial review for optimizing and improving the automatic rating algorithm and improving the overall performance of the system.

[0087] Furthermore, as new film data accumulates continuously and the industry environment changes, various types of data in the system are updated regularly. For example, the latest film-related data, including the basic information and word-of-mouth data of newly released films, is obtained from the upstream new film warehousing system every week, and historical data of the same theme and online public opinion data are updated in a timely manner. This helps ensure that the data on which the system bases its ratings is up-to-date and comprehensive, thus guaranteeing that the rating results can reflect the current market and industry conditions.

[0088] Regarding users' perception of word-of-mouth, the latest user reviews are continuously collected through web crawlers, and the results of sentiment analysis and topic mining are updated. If there is a significant change in the word-of-mouth of a certain film during subsequent playback (such as the user evaluation changing from positive to negative due to a plot twist), the system can promptly capture this change and take it into account during re-evaluation.

[0089] The algorithm models in the intelligent prediction model and content rating engine are optimized based on the new data. For example, the deep learning model for rating intelligent prediction is retrained using the newly accumulated film script data and the corresponding actual rating results, and the parameters of the model are adjusted to improve the accuracy of the model's analysis of the script content.

[0090] For the machine learning classification algorithm in the content rating engine, such as the support vector machine, the classification boundary is continuously adjusted according to the new labeled samples to optimize the classification performance. At the same time, combined with the feedback and adjustment results of the reviewers during the manual review process, the reasons for the deviation of the system rating are analyzed, and the feature selection and weight assignment mechanisms of the model are improved targeted to further enhance the accuracy and reliability of the rating.

[0091] Furthermore, the final rating results of the films to be evaluated in the present invention can be applied to the following scenarios:

[0092] (1) Content procurement decision support:

[0093] Before a film is released, when considering content procurement, film production companies, distributors or video platforms can use the method of the present invention to pre-rate the film. For example, a video platform can input the film to be procured into the system to obtain its pre-rating results and a detailed analysis report. If the pre-rating of the film is relatively high and it performs well in aspects such as historical trends of the same theme and perception of the main cast and crew trends, the platform can be more confident in procurement, reducing procurement risks and improving content quality.

[0094] For content production companies, they can also refer to the pre-rating feedback of the present invention during the film production process to adjust and optimize the film content. For example, if the pre-rating indicates that inappropriate content in the film may lead to a lower rating, the production side can consider modifying relevant plots or treatment methods to improve the overall rating and market competitiveness of the film.

[0095] (2) Content traffic operation and settlement basis:

[0096] After the film is released, the video platform can manage the traffic operation of the film according to the rating results of the present invention. For example, for films with higher ratings, more recommended positions, homepage display opportunities, etc. can be given to increase the exposure and play volume of the film. For films with lower ratings, corresponding restrictive measures can be taken, such as restricting the recommended range or adding content warning labels, etc.

[0097] In terms of content settlement, the platform and the content provider can negotiate the settlement price and method according to the rating results of the film. Generally, films with higher ratings may obtain higher settlement prices because they have higher content quality and market value. The rating results of the present invention provide an objective and fair basis for such settlement, which helps to reduce disputes between the two parties during the settlement process.

[0098] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to user analysis data, user stored data, user displayed data, etc.) and signals involved in the present invention are all information, data and signals authorized by the user or fully authorized by all parties; and the collection, use and processing of relevant information, data and signals comply with the laws, regulations and standards of relevant countries and regions. Specifically include:

[0099] (1) Data security:

[0100] During the data collection, storage and processing processes, strict data security measures are taken. For data obtained from external sources (such as user comment data crawled from social media platforms), encryption protocols are used during the transmission process to prevent data from being stolen or tampered with. For example, SSL / TLS encryption technology is used to ensure the security of data during network transmission.

[0101] In terms of data storage, a secure database management system is adopted and strict access permissions are set. Only authorized personnel can access and operate the data in the system to prevent data leakage. For example, different database operation permissions are set for different levels of users (such as system administrators, data analysts, etc.) to ensure the confidentiality and integrity of the data.

[0102] (2) Privacy protection:

[0103] When processing user-related data (such as user comment data in user word-of-mouth perception), the privacy of users is fully respected. The collected user data is anonymized to ensure that the specific user identity cannot be identified during the data analysis process. For example, when performing sentiment analysis on user comments, only the comment content itself is concerned, rather than the personal information of the user, such as the username, IP address, etc.

[0104] When using user data for model training and optimization, relevant privacy protection regulations and industry norms are followed to ensure the legal use of user data. At the same time, the purpose and method of data use are clearly informed to users, and users' consent is obtained to protect users' right to know and right to choose.

[0105] The innovations of the present invention include:

[0106] (1) Deep fusion and intelligent perception method for multi-source heterogeneous data:

[0107] A set of efficient data fusion algorithms that can integrate multi-source heterogeneous data including film scripts, basic information, film personnel information, historical data of the same theme, public opinion across the network, user reviews, etc. For example, when fusing script data with film visual picture data, by establishing a semantic mapping model, the scene descriptions in the script are associated with the visual elements in the picture (such as scene layout, character actions, etc.), so as to comprehensively understand the plot presentation and expression techniques of the film, overcoming the one-sidedness of traditional single data type evaluation.

[0108] Innovative multi-modal data perception technology that can dynamically perceive data changes of the film in different dimensions in real time. For example, using a real-time public opinion monitoring system, it can not only timely capture negative news related to the film, but also analyze the heat trend, dissemination path and audience emotion tendency of public opinion, providing a basis for quickly adjusting the rating strategy and solving the problem of lagging response to external factors such as public opinion in the existing technology.

[0109] (2) Deep semantic analysis and pre-rating method for film scripts based on deep learning:

[0110] An advanced deep learning architecture (such as a bidirectional LSTM network combined with an attention mechanism) is used to perform deep semantic analysis on the film script. This architecture can focus on the key plots and emotional contexts in the script, accurately identify potential sensitive content and plot trends, for example, accurately judge the impact of metaphorical bad hints or complex emotional conflicts on the film rating, and greatly improve the accuracy and in-depth understanding ability of pre-rating compared with traditional script analysis methods based on keyword matching.

[0111] (3) Rating algorithm integrating dynamic weighting and logical reasoning:

[0112] A unique dynamic weighting algorithm is designed, which can automatically adjust the weights of various rating factors (such as sensitive content, user word-of-mouth, historical data of the same theme, etc.) according to factors such as the type, theme, target audience, and market environment of the film. For example, for films targeted at teenagers, the weights of user word-of-mouth and positive value orientation factors will be increased; while for art films, more emphasis will be placed on the consideration of content innovation and depth, so as to ensure that the rating results can accurately reflect the value and suitability of the film in a specific context, overcoming the rigidity of the traditional fixed weighting method.

[0113] Deeply integrate rule-based logical reasoning with machine learning classification algorithms. In the rating process, both use machine learning models to learn and judge a large number of data features, and conduct logical reasoning based on industry norms, moral standards, and specific review rules. For example, when judging whether inappropriate content is compliant, in addition to referring to the quantitative analysis of inappropriate scenes by the model, it will also comprehensively judge the presentation method, degree of inappropriateness, and its rationality in the overall context of the film based on relevant laws, regulations, and social ethics, effectively avoiding rating biases caused by pure data-driven, and improving the accuracy and fairness of ratings.

[0114] (4) Comprehensive and detailed rating dimension and standard system method:

[0115] Construct a rating dimension system covering various aspects of the film, including production quality (evaluated from multiple levels such as picture resolution, color calibration, surround sound effect, and editing rhythm), content theme (deeply analyzing the novelty, ideological nature of the theme, and the positive or negative transmission of values), and social impact (comprehensively considering multi-dimensional analysis of user word-of-mouth, positive and negative effects of public opinion, and market follow-up or leading trends of films of the same theme). Detailed and quantifiable evaluation indicators and methods are formulated for each dimension. For example, when evaluating the innovation of the content theme, by comparing and analyzing with the theme database of the same type of films, the uniqueness score of the theme is calculated, enabling in-depth and detailed rating analysis, and solving the problem of single and general existing rating standards.

[0116] Establish a flexible and dynamically adjustable quantitative standard for sensitive content. With the help of advanced image recognition technology, semantic analysis technology, and multi-modal data correlation analysis, it can accurately quantify inappropriate and other sensitive content in the film. And it can be dynamically adjusted according to changes in different cultural backgrounds, social values, and updates of industry norms. For example, as society's tolerance for certain types of inappropriateness (such as cartoonized inappropriate behaviors) changes, the system can timely update the corresponding quantitative parameters to ensure that the ratings always meet current social cognition and industry requirements.

[0117] The beneficial effects of the present invention:

[0118] (1) Efficiency improvement: Through an automated processing flow, the present invention can rate a large number of films in a short period of time. For example, in actual tests, the system can process the rating tasks of hundreds of films per hour. Compared with manual rating, the processing speed has been greatly improved, which can meet the timeliness requirements of content rating in the rapid development of the film and television industry, and ensure that films can be launched in a timely manner or carry out relevant operation operations.

[0119] (2) Standard unification: Based on strictly defined rating standards, the entire rating process is carried out according to unified algorithms and rules, ensuring the consistency and comparability of rating results among different films. Regardless of the type, style, and source of the film, it is evaluated according to the same sensitive content quantification standard, weighting mechanism, and rating dimension, avoiding rating differences caused by human factors or different methods, and improving the reliability and fairness of the rating results.

[0120] (3) Enhanced objectivity: With the help of multi-modal data and automated algorithms, the rating process is not affected by personal subjective factors. The system's rating of films is completely based on the calculation results of data and algorithms, and will not change due to factors such as the personal preferences, cultural backgrounds, or emotional states of reviewers, ensuring the objectivity and fairness of the rating results and being able to truly reflect the actual content quality and suitable viewing population of the films.

[0121] (4) Improved comprehensiveness and accuracy: By comprehensively rating with multi-modal data, it can comprehensively consider various factors of the film from content plot to market performance, public opinion impact, etc., thereby improving the accuracy of the rating. For example, when rating a controversial film, the system not only considers the content of the film itself (such as plot, picture, sound effects, etc.), but also combines external factors such as user reviews, historical data of the same theme, and public opinion risks, and can more accurately judge the actual impact of the film and the suitable viewing population, providing a more reliable basis for content procurement, traffic operation, and settlement of the film.

[0122] In addition, the possible design change directions or deformation schemes of the present invention include:

[0123] (1) Alternative solutions for data fusion and perception:

[0124] 1) Adopt a federated learning framework for data fusion and model training:

[0125] Federated learning allows different data owners (such as major film and television platforms, social media platforms, etc.) to jointly participate in model training without sharing the original data. In the present invention, a federated learning framework can be utilized to enable each platform to upload the model parameters trained locally to a central server. The central server performs model aggregation and optimization, and then transmits the updated model parameters back to each platform. In this way, data privacy can be protected, and the integration and utilization of multi-source data can be achieved to rate the film content. For example, video platform A has the playback data of a film, and social media platform B has user comment data. Through federated learning, the two parties can rate the film by comprehensively considering the playback data and comment data without directly sharing the data.

[0126] 2), Multi-modal data association and perception based on knowledge graph:

[0127] Construct a knowledge graph in the film and television field to model entities such as films, actors, directors, themes, plot elements, etc. and the relationships between them. In the data perception stage, map the collected multi-modal data (such as scripts, public opinions, user word-of-mouth, etc.) into the knowledge graph, and through the reasoning and analysis capabilities of the knowledge graph, explore the potential associations and influences between the data. For example, when the director of a film has a frequent cooperation relationship with successful works of a specific theme, the knowledge graph can infer that the film may have a high quality expectation in this theme, thus providing a reference for rating. This method can more intuitively display the logical relationships between the data, improve the data utilization efficiency and the accuracy of rating.

[0128] (2), Variant schemes of intelligent prediction and rating algorithms:

[0129] 1), Application of reinforcement learning algorithm to rating strategy optimization:

[0130] Introduce a reinforcement learning algorithm and regard the rating process as a decision-making process of an intelligent agent in a specific environment (constituted by film data and rating criteria). The intelligent agent (rating system) continuously interacts with the environment, takes different rating actions according to different film characteristics, and learns and optimizes the rating strategy according to the obtained rewards (such as accuracy feedback of rating, market feedback, etc.). For example, when the system gives a high rating to a film and the film has a good reputation and high box office in the market, the intelligent agent gets a positive reward, thus strengthening the current rating strategy; conversely, if the rating does not match the market reaction, a negative reward is obtained, prompting the intelligent agent to adjust the rating strategy. In this way, the system can continuously adapt to market changes and film diversity, improving the accuracy and adaptability of rating.

[0131] 2), Sample expansion and rating model training based on generative adversarial network (GAN):

[0132] Use the generator in the GAN network to generate more virtual movie sample data. While maintaining a similar feature distribution to real movie data, these sample data can expand the training dataset. For example, the generator can generate virtual script, scene description and other data based on existing movie types, plot patterns, actor styles and other information. Then, mix these generated data with real data for training the rating model. The discriminator is used to distinguish between real and generated data. In this adversarial process, the rating model can learn a wider and deeper movie feature pattern, improving the generalization ability of rating different types of movies. Especially for some niche or emerging types of movies, it can better perform rating predictions.

[0133] (3) System architecture adjustment plan:

[0134] 1) Distributed architecture design:

[0135] Design the entire rating system as a distributed architecture, with each part (such as data collection, data processing, rating, etc.) distributed on different computing nodes. This can improve the system's processing capacity and scalability to meet the rating needs of large-scale movie data. For example, data collection can be distributed on multiple web crawler nodes to collect data from different data sources (such as major film and television websites, social media platforms, etc.) simultaneously, and then transfer the data to the data processing nodes for cleaning, fusion and other operations. Finally, the rating nodes perform rating calculations based on the processed data. This distributed architecture can effectively improve the system's concurrent processing capacity, reduce data processing time, and improve rating efficiency.

[0136] 2) Module splitting and reorganization based on microservice architecture:

[0137] Adopt a microservice architecture to split the existing system modules, and develop each functional part (such as user input, multimodal data perception, etc.) into an independent microservice. These microservices can be flexibly combined and deployed according to requirements, facilitating system maintenance and upgrade. For example, if functional optimization or technical upgrade is needed for user word-of-mouth perception, only the corresponding microservice needs to be modified without affecting the normal operation of other modules. At the same time, the microservice architecture is also convenient for integration with other external systems, such as integration with the internal management system of film production companies to achieve sharing and interaction of movie rating data, providing more comprehensive support for all links in the film production process.

[0138] (4) Data source expansion plan:

[0139] 1) Introduce structured data of professional film critics:

[0140] In addition to the comment data of ordinary users, cooperate with professional film critic platforms to obtain structured evaluation data of professional film critics. These data usually include more in-depth film analysis, professional art evaluations, and insights from an industry perspective. For example, film critics may conduct detailed evaluations on aspects such as the director's techniques, cinematography skills, and narrative structure of a film, and give corresponding scores and comments. Incorporating these data into the data sources of the rating system can improve the professionalism and depth of the rating. Especially for some films with high artistic value or complex connotations, more accurate ratings can be obtained.

[0141] 2) Mine internal data resources in the film and television industry:

[0142] Deeply mine internal data resources in the film and television industry, such as shooting logs, post-production data, and distribution channel data of film and television production companies. For example, information such as the shooting cycle, shooting location, and shooting difficulty in the shooting log can reflect the investment and attentiveness of film production; the special effects production time and the number of editing versions in the post-production data can reflect the fineness of the film in production technology; the distribution channel data can reflect the market positioning and expected audience of the film. Combining these internal data with external data (such as public opinion, user word-of-mouth, etc.) can evaluate the quality and value of the film from more dimensions and provide a more comprehensive basis for rating.

[0143] Furthermore, as Figure 3 shown, based on the above-mentioned automated film content rating method based on multi-modal data, the present invention also correspondingly provides an automated film content rating system based on multi-modal data prediction. Among them, the automated film content rating system based on multi-modal data prediction includes:

[0144] An information input and completion module 51, configured to receive film-related information of the film to be evaluated input by a user, connect to an upstream new film warehousing system, obtain multi-dimensional data of the film to be evaluated, and integrate the multi-dimensional data with the film-related information through a data fusion and completion algorithm to obtain a film data set;

[0145] A multi-modal data perception module 52, configured to analyze the film data of the film to be evaluated from different preset angles according to the film data set to obtain multiple film analysis results;

[0146] A rating intelligent prediction module 53, configured to perform a preliminary rating prediction on the film to be evaluated based on the film-related information by using natural language processing technology and machine learning algorithms, and generate a preliminary rating prediction result;

[0147] The content rating engine module 54 is used to construct a multi-factor weighted model, and perform content rating on the to-be-evaluated film based on the film data set, the plurality of film analysis results and the preliminary rating prediction results, and based on the rating criteria to obtain a final rating result.

[0148] Furthermore, the automatic film content rating system based on multimodal data prediction of the present invention also includes:

[0149] The evaluation result display module is used to output the final rating result in an intuitive manner, including the final rating conclusion and detailed rating reasons of the film to be evaluated.

[0150] The manual review module is used to provide a manual review mechanism when there is a dispute over the final rating result. Professional reviewers will review the film to be rated again and adjust the final rating result according to the actual situation. At the same time, the process and results of the manual review are recorded to optimize and improve the automatic rating algorithm.

[0151] The data update and model optimization module is used to regularly update various types of data in the system, optimize the multi-factor weighted model according to the new data, analyze the reasons for the deviation of the system rating in combination with the feedback and adjustment results of the reviewers during the manual review process, and improve the feature selection and weight allocation mechanism of the multi-factor weighted model in a targeted manner.

[0152] Furthermore, if Figure 4 As shown, based on the above-mentioned automatic film content rating method and system based on multi-modal data prediction, the present invention also provides a terminal accordingly, and the terminal includes a processor 10, a memory 20 and a display 30. Figure 4 Only some components of the terminal are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0153] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as the hard disk or memory of the terminal. In some other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the terminal. Further, the memory 20 may also include both the internal storage unit and the external storage device of the terminal. The memory 20 is used to store application software installed on the terminal and various types of data, such as program codes for installing the terminal. The memory 20 may also be used to temporarily store data that has been output or will be output. In one embodiment, an automated movie content rating program 40 based on multi-modal data prediction is stored on the memory 20, and the automated movie content rating program 40 based on multi-modal data prediction can be executed by the processor 10, so as to implement the automated movie content rating method based on multi-modal data prediction in the present application.

[0154] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor or other data processing chips, and is used to run the program codes stored in the memory 20 or process data, such as executing the automated movie content rating method based on multi-modal data prediction, etc.

[0155] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. The display 30 is used to display information on the terminal and to display a visual user interface. The processor 10, the memory 20, and the display 30 of the terminal communicate with each other through a system bus.

[0156] In one embodiment, when the processor 10 executes the automated movie content rating program 40 based on multi-modal data prediction in the memory 20, the steps of the automated movie content rating method based on multi-modal data prediction as described above are implemented.

[0157] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores an automated movie content rating program based on multi-modal data prediction, and when the automated movie content rating program based on multi-modal data prediction is executed by a processor, the steps of the automated movie content rating method based on multi-modal data prediction as described above are implemented.

[0158] In summary, the present invention provides an automated film content rating method, system, terminal and computer-readable storage medium based on multi-mode data prediction. The method includes: receiving film-related information of a film to be evaluated input by a user, connecting to an upstream new film warehousing system to obtain multi-dimensional data of the film to be evaluated, and integrating the multi-dimensional data with the film-related information through a data fusion and completion algorithm to obtain a film data set; analyzing the film data of the film to be evaluated from different preset perspectives according to the film data set to obtain multiple film analysis results; based on the film-related information, using natural language processing technology and machine learning algorithms to perform a preliminary rating prediction on the film to be evaluated to generate a preliminary rating prediction result; constructing a multi-factor weighted model, and performing content rating on the film to be evaluated according to the film data set, multiple film analysis results and the preliminary rating prediction result, and based on a rating standard to obtain a final rating result. The present invention considers the information of the film from multiple perspectives, rates the film content comprehensively, accurately, efficiently and objectively, can rate a large number of films in a short time, and improves the accuracy and reliability of the rating.

[0159] It should be noted that in this article, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or terminal. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or terminal including that element.

[0160] Of course, those of ordinary skill in the art can understand that all or part of the processes of implementing the above method embodiments can be completed by instructing relevant hardware (such as a processor, a controller, etc.) through a computer program. The program can be stored in a computer-readable storage medium readable by a computer. When the program is executed, it can include the processes of the above method embodiments. The computer-readable storage medium can be a memory, a magnetic disk, an optical disc, etc.

[0161] It should be understood that the application of the present invention is not limited to the above examples. For those of ordinary skill in the art, improvements or transformations can be made according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of the present invention.

Claims

1. An automated film content rating method based on multimodal data prediction, characterized in that: The automatic film content rating method based on multimodal data prediction includes: Receive the film-related information of the film to be evaluated input by the user, connect to the upstream new film storage system, obtain the multi-dimensional data of the film to be evaluated, and integrate the multi-dimensional data with the film-related information through data fusion and completion algorithms to obtain a film data set; According to the film data set, analyzing the film data of the film to be evaluated at different preset angles to obtain multiple film analysis results; Based on the film-related information, using natural language processing technology and machine learning algorithms to make a preliminary rating prediction for the film to be evaluated, and generate a preliminary rating prediction result; A multi-factor weighted model is constructed, and the content of the film to be evaluated is rated based on the film data set, the multiple film analysis results and the preliminary rating prediction results, and based on the rating criteria to obtain a final rating result.

2. The method for automatic film content rating based on multimodal data prediction according to claim 1, characterized in that: The multi-dimensional data includes: basic information of the film, basic information of the filmmakers, peripheral information of the film, historical data of the same theme, historical data of the main staff and public opinion data of the entire network.

3. The automatic film content rating method based on multimodal data prediction according to claim 1, characterized in that: The step of analyzing the video data of the video to be evaluated at different preset angles according to the video data set to obtain multiple video analysis results specifically includes: Collecting user comment data on the film to be evaluated by using web crawler technology, analyzing the comment data by using natural language processing technology, and obtaining comment analysis results; Setting keyword filtering and event monitoring algorithms to monitor negative public opinion events related to the film to be evaluated in real time; Conduct data mining and analysis on the market performance and ratings of films of the same type as the film to be evaluated in historical periods, and calculate statistical indicators; Analyze the popular trends and hot topics of current films of the same type as the film to be evaluated, and obtain the impact on the film to be evaluated based on the popular trends and hot topics of current films of the same type; Through industry reports and market data, the average commercial revenue and profit model of films with the same theme are calculated to evaluate the potential of the film to be evaluated in terms of commercial value and rating impact; Track the recent creative style and public opinion image of specific creators of the film to be evaluated, and evaluate the potential influence of specific creators on the film to be evaluated.

4. The method for automatic film content rating based on multimodal data prediction according to claim 1, characterized in that: The rating standards include: level definition, sensitive content quantification standards, weighting or downgrading mechanism, rating dimension definition, rating scoring standards and rating confidence standards.

5. The method for automatic film content rating based on multimodal data prediction according to claim 1, characterized in that: The multi-factor weighted model is constructed to rate the content of the film to be evaluated based on the plurality of film analysis results and the preliminary rating prediction results, and the final rating result is obtained, and then further includes: The final rating result is output in an intuitive manner, including the final rating conclusion and detailed rating reasons of the film to be evaluated.

6. The method for automatic film content rating based on multimodal data prediction according to claim 1, characterized in that: The multi-factor weighted model is constructed to rate the content of the film to be evaluated based on the plurality of film analysis results and the preliminary rating prediction results, and the final rating result is obtained, and then further includes: When there is a dispute over the final rating result, a manual review mechanism is provided, and professional reviewers will review the film to be rated again, and adjust the final rating result according to the actual situation. At the same time, the process and results of the manual review are recorded for optimizing and improving the automatic rating algorithm.

7. The method for automatic film content rating based on multimodal data prediction according to claim 6, characterized in that: The automatic film content rating method based on multimodal data prediction also includes: Regularly update various data in the system and optimize the multi-factor weighted model according to the new data; Combined with the reviewers' feedback and adjustment results during the manual review process, the reasons for the deviation in the system rating are analyzed, and the feature selection and weight allocation mechanism of the multi-factor weighted model are improved in a targeted manner.

8. An automated film content rating system based on multimodal data prediction, characterized in that: The automatic film content rating system based on multimodal data prediction includes: An information input and completion module is used to receive the film-related information of the film to be evaluated input by the user, connect to the upstream new film storage system, obtain the multi-dimensional data of the film to be evaluated, and integrate the multi-dimensional data with the film-related information through data fusion and completion algorithms to obtain a film data set; A multi-mode data perception module, used to analyze the film data of the film to be evaluated at different preset angles according to the film data set to obtain multiple film analysis results; An intelligent rating prediction module is used to make a preliminary rating prediction for the film to be evaluated based on the relevant information of the film using natural language processing technology and machine learning algorithms to generate a preliminary rating prediction result; The content rating engine module is used to construct a multi-factor weighted model, and to perform content rating on the film to be evaluated based on the film data set, the multiple film analysis results and the preliminary rating prediction results, and to obtain a final rating result based on the rating criteria.

9. A terminal, characterized in that: The terminal includes: a memory, a processor, and an automatic film content rating program based on multimodal data prediction stored in the memory and executable on the processor, wherein the automatic film content rating program based on multimodal data prediction, when executed by the processor, implements the steps of the automatic film content rating method based on multimodal data prediction as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an automated film content rating program based on multimodal data prediction, and when the automated film content rating program based on multimodal data prediction is executed by a processor, the automated film content rating program based on multimodal data prediction implements the steps of the automated film content rating method based on multimodal data prediction as described in any one of claims 1-7.

Images