Film and television simulation evaluation method and system based on large language model, terminal and storage medium

Abstract

The invention discloses a movie and television simulation evaluation method and system based on a large language model, a terminal and a storage medium, and the method comprises the steps: obtaining user film watching data and corresponding movie and television content data, carrying out the preprocessing of the user film watching data and the movie and television content data, and obtaining target user film watching data and movie and television key features; determining a pre-training model, and performing model training and model fine tuning on the pre-training model according to the target user film watching data and the film and television key features to obtain a film and television simulation evaluation model; and obtaining current film and television content data, inputting the current film and television content data into the film and television simulation evaluation model, and outputting a film and television simulation evaluation result. The movie and television simulation evaluation model is constructed to carry out movie and television simulation evaluation prediction on the current movie and television content data, potential evaluation of different user groups on new movie and television content can be effectively reflected, and the movie and television simulation evaluation result output accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a film and television simulation evaluation method, system, terminal and computer-readable storage medium based on a large language model. Background Art

[0002] With rising film and television production costs and increasingly fierce market competition, film and television content investors and distributors face increasing investment risks. Accurate film and television review predictions have become a crucial tool in the industry, not only helping investors make more rational investment decisions but also guiding audiences to choose films that truly meet their preferences and are worth watching.

[0003] However, existing film and television rating prediction methods have obvious limitations.

[0004] For example, data mining-based prediction methods use regression models to predict reputation and box office by analyzing factors such as a film's production cost, cast, and director's reputation. However, these methods rely too heavily on the correlation of historical data and struggle to accurately capture the artistic value of the film's content and the audience's emotional experience.

[0005] Another example is the preview screening system, which collects reviews from a limited number of audience members or staff before a film's official release. This method is limited by the small sample size and lack of representativeness, making it difficult to fully reflect the viewing reactions of different groups of people.

[0006] As can be seen, existing film and television evaluation technology has significant shortcomings in terms of representativeness of evaluation samples, scientific nature of the evaluation process, and personalized evaluation results. This is particularly true during the predictive evaluation phase before a film's official release, where accuracy is severely challenged. It is impossible to accurately simulate the differentiated evaluations of different user groups on the same film, nor is it possible to effectively integrate the deep connection between film content features and user viewing preferences. As a result, the generated evaluation predictions fail to meet the actual needs of investors, distributors, and audiences.

[0007] Therefore, there is an urgent need to develop a technical solution that can comprehensively analyze the content characteristics of the film and make intelligent evaluation predictions based on the user's personalized viewing habits. Summary of the Invention

[0008] The main purpose of the present invention is to provide a film and television simulation evaluation method, system, terminal and computer-readable storage medium based on a large language model, aiming to solve the problems existing in the prior art such as insufficient representativeness of evaluation samples, difficulty in capturing the deep connection between film content and user preferences, lack of personalized analysis, etc., especially in the predictive evaluation stage before the official release of the film, it is difficult to accurately simulate the real reaction of a diverse audience group to the film.

[0009] To achieve the above-mentioned object, the present invention provides a film and television simulation evaluation method based on a large language model, the film and television simulation evaluation method based on a large language model comprising the following steps:

[0010] Obtaining user viewing data and corresponding film and television content data, and preprocessing the user viewing data and the film and television content data to obtain target user viewing data and film and television key features;

[0011] Determine a pre-trained model, and perform model training and model fine-tuning on the pre-trained model based on the target user's viewing data and the key features of the film and television to obtain a film and television simulation evaluation model;

[0012] The current film and television content data is obtained, and the current film and television content data is input into the film and television simulation evaluation model, and the film and television simulation evaluation result is output.

[0013] Optionally, in the film and television simulation evaluation method based on a large language model, the preprocessing includes a first preprocessing and a feature extraction process;

[0014] The step of obtaining user viewing data and corresponding film and television content data, and preprocessing the user viewing data and the film and television content data to obtain target user viewing data and film and television key features, specifically includes:

[0015] Acquiring user movie viewing data using a web crawler technology or a data interface calling technology, and performing the first preprocessing on the user movie viewing data to obtain target user movie viewing data, wherein the first preprocessing includes data cleaning, denoising, and normalization.

[0016] The film and television content data corresponding to the target user's viewing data is obtained, and the feature extraction process is performed on the film and television content data to obtain film and television key features.

[0017] Optionally, the film and television simulation evaluation method based on a large language model, wherein the step of obtaining the film and television content data corresponding to the target user's viewing data and performing the feature extraction process on the film and television content data to obtain key features of the film and television, specifically includes:

[0018] Obtaining film information from the target user's viewing data, and obtaining corresponding film and television content data based on the film information;

[0019] The feature extraction process is performed on the film and television content data to obtain key film and television features, wherein the feature extraction process includes text content analysis process, audio-visual content analysis process, structured data analysis process and narrative structure analysis process.

[0020] Optionally, the film and television simulation evaluation method based on a large language model, wherein the determining of a pre-trained model and the model training and model fine-tuning of the pre-trained model according to the target user viewing data and the key features of the film and television to obtain the film and television simulation evaluation model, specifically includes:

[0021] Associating the target user's movie viewing data with the key features of the film and television to obtain a training sample data set;

[0022] Performing data segmentation processing on the training sample data set to obtain a target sample data set, wherein the target sample data set includes a training set, a validation set, and a test set;

[0023] A pre-trained model is determined, and model training processing, model fine-tuning processing and model optimization processing are performed on the pre-trained model according to the target sample data set to obtain a film and television simulation evaluation model.

[0024] Optionally, the film and television simulation evaluation method based on a large language model, wherein the pre-trained model is subjected to model training, model fine-tuning, and model optimization according to the target sample data set to obtain a film and television simulation evaluation model, specifically includes:

[0025] Performing model training processing on the pre-trained model according to the training set to obtain an initial film and television simulation evaluation model;

[0026] Performing model fine-tuning processing on the initial film and television simulation evaluation model according to the verification set to obtain a fine-tuned film and television simulation evaluation model;

[0027] The fine-tuned film and television simulation evaluation model is optimized according to the test set to obtain a film and television simulation evaluation model.

[0028] Optionally, the film and television simulation evaluation method based on a large language model, wherein the fine-tuning of the initial film and television simulation evaluation model according to the validation set to obtain the fine-tuned film and television simulation evaluation model, specifically includes:

[0029] Inputting the verification set into the initial film and television simulation evaluation model, and outputting the initial film and television simulation evaluation result;

[0030] Obtaining actual film and television evaluation results, and comparing the initial film and television simulation evaluation results with the actual film and television evaluation results to obtain a comparison result;

[0031] The model parameters of the initial film and television simulation evaluation model are adjusted according to the comparison result to obtain a fine-tuned film and television simulation evaluation model.

[0032] Optionally, the film and television simulation evaluation method based on a large language model, wherein the step of obtaining current film and television content data, inputting the current film and television content data into the film and television simulation evaluation model, and outputting the film and television simulation evaluation result, specifically includes:

[0033] Acquire current film and television content data, extract target film and television key features corresponding to the current film and television content data, and input the target film and television key features into the film and television simulation evaluation model;

[0034] Generate multiple simulated user agents through the film and television simulation evaluation model, and perform simulated film and television evaluation generation processing based on the current film and television content data by the multiple simulated user agents to obtain simulated film and television evaluations of the multiple simulated user agents;

[0035] A plurality of the simulated film and television evaluations are aggregated to obtain a film and television simulation evaluation result.

[0036] In addition, to achieve the above-mentioned purpose, the present invention further provides a film and television simulation evaluation system based on a large language model, wherein the film and television simulation evaluation system based on a large language model includes:

[0037] A data preprocessing module is used to obtain user viewing data and corresponding film and television content data, and preprocess the user viewing data and the film and television content data to obtain target user viewing data and film and television key features;

[0038] A film and television simulation evaluation model generation module is used to determine a pre-trained model, and perform model training and model fine-tuning on the pre-trained model based on the target user viewing data and the key features of the film and television to obtain a film and television simulation evaluation model;

[0039] The film and television simulation evaluation result output module is used to obtain current film and television content data, input the current film and television content data into the film and television simulation evaluation model, and output the film and television simulation evaluation result.

[0040] In the present invention, user viewing data and corresponding film and television content data are obtained, and the user viewing data and the film and television content data are preprocessed to obtain target user viewing data and film and television key features; a pre-trained model is determined, and model training and model fine-tuning are performed on the pre-trained model according to the target user viewing data and the film and television key features to obtain a film and television simulation evaluation model; current film and television content data is obtained, and the current film and television content data is input into the film and television simulation evaluation model, and a film and television simulation evaluation result is output. The present invention trains the pre-trained model by obtaining user viewing data and film and television content data, thereby constructing a film and television simulation evaluation model. The film and television simulation evaluation model can achieve accurate analysis of the current film and television content data, thereby generating accurate film and television simulation evaluation results, which can reflect the potential evaluation of new film and television content by different user groups, provide a reliable basis for the early evaluation of film and television projects, and help film and television distributors formulate accurate promotion strategies based on the output film and television simulation evaluation results, thereby improving the promotion effect and resource utilization efficiency of new film and television works. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of a preferred embodiment of the film and television simulation evaluation method based on a large language model of the present invention;

[0042] Figure 2 This is a schematic diagram of the overall process of a preferred embodiment of the film and television simulation evaluation method based on a large language model of the present invention;

[0043] Figure 3 1 is a structural diagram of a preferred embodiment of a film and television simulation evaluation system based on a large language model according to the present invention;

[0044] Figure 4 It is a structural diagram of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the present invention more clear and distinct, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0046] Existing film and television rating technologies suffer from significant deficiencies in the representativeness of evaluation samples, the scientific nature of the evaluation process, and the personalization of evaluation results. This is particularly true during the predictive evaluation phase before a film's official release, where accuracy is severely challenged. These technologies are unable to accurately simulate the differentiated evaluations of different user groups regarding the same film, nor can they effectively integrate the deep connections between film content features and user viewing preferences. Consequently, the generated evaluation predictions fail to meet the actual needs of investors, distributors, and audiences.

[0047] To solve the above problems, the present invention proposes a film and television simulation evaluation method based on a large language model. By collecting user viewing data, including age, gender, region, viewing history and evaluation preferences, as well as corresponding film and television content data, as training data for the large language model, the large model learns the evaluation of films and television by different user groups, thereby generating virtual evaluations of new film and television content that conform to different user portraits, which can be used to evaluate the potential of film and television projects and formulate promotion plans.

[0048] The film and television simulation evaluation method based on the large language model described in the preferred embodiment of the present invention is as follows: Figure 1 As shown, the film and television simulation evaluation method based on the large language model includes the following steps:

[0049] Step S10: Obtain user viewing data and corresponding film and television content data, and pre-process the user viewing data and the film and television content data to obtain target user viewing data and film and television key features. The pre-processing includes a first pre-processing and a feature extraction process.

[0050] like Figure 2 As shown, the present invention first needs to collect user viewing data from multiple data sources, where the user viewing data covers the user's basic information (including age, gender, and region, etc.), viewing history (including movies watched, viewing time, and viewing platform, etc.), and ratings and evaluations; then, it is necessary to collect film and television content data corresponding to the user viewing data, and then analyze the film and television content data to extract key features in the film and television content data, such as plot keywords, actors, lines, picture style, and sound effects. After preprocessing the above data, it can be used as training data for the large language model.

[0051] Specifically, web crawler technology or data interface calling technology is used to obtain user movie viewing data, and the first preprocessing is performed on the user movie viewing data to obtain target user movie viewing data, wherein the first preprocessing includes data cleaning processing, denoising processing and normalization processing.

[0052] The present invention is provided with a data collection module: by cooperating with various film and television platforms, social media platforms and professional film and television data research institutions, it uses web crawler technology, data interface calling technology and other methods to collect user viewing data and store it in the user's original database; at the same time, it also collects corresponding film and television content data and stores it in the film and television original database.

[0053] like Figure 2 As shown, a user data preprocessing module is provided in the present invention: it is used to perform operations such as cleaning, denoising, and normalization on the user's original data (i.e., the user's movie viewing data in the present invention), extract key features in the user's movie viewing data, and convert it into structured data, which is stored in the user preprocessing database.

[0054] Obtain film information from the target user's viewing data, and obtain corresponding film and television content data based on the film information; perform feature extraction processing on the film and television content data to obtain film and television key features, wherein the feature extraction processing includes text content analysis processing, audio-visual content analysis processing, structured data analysis processing, and narrative structure analysis processing.

[0055] The user viewing data collected by the present invention includes information about the movies watched by the user, and the remaining data (ie, the movie content data in the present invention) can be collected through the movie information (or movie titles).

[0056] Among them, the specific processing process of text content analysis is as follows: natural language processing of the scripts, lines, subtitles and other text contents of film and television works, including word segmentation, word frequency statistics, sentiment analysis and topic extraction.

[0057] The specific processing process of audio-visual content analysis is as follows: computer vision and audio analysis of the video and audio content of film and television works, including scene recognition, shot switching frequency, color distribution, and audio feature extraction.

[0058] The specific processing process of structured data analysis is as follows: analyze the metadata of film and television works, including director, actors, release time, and length.

[0059] The specific process of narrative structure analysis is as follows: Identify the narrative structure of film and television works, including plot development, climax turning points, and ending type.

[0060] like Figure 2 As shown, it can be understood that the present invention provides a film and television feature extraction module: it is used to analyze film and television content data, extract its key features, such as plot keywords, cast characteristics, visual style, sound effects, etc., and store them in the film and television feature database. Among them, the key feature extraction methods include: 1. Using algorithms such as TF-IDF and TextRank to extract keywords from scripts and lines; 2. Applying topic models (such as LDA) to identify potential topics; 3. Using named entity recognition technology to extract entities such as people, places, and events; 4. Using large models for parsing and extraction.

[0061] Step S20: determine a pre-trained model, and perform model training and model fine-tuning on the pre-trained model according to the target user's viewing data and the key features of the film and television to obtain a film and television simulation evaluation model.

[0062] The pre-trained model in this invention preferably uses an open-source large-scale language model based on the Transformer architecture, such as the Qwen series, Llama-3, and Mistral. This invention constructs an association model between different user groups and film and television content features (i.e., the film and television simulation evaluation model in this invention), and then uses this model to effectively learn how different user groups evaluate elements in film and television content.

[0063] Specifically, the target user's movie viewing data is associated with the key features of the film and television to obtain a training sample data set; the training sample data set is segmented to obtain a target sample data set, wherein the target sample data set includes a training set, a validation set and a test set.

[0064] Determine a pre-trained model, and perform model training processing on the pre-trained model according to the training set to obtain an initial film and television simulation evaluation model; input the verification set into the initial film and television simulation evaluation model, and output an initial film and television simulation evaluation result; obtain the actual film and television evaluation result, and compare the initial film and television simulation evaluation result with the actual film and television evaluation result to obtain a comparison result; adjust the model parameters of the initial film and television simulation evaluation model according to the comparison result to obtain a fine-tuned film and television simulation evaluation model; perform model optimization processing on the fine-tuned film and television simulation evaluation model according to the test set to obtain a film and television simulation evaluation model.

[0065] like Figure 2 As shown, a large model training module is set up in the present invention: large model training is performed using user preprocessing data (i.e., target user viewing data in the present invention) and film and television feature data (i.e., key film and television features in the present invention), and the trained model is stored in a model library.

[0066] Specifically, the training process of the large model is as follows: 1. Data preparation: Associating user pre-processed data (i.e., the target user viewing data in the present invention, including user portraits, viewing history, and evaluation preferences) with film and television feature data (i.e., the key features of film and television in the present invention, including film and television content features); 2. Constructing a training sample format: The input part includes user features and film and television features, and the output part is the user's actual evaluation of the film and television; 3. Data segmentation: Divide the data set into a training set (ratio: 70%), a validation set (ratio: 15%), and a test set (ratio: 15%). 4. Selection of pre-training model: The Qwen-72B model is preferably used in the present invention. 5. Model Fine-tuning: Design a specific prompt template for the film and television review task, including: a. User information: [User age], [User gender], [User region], [User occupation]; b. Movie viewing history: [Film 1], [Film 2], ...; c. Historical review style: [Review example 1], [Review example 2], ...; d. Target film features: [Film title], [Genre], [Director], [Actor], [Plot summary], [Special elements]; e. Generate the user's review of the target film. This includes fine-tuning using PEFT (Parameter-Efficient Fine-Tuning) technology, employing supervised fine-tuning (SFT) methods to compare model output with actual user reviews and adjust model parameters through backpropagation. 6. Model Evaluation and Optimization: a. Evaluate model performance on the test set; b. Conduct manual evaluation to compare the similarity between model-generated reviews and real user reviews; c. Adjust the model architecture and training strategy based on the evaluation results.

[0067] In addition to pre-trained models, other machine learning algorithms or models can also be used to replace the current large language model (such as user-based collaborative filtering to predict user ratings of movies and TV shows, specifically using XGBoost to predict user ratings of movies), or combine multiple different types of models to build a hybrid model system to give full play to the advantages of different models and achieve more accurate and intelligent film and television evaluation simulation.

[0068] Step S30: Acquire current film and television content data, input the current film and television content data into the film and television simulation evaluation model, and output a film and television simulation evaluation result.

[0069] like Figure 2As shown, when there is new film and television content (i.e., the current film and television content data in the present invention) that needs to be simulated and evaluated, the new film and television content is analyzed, and its key features (i.e., the target film and television key features in the present invention) are extracted. These feature data are input into the big model (i.e., the film and television simulation evaluation model in the present invention). The big model generates corresponding intelligent agents for different user portraits to simulate users (i.e., multiple simulated user intelligent agents in the present invention) and evaluate the new content.

[0070] Specifically, the current film and television content data is obtained, the target film and television key features corresponding to the current film and television content data are extracted, and the target film and television key features are input into the film and television simulation evaluation model; multiple simulated user agents are generated through the film and television simulation evaluation model, and multiple simulated user agents are used to perform simulated film and television evaluation generation processing based on the current film and television content data to obtain simulated film and television evaluations of multiple simulated user agents; multiple simulated film and television evaluations are aggregated to obtain film and television simulation evaluation results.

[0071] like Figure 2 As shown, an evaluation simulation module is provided in the present invention: after receiving the feature data of new film and television content, the film and television simulation evaluation model is called to generate intelligent agents simulating different users, evaluate the new film and television, generate virtual evaluations, and store the results in the evaluation result database.

[0072] Generally, multiple different intelligent agents can be generated based on demand and computing resources. For example, they can be divided into basic evaluation mode and deep evaluation mode. Among them, the basic evaluation mode: usually generates 5-10 intelligent agents representing the main user groups (age and gender divisions), young women aged 18-24, adult men aged 25-34, etc.; deep evaluation mode: can generate more intelligent agents, such as 30-100, further subdivide user groups, and consider differences in more dimensions, such as different regions, different occupations, different preferences, and different viewing platforms (cinema, online).

[0073] The following is an example of generating an agent prompt: You are a 22-25-year-old white-collar female from XX City. You enjoy watching urban romance dramas and have watched "Love Myth." You commented on "Actor 1, Actor 2, Actor 3. Each of these three actresses, no longer young, possesses their own charm. The film's portrayal of Shanghai is truly captivating, with its streetside cafes, imported food shops, pajamas-clad neighbors, and narrow alleyways—detailed yet unforced. The dialogue is full of golden lines, and I laughed throughout. As a female viewer, I felt understood and moved." Please make an evaluation based on the input film and television information (this is a simulated user agent generated by the large model. The user role simulated by the agent is a {22-25}-year-old {white-collar} {female} in {XX city}. It can learn the evaluation of the drama "Love Myth" by the {22-25}-year-old {white-collar} {female} in {XX city}. Once the learning is completed, it can evaluate other related film and television dramas from the perspective of the {22-25}-year-old {white-collar} {female} in {XX city}).

[0074] Example review prompt: The following is film and television information: {Movie} Title {"Good Stuff"}, Genre {Drama, Romance}, Director {XX Director}, Screenwriter {XX Screenwriter}, Starring {Actor 4, Actor 5, Actor 6, Actor 7, Actor 8, Actor 9, Actor 10, Actor 11, Actor 12, Actor 13, Actor 14, Actor 15, Actor 16, Actor 17}, Plot Summary: "Single mother Wang Tiemei (actor 4) moves into a new home with her child Wang Moli (actor 6) and meets her supposedly clear-headed neighbor Xiaoye (actor 5). These two women have very different personalities: one strong, the other soft; one skilled at mothering, the other a perpetual liar. Facing old traumas and new challenges, they find warmth and comfort in each other. Meanwhile, Wang Tiemei's ex-husband (actor 8) constantly stirs up trouble, while her daughter's drum teacher (actor 7) seems to hold new possibilities." As awakened women and men who have learned about gender issues, what new problems will they encounter and how will they view themselves and the world? Lines (text), soundtrack (audio can be attached if supported), keyframes (keyframe features or supported images can be attached directly). In this scenario, the simulated film review output is a simulated evaluation of the movie "Good Stuff" by white-collar women aged 22-25 in XX City.

[0075] Furthermore, the present invention includes a data update module that regularly acquires new data from the data source to update the preprocessing database and model library. It is understood that the present invention regularly acquires new user viewing data and film content data from the data source to retrain and optimize the large model, ensuring that its evaluation and simulation capabilities can keep pace with market changes and evolving user habits.

[0076] The present invention also provides a user interface module: it provides an operating interface for system users to facilitate operations such as uploading film and television content, viewing evaluation results, setting system parameters (parameters of the film review system, such as the number and type of intelligent agents, etc.), and obtaining data update reports.

[0077] Beneficial effects of the present invention:

[0078] By collecting multi-dimensional user viewing data and using a large model for learning and simulated evaluation, this invention can reflect the potential evaluations of new film and television content by different user groups, providing a reliable basis for the early evaluation of film and television projects and helping producers reduce investment risks. Furthermore, the evaluation results based on the large model can help film and television distributors formulate precise promotional strategies, improving promotional effectiveness and resource utilization efficiency.

[0079] The main innovations of the present invention include: 1. Multi-dimensional user data fusion: The present invention comprehensively collects and integrates multi-dimensional data such as user age, gender, region, viewing history and evaluation preferences, so that the large model can learn the evaluations of different user groups, thereby generating simulated evaluations that are more in line with actual user evaluations. 2. Continuous learning and updating of the large model: The present invention can regularly incorporate new user viewing data and film and television information, update the training data of the large model, and dynamically adjust the evaluation simulation strategy. It can effectively respond to the rapid changes in the film and television market and the continuous evolution of user habits, ensuring that the rating results always maintain a high degree of consistency with the actual situation.

[0080] Further, if Figure 3 As shown, based on the above-mentioned film and television simulation evaluation method based on a large language model, the present invention also provides a film and television simulation evaluation system based on a large language model, wherein the film and television simulation evaluation system based on a large language model includes:

[0081] The data preprocessing module 51 is used to obtain user viewing data and corresponding film and television content data, and preprocess the user viewing data and the film and television content data to obtain target user viewing data and film and television key features;

[0082] The film and television simulation evaluation model generation module 52 is used to determine a pre-trained model, and perform model training and model fine-tuning on the pre-trained model according to the target user viewing data and the key features of the film and television to obtain a film and television simulation evaluation model;

[0083] The film and television simulation evaluation result output module 53 is used to obtain current film and television content data, input the current film and television content data into the film and television simulation evaluation model, and output the film and television simulation evaluation result.

[0084] Further, if Figure 4As shown, based on the above-mentioned film and television simulation evaluation method and system based on the large language model, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 4 Only some of the components of the terminal are shown, but it should be understood that implementation of all of the shown components is not required, and more or fewer components may be implemented instead.

[0085] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Furthermore, the memory 20 may also include both an internal storage unit of the terminal and an external storage device. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code of the installation terminal. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a film and television simulation evaluation program 40 based on a large language model is stored on the memory 20, and the film and television simulation evaluation program 40 based on a large language model can be executed by the processor 10, thereby realizing the film and television simulation evaluation method based on a large language model in the present application.

[0086] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run the program code or process data stored in the memory 20, such as executing the film and television simulation evaluation method based on the large language model.

[0087] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch screen, etc. The display 30 is used to display information on the terminal and to display a visual user interface.

[0088] In one embodiment, when the processor 10 executes the film and television simulation evaluation program 40 based on the large language model in the memory 20, the following steps are implemented:

[0089] Obtaining user viewing data and corresponding film and television content data, and preprocessing the user viewing data and the film and television content data to obtain target user viewing data and film and television key features;

[0090] Determine a pre-trained model, and perform model training and model fine-tuning on the pre-trained model based on the target user's viewing data and the key features of the film and television to obtain a film and television simulation evaluation model;

[0091] The current film and television content data is obtained, and the current film and television content data is input into the film and television simulation evaluation model, and the film and television simulation evaluation result is output.

[0092] Wherein, the preprocessing includes a first preprocessing and a feature extraction process;

[0093] The step of obtaining user viewing data and corresponding film and television content data, and preprocessing the user viewing data and the film and television content data to obtain target user viewing data and film and television key features, specifically includes:

[0094] Acquiring user movie viewing data using a web crawler technology or a data interface calling technology, and performing the first preprocessing on the user movie viewing data to obtain target user movie viewing data, wherein the first preprocessing includes data cleaning, denoising, and normalization.

[0095] The film and television content data corresponding to the target user's viewing data is obtained, and the feature extraction process is performed on the film and television content data to obtain film and television key features.

[0096] The step of obtaining the film and television content data corresponding to the target user's viewing data and performing the feature extraction process on the film and television content data to obtain key film and television features specifically includes:

[0097] Obtaining film information from the target user's viewing data, and obtaining corresponding film and television content data based on the film information;

[0098] The feature extraction process is performed on the film and television content data to obtain key film and television features, wherein the feature extraction process includes text content analysis process, audio-visual content analysis process, structured data analysis process and narrative structure analysis process.

[0099] The step of determining a pre-trained model and performing model training and model fine-tuning on the pre-trained model based on the target user's viewing data and the key features of the film and television to obtain a film and television simulation evaluation model specifically includes:

[0100] Associating the target user's movie viewing data with the key features of the film and television to obtain a training sample data set;

[0101] Performing data segmentation processing on the training sample data set to obtain a target sample data set, wherein the target sample data set includes a training set, a validation set, and a test set;

[0102] A pre-trained model is determined, and model training processing, model fine-tuning processing and model optimization processing are performed on the pre-trained model according to the target sample data set to obtain a film and television simulation evaluation model.

[0103] The step of performing model training, model fine-tuning, and model optimization on the pre-trained model according to the target sample data set to obtain a film and television simulation evaluation model specifically includes:

[0104] Performing model training processing on the pre-trained model according to the training set to obtain an initial film and television simulation evaluation model;

[0105] Performing model fine-tuning processing on the initial film and television simulation evaluation model according to the verification set to obtain a fine-tuned film and television simulation evaluation model;

[0106] The fine-tuned film and television simulation evaluation model is optimized according to the test set to obtain a film and television simulation evaluation model.

[0107] The fine-tuning of the initial film and television simulation evaluation model according to the validation set to obtain the fine-tuned film and television simulation evaluation model specifically includes:

[0108] Inputting the verification set into the initial film and television simulation evaluation model, and outputting the initial film and television simulation evaluation result;

[0109] Obtaining actual film and television evaluation results, and comparing the initial film and television simulation evaluation results with the actual film and television evaluation results to obtain a comparison result;

[0110] The model parameters of the initial film and television simulation evaluation model are adjusted according to the comparison result to obtain a fine-tuned film and television simulation evaluation model.

[0111] The step of obtaining current film and television content data, inputting the current film and television content data into the film and television simulation evaluation model, and outputting the film and television simulation evaluation results specifically includes:

[0112] Acquire current film and television content data, extract target film and television key features corresponding to the current film and television content data, and input the target film and television key features into the film and television simulation evaluation model;

[0113] Generate multiple simulated user agents through the film and television simulation evaluation model, and perform simulated film and television evaluation generation processing based on the current film and television content data by the multiple simulated user agents to obtain simulated film and television evaluations of the multiple simulated user agents;

[0114] A plurality of the simulated film and television evaluations are aggregated to obtain a film and television simulation evaluation result.

[0115] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a film and television simulation evaluation program based on a large language model, and when the film and television simulation evaluation program based on a large language model is executed by a processor, the steps of the film and television simulation evaluation method based on a large language model as described above are implemented.

[0116] In summary, the present invention provides a film and television simulation evaluation method, system, and terminal based on a large language model. The method includes: obtaining user viewing data and corresponding film and television content data, and preprocessing the user viewing data and the film and television content data to obtain target user viewing data and film and television key features; determining a pre-trained model, and performing model training and model fine-tuning on the pre-trained model based on the target user viewing data and the film and television key features to obtain a film and television simulation evaluation model; obtaining current film and television content data, and inputting the current film and television content data into the film and television simulation evaluation model, and outputting a film and television simulation evaluation result. The present invention trains the pre-trained model by obtaining user viewing data and film and television content data, thereby constructing a film and television simulation evaluation model. The film and television simulation evaluation model can achieve accurate analysis of the current film and television content data, thereby generating accurate film and television simulation evaluation results, which can reflect the potential evaluation of new film and television content by different user groups, provide a reliable basis for the early evaluation of film and television projects, and help film and television distributors formulate accurate promotion strategies based on the output film and television simulation evaluation results, thereby improving the promotion effect and resource utilization efficiency of new film and television works.

[0117] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or terminal comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or terminal comprising the element.

[0118] Of course, those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware (such as a processor, controller, etc.) through a computer program. The program can be stored in a computer-readable storage medium that can be read by a computer. When the program is executed, it can include the processes in the above-described method embodiments. The computer-readable storage medium can be a memory, a magnetic disk, an optical disk, etc.

[0119] It should be understood that the application of the present invention is not limited to the above examples. For those skilled in the art, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A film and television simulation evaluation method based on a large language model, characterized in that: The film and television simulation evaluation method based on the large language model includes: Obtaining user viewing data and corresponding film and television content data, and preprocessing the user viewing data and the film and television content data to obtain target user viewing data and film and television key features; Determine a pre-trained model, and perform model training and model fine-tuning on the pre-trained model based on the target user's viewing data and the key features of the film and television to obtain a film and television simulation evaluation model; Current film and television content data is acquired, and the current film and television content data is input into the film and television simulation evaluation model, and a film and television simulation evaluation result is output.

2. The film and television simulation evaluation method based on a large language model according to claim 1, characterized in that: The preprocessing includes a first preprocessing and a feature extraction process; The acquiring of user viewing data and corresponding film and television content data, and preprocessing of the user viewing data and the film and television content data to obtain target user viewing data and film and television key features, specifically includes: Acquiring user movie viewing data using a web crawler technology or a data interface calling technology, and performing the first preprocessing on the user movie viewing data to obtain target user movie viewing data, wherein the first preprocessing includes data cleaning, denoising, and normalization. The film and television content data corresponding to the target user's viewing data is obtained, and the feature extraction process is performed on the film and television content data to obtain film and television key features.

3. The film and television simulation evaluation method based on a large language model according to claim 2, characterized in that: The obtaining of the film and television content data corresponding to the target user's viewing data, and performing the feature extraction process on the film and television content data to obtain key film and television features, specifically includes: Obtaining film information from the target user's viewing data, and obtaining corresponding film and television content data based on the film information; The feature extraction process is performed on the film and television content data to obtain key features of the film and television, wherein the feature extraction process includes text content analysis process, audio-visual content analysis process, structured data analysis process and narrative structure analysis process.

4. The film and television simulation evaluation method based on a large language model according to claim 1, characterized in that: The determining of the pre-trained model, and performing model training and model fine-tuning on the pre-trained model according to the target user viewing data and the key features of the film and television to obtain a film and television simulation evaluation model specifically includes: Associating the target user's movie viewing data with the key features of the film and television to obtain a training sample data set; Performing data segmentation processing on the training sample data set to obtain a target sample data set, wherein the target sample data set includes a training set, a validation set, and a test set; A pre-trained model is determined, and model training processing, model fine-tuning processing and model optimization processing are performed on the pre-trained model according to the target sample data set to obtain a film and television simulation evaluation model.

5. The film and television simulation evaluation method based on a large language model according to claim 4 is characterized in that: The method further comprises: performing model training, model fine-tuning, and model optimization on the pre-trained model according to the target sample data set to obtain a film and television simulation evaluation model; Performing model training processing on the pre-trained model according to the training set to obtain an initial film and television simulation evaluation model; Performing model fine-tuning processing on the initial film and television simulation evaluation model according to the verification set to obtain a fine-tuned film and television simulation evaluation model; The fine-tuned film and television simulation evaluation model is optimized according to the test set to obtain a film and television simulation evaluation model.

6. The film and television simulation evaluation method based on a large language model according to claim 5, characterized in that: The fine-tuning of the initial film and television simulation evaluation model according to the validation set to obtain the fine-tuned film and television simulation evaluation model specifically includes: Inputting the verification set into the initial film and television simulation evaluation model, and outputting the initial film and television simulation evaluation result; Obtaining actual film and television evaluation results, and comparing the initial film and television simulation evaluation results with the actual film and television evaluation results to obtain a comparison result; The model parameters of the initial film and television simulation evaluation model are adjusted according to the comparison result to obtain a fine-tuned film and television simulation evaluation model.

7. The film and television simulation evaluation method based on a large language model according to claim 1, characterized in that: The obtaining of current film and television content data, inputting the current film and television content data into the film and television simulation evaluation model, and outputting the film and television simulation evaluation result specifically includes: Acquire current film and television content data, extract target film and television key features corresponding to the current film and television content data, and input the target film and television key features into the film and television simulation evaluation model; Generate multiple simulated user agents through the film and television simulation evaluation model, and perform simulated film and television evaluation generation processing based on the current film and television content data by the multiple simulated user agents to obtain simulated film and television evaluations of the multiple simulated user agents; A plurality of the simulated film and television evaluations are aggregated to obtain a film and television simulation evaluation result.

8. A film and television simulation evaluation system based on a large language model, characterized in that: The film and television simulation evaluation system based on the large language model includes: A data preprocessing module is used to obtain user viewing data and corresponding film and television content data, and preprocess the user viewing data and the film and television content data to obtain target user viewing data and film and television key features; A film and television simulation evaluation model generation module is used to determine a pre-trained model, and perform model training and model fine-tuning on the pre-trained model based on the target user viewing data and the key features of the film and television to obtain a film and television simulation evaluation model; The film and television simulation evaluation result output module is used to obtain current film and television content data, input the current film and television content data into the film and television simulation evaluation model, and output the film and television simulation evaluation result.

9. A terminal, characterized in that: The terminal includes: a memory, a processor, and a film and television simulation evaluation program based on a large language model stored in the memory and runnable on the processor. When the film and television simulation evaluation program based on a large language model is executed by the processor, the steps of the film and television simulation evaluation method based on a large language model are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a film and television simulation evaluation program based on a large language model. When the film and television simulation evaluation program based on a large language model is executed by a processor, the steps of the film and television simulation evaluation method based on a large language model are implemented as described in any one of claims 1 to 7.

Images