System for automatically generating video script creation

Abstract

The invention discloses a system for automatically generating video script creation, which relates to the technical field of script generation and comprises a content generation module for automatically generating a video script by adopting a script generation algorithm according to a theme and keywords input by a user; the retrieval experience enhancement module is used for collecting feedback data of the user to the video script and processing the feedback data to obtain quantized data, and adjusting a script generation algorithm based on the quantized data; the preference learning module is used for collecting historical creation data of a plurality of video scripts generated by using a script generation algorithm and corresponding feedback data, and training an online preference model; and the online iterative optimization module is used for acquiring feedback data of the user on the generated video script and the trained online preference model, and carrying out iterative optimization on a script generation algorithm. Through the scheme, multiple video scripts can be quickly generated according to user requirements, the topic selection direction of video creation is enriched, the video creation time is shortened, and the creation quality of the video scripts is continuously improved according to historical experience summarization.

Description

Technical Field

[0001] The present invention relates to the technical field of script generation, and in particular to a system for automatically generating video script creation. Background Art

[0002] With the popularity of short videos on self-media, most video shootings require the creation of corresponding video scripts. Currently, video script creation is usually done by manpower. According to the content and theme you want to express, you need to combine hot topics to conceive the core, and then specify the lens, scene, content, lines, duration, camera movement and props to form a video script draft. The time required to produce a video script needs to increase linearly, the scope of topic selection is affected by personal differences, successful experiences are not well accumulated and summarized for reuse, and there is a lack of automated content-assisted generation capabilities.

[0003] The existing video script generation lacks automated content-assisted creation methods, resulting in a linear increase in time consumption, low topic selection efficiency, and inability to reuse experience during the creation process. Therefore, there is an urgent need for a system that automatically generates video scripts to solve the above problems. Summary of the invention

[0004] In order to solve the above-mentioned deficiencies of the prior art, the present invention provides a system for automatically generating video script creation, the system comprising:

[0005] Content generation module: Based on the topics and keywords input by the user, the script generation algorithm is used to automatically generate video scripts;

[0006] Retrieval experience enhancement module: collects user feedback data on generated video scripts and processes them to obtain quantitative data, and adjusts and optimizes the script generation algorithm based on the quantitative data;

[0007] Preference learning module: collects historical creation data of multiple video scripts generated by the script generation algorithm and corresponding feedback data to obtain collection results, and trains an online preference model based on the collection results;

[0008] Online iterative optimization module: obtains user feedback data on generated video scripts and the trained online preference model, and iteratively optimizes the script generation algorithm.

[0009] The present invention is implemented through the following technical scheme: This scheme provides a system for automatically generating video script creation, including: a content generation module, a retrieval experience enhancement module, a preference learning module and an online iterative optimization module.

[0010] Content generation module: used to automatically generate video scripts using a script generation algorithm based on the topics and keywords input by the user; this module uses deep learning technology to analyze a large amount of video script data, learn and simulate the thinking mode and expression habits of human creation, and thus generate script content that meets user needs;

[0011] Retrieval experience enhancement module: collects user feedback data on generated video scripts and processes them to obtain quantitative data. The feedback data includes but is not limited to the creativity, structure and language style of the scripts. The script generation algorithm is then adjusted and optimized based on the quantified feedback data.

[0012] Preference learning module: collects historical creation data and corresponding feedback data of multiple video scripts generated by the script generation algorithm, and trains the online preference model to meet the personalized creation needs of users; this module can identify the user's style preference, theme preference, language preference, etc., so as to provide more personalized content in the subsequent script generation;

[0013] Online iterative optimization module: obtains user feedback data on generated video scripts and the trained online preference model, and continuously iterates and optimizes the script generation algorithm; through machine learning technology, the script generation algorithm can self-learn and evolve, thereby continuously improving the quality and creation efficiency of the script.

[0014] As an optional technical solution, the automatic generation of video scripts using a script generation algorithm includes: building a deep language learning model; inputting a number of video script data into the deep language learning model for analysis and processing, and training the deep language learning model; obtaining the topic and keywords input by the user, and inputting them into the trained deep language learning model to automatically generate a video script.

[0015] As an optional technical solution, the script algorithm is specifically as follows:

[0016] (y 0 ,y 1 , ..., y i )=ρ(x),i∈N

[0017] y is the video script output by the deep language learning model, ρ is the deep language learning model, x is the input of the deep language learning model, and i is the number of generated video scripts.

[0018] As an optional technical solution, the script generation algorithm is adjusted and optimized based on the quantitative data as follows:

[0019] format n =ρ(prompt n )

[0020] Define the user to enter any question as n, prompt n is the format description of problem n, format n Formatted data generated for deep language learning models;

[0021]

[0022] retrieval(format n ) is used to retrieve formatted data, is the retrieval result for formatted data, k is the k retrieval results with the highest relevance among the retrieval results for formatted data;

[0023] (y 0 ,y 1 , ..., y i )=ρ(a+experience 1,2，...，k ), i∈N

[0024] (a+experience 1,2，...，k ) is the input of the deep language learning model, and a is the content input by the user.

[0025] As an optional technical solution, based on the collection results, training the online preference model is specifically as follows:

[0026] score 1,2，...，i =RM(y 1,2，...，i ), i∈N

[0027] RM is a reward model, which is used to score multiple batch video scripts y. 1,2，...，i The scoring result output by the reward model;

[0028]

[0029] E A is the winning expectation of player A, R A Output the score of participant A for the reward model, R B Outputting a score of participant B for the reward model;

[0030]

[0031] is the updated score of participant A, and K is the dynamic adjustment coefficient.

[0032] As an optional technical solution, the iterative optimization of the script generation algorithm is specifically as follows:

[0033]

[0034] L (r) is the loss function, E (x，experience，y)～D represents the expectation of all pairs (a, experience, y) sampled from the data distribution D; (a, experience, y) is a data pair, representing the input, latent variable and output respectively; log σ is the logarithm of the output of the Sigmoid function, σ is the sigmoid function; r(a, experience, y c ) indicates that the model takes (a+experience) as input, y c is the score at the time of output; r(a, experience, y r ) indicates that the model takes (a+experience) as input, y r is the score at the time of output; μ^(x, y) is the marginal function.

[0035] One or more technical solutions provided by the present invention have at least the following technical effects or advantages:

[0036] The present invention uses a deep learning language model as its basis, trains an online preference model according to industry-related technologies and human feedback data, and generates corresponding experience essentials, thereby enhancing the quality of the script content generated next time. Through this system, multiple video scripts can be quickly generated according to user needs, enriching the topic selection direction of video creation, shortening the video creation time, and continuously improving the creation quality of video scripts based on historical experience summaries. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation on the embodiments of the present invention;

[0038] Figure 1 It is a schematic diagram of the composition of a system for automatically generating video script creation in the present invention. DETAILED DESCRIPTION

[0039] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those within the scope of this description. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0041] Example

[0042] Please refer to Figure 1 , Figure 1 This is a schematic diagram of the composition of a system for automatically generating video script creation in the present invention. The system includes:

[0043] Content generation module: Based on the topics and keywords input by the user, the script generation algorithm is used to automatically generate video scripts;

[0044] Retrieval experience enhancement module: collects user feedback data on generated video scripts and processes them to obtain quantitative data, and adjusts and optimizes the script generation algorithm based on the quantitative data;

[0045] Preference learning module: collects historical creation data of multiple video scripts generated by the script generation algorithm and corresponding feedback data to obtain collection results, and trains an online preference model based on the collection results;

[0046] Online iterative optimization module: obtains user feedback data on generated video scripts and the trained online preference model, and iteratively optimizes the script generation algorithm.

[0047] The specific embodiments of the present invention are as follows:

[0048] Content generation module: used to automatically generate video scripts using a script generation algorithm based on the topics and keywords input by the user; this module uses deep learning technology to analyze a large amount of video script data, learn and simulate the thinking mode and expression habits of human creation, and thus generate script content that meets user needs. Automatically generating video scripts using a script generation algorithm includes: building a deep language learning model; inputting a number of video script data into the deep language learning model for analysis and processing, and training the deep language learning model; obtaining the topics and keywords input by the user, and inputting them into the trained deep language learning model to automatically generate video scripts.

[0049] The script algorithm is specifically as follows:

[0050] (y 0 ,y 1 , ..., y i )=ρ(x),i∈N

[0051] y is the video script output by the deep language learning model, ρ is the deep language learning model, x is the input of the deep language learning model, and i is the number of generated video scripts.

[0052] Among them, this embodiment uses an automated content generation module to batch generate preliminary drafts of video scripts of different styles and angles for the same topic based on experience, which greatly shortens the creation time and improves the creation efficiency.

[0053] Retrieval experience enhancement module: collects user feedback data on generated video scripts and processes them to obtain quantitative data. The feedback data includes but is not limited to the creativity, structure and language style of the scripts. The script generation algorithm is then adjusted and optimized based on the quantified feedback data.

[0054] The script generation algorithm is adjusted and optimized based on the quantitative data as follows:

[0055] format n =ρ(prompt n )

[0056] Define the user to enter any question as n, prompt n is the format description of problem n, format n Formatted data generated for deep language learning models;

[0057]

[0058] retrieval(format n ) is used to retrieve formatted data, is the retrieval result for formatted data, k is the k retrieval results with the highest relevance among the retrieval results for formatted data;

[0059] (y 0 ,y 1 , ..., y i )=ρ(a+experience 1,2，...，k ), i∈N

[0060] (a+experience 1,2，...，k ) is the input of the deep language learning model, and a is the content input by the user.

[0061] The retrieval experience enhancement module is a machine learning technology that enhances the output of the model by retrieving relevant information. Specifically, it first analyzes the user's query, then retrieves the most relevant information, and inputs the retrieved information into the model together with the query in the hope of obtaining more accurate and richer output.

[0062] Among them, the retrieval experience enhancement module of this embodiment enables the entire system to continuously learn and adapt to the user's creative style, realizes the effective reuse of creative experience and materials, and reduces resource waste.

[0063] Preference learning module: collects historical creation data and corresponding feedback data of multiple video scripts generated by the script generation algorithm, and trains the online preference model to adapt to the user's personalized creation needs; this module can identify the user's style preferences, theme preferences, language preferences, etc., so as to provide more personalized content in subsequent script generation.

[0064] Based on the collected results, the online preference model is trained as follows:

[0065] score 1,2，...，i =RM(y 1,2，...，i ), i∈N

[0066] RM is a reward model, which is used to score multiple batch video scripts y. 1,2，...，i The scoring result output by the reward model;

[0067]

[0068] E A is the winning expectation of player A, R A Output the score of participant A for the reward model, R B Outputting a score of participant B for the reward model;

[0069]

[0070] is the updated score of participant A, and K is the dynamic adjustment coefficient.

[0071] Among them, the Elo ranking system is a scoring system used to evaluate and compare the relative skills of participants. In machine learning, the Elo system can be used to evaluate the quality of model output and update the model's score through comparison. It is time-sensitive and will change with the hot cycle, so that the created video script can be more in line with the hot spot. A is the victory expectation of participant A in the Elo system, which is used to obtain the expected performance of participant A when compared with another participant B. A Output the score of participant A for the reward model, R B Output the score of participant B for the reward model; use E A The score of participant A is updated based on the output result, and K is dynamically adjusted with the amount of data in the Elo system.

[0072] Among them, the preference learning module in this embodiment can generate scripts that better meet user needs based on the user's creative habits and preferences combined with a large amount of accumulated experience data, broaden the creative space, and improve the novelty and diversity of the scripts.

[0073] Online iterative optimization module: obtains user feedback data on generated video scripts and the trained online preference model, and continuously iterates and optimizes the script generation algorithm; through machine learning technology, the script generation algorithm can self-learn and evolve, thereby continuously improving the quality and creation efficiency of the script.

[0074] The iterative optimization of the script generation algorithm is specifically as follows:

[0075]

[0076] L (r) is the loss function, E (x，experience，y)～D represents the expectation of all pairs (a, experience, y) sampled from the data distribution D; (a, experience, y) is a data pair, representing the input, latent variable and output respectively; log σ is the logarithm of the output of the Sigmoid function, σ is the Sigmoid function; r(a, experience, y c ) indicates that the model takes (a+experience) as input, y c is the score at the time of output; r(a, experience, y r ) indicates that the model takes (a+experience) as input, y r is the score at the time of output; μ^(x, y) is the marginal function.

[0077] Among them, μ^(x, y) is the marginal function, which enables the model to flexibly and effectively process data with different preference strengths, thereby improving the performance and accuracy of the model in preference modeling tasks.

[0078] Among them, the iterative optimization module in this embodiment ensures that the entire system can continue to improve based on user feedback, and the generated scripts will increasingly meet the user's personalized needs, thereby improving the quality of creation.

[0079] Furthermore, the script generation algorithm after adjustment, optimization and iterative optimization has the following improvements on script generation:

[0080] Obtain user rewriting feedback data and expansion feedback data on the initial script, the rewriting feedback data includes other data such as script style change data, script concept adjustment data, script framework adjustment data and script plot adjustment data, and the expansion feedback data includes other data such as background addition description data and character depth increase data; perform multi-dimensional analysis on the initial script to obtain analysis results, and determine optimization data for the initial script based on the analysis results; based on the optimization data for the initial script and the optimized script generation algorithm, rewrite and expand the initial script to obtain a first script, obtain new rewriting and expansion feedback data for the first script based on the user's interactive operation, and rewrite and expand the first script again based on the new rewriting and expansion feedback data and the optimized script generation algorithm, until a target script that meets the user's requirements is obtained.

[0081] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0082] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A system for automatically generating video script creation, characterized in that: The system comprises: Content generation module: Based on the topics and keywords input by the user, the script generation algorithm is used to automatically generate video scripts; Retrieval experience enhancement module: collects user feedback data on generated video scripts and processes them to obtain quantitative data, and adjusts and optimizes the script generation algorithm based on the quantitative data; Preference learning module: collects historical creation data of multiple video scripts generated by the script generation algorithm and corresponding feedback data to obtain collection results, and trains an online preference model based on the collection results; Online iterative optimization module: obtains user feedback data on generated video scripts and the trained online preference model, and iteratively optimizes the script generation algorithm.

2. A system for automatically generating video script creation according to claim 1, characterized in that: Automatically generating video scripts using a script generation algorithm includes: building a deep language learning model; inputting a number of video script data into the deep language learning model for analysis and processing, and training the deep language learning model; obtaining topics and keywords input by the user, and inputting them into the trained deep language learning model to automatically generate video scripts.

3. A system for automatically generating video script creation according to claim 2, characterized in that: The script algorithm is specifically as follows: (y0,y1,…,y i )=ρ(x),i∈N y is the video script output by the deep language learning model, ρ is the deep language learning model, x is the input of the deep language learning model, and i is the number of generated video scripts.

4. A system for automatically generating video script creation according to claim 3, characterized in that: The script generation algorithm is adjusted and optimized based on the quantitative data as follows: format n =ρ(prompt n ) Define the user to enter any question as n, prompt n is the format description of problem n, format n Formatted data generated for deep language learning models; retrieval(format n ) is used to retrieve formatted data, is the retrieval result for formatted data, k is the k retrieval results with the highest relevance among the retrieval results for formatted data; (y0,y1,…,y i )=ρ(a+experience 1,2，…，k ),i∈N (a+experience 1,2，…，k ) is the input of the deep language learning model, and a is the content input by the user.

5. A system for automatically generating video script creation according to claim 4, characterized in that: Based on the collected results, the online preference model is trained as follows: score 1,2，…，i =RM(y 1,2，…，i ),i∈N RM is a reward model, which is used to score multiple batch video scripts y. 1,2，…，i The scoring result output by the reward model; E A is the winning expectation of player A, R A Output the score of participant A for the reward model, R B Outputting a score of participant B for the reward model; R A =R A +K(S A ―E A ),S A ∈(1,0.5,0) R′ A is the updated score of participant A, and K is the dynamic adjustment coefficient.

6. A system for automatically generating video script creation according to claim 5, characterized in that: The iterative optimization of the script generation algorithm is specifically as follows: L (r) is the loss function, E (x，experience，y)～D represents the expectation of all pairs (a, experience, y) sampled from the data distribution D; (a, experience, y) is a data pair, representing the input, latent variable and output respectively; log σ is the logarithm of the output of the Sigmoid function, σ is the Sigmoid function; r(a, experience, y c ) indicates that the model takes (a+experience) as input, y c is the score at the time of output; r(a, experience, y r ) indicates that the model takes (a+experience) as input, y r is the score at the time of output; μ^(x, y) is the marginal function.

Images