Method for generating content, related device and computer program product

Abstract

The invention provides a content generation method, a related device and a computer program product. The method comprises the following steps: acquiring demand description information of target generated content; performing demand decomposition on the demand description information through a demand processing model to obtain material demand information, training an initial demand processing model according to the difference between the training material and the sample material to obtain the demand processing model, and performing demand decomposition on the sample demand description information by the training material according to the initial demand processing model to obtain the material demand information; after the training material demand information is obtained, the sample material is extracted from a material library based on the training material demand information, and the sample material is used for meeting the material demand of the sample demand description information and is extracted from the material library; extracting a target material corresponding to the material demand information from a material library; and generating target generation content based on the target material. Therefore, the demand processing model with stronger perception ability and stronger material selection ability for the material library can be utilized to more accurately generate the material demand information for obtaining the materials required by the processing demand of the user according to the content in the material library, so that the model illusion is reduced, and the user experience is improved. Therefore, the task of demand processing and content generation can be better completed.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, electronic device, computer-readable medium, and computer program product for generating content. Background Technology

[0002] In today's era of rapid development in information technology, artificial intelligence technology has been widely applied in many fields, especially in the area of ​​personalized content generation, where it has demonstrated strong application potential.

[0003] For example, AI technology based on large language models can generate content in the form of text, video, etc., according to the user's description information, creating customized content for the user. This approach greatly reduces the user's creation and execution costs, enabling them to produce media content for viewing and sharing more cost-effectively and efficiently.

[0004] Against this backdrop, how to better fulfill these needs and generate content is a matter of concern and urgent need. Summary of the Invention

[0005] This application provides a method, apparatus, electronic device, computer-readable storage medium, and computer program product for generating content, which can utilize a demand processing model with stronger perception and selection capabilities of a material library to more accurately generate material demand information from the content in the material library to obtain the materials needed to meet the user's processing needs, reduce model illusion, and thus better complete the demand processing and content generation tasks.

[0006] One aspect of this application provides a method for generating content, comprising: obtaining requirement description information of the target generated content; decomposing the requirement description information into requirement information using a requirement processing model to obtain material requirement information, wherein the requirement processing model is obtained by training an initial requirement processing model based on the difference between training materials and sample materials; after the training materials decompose the sample requirement description information into requirement information using the initial requirement processing model to obtain training material requirement information, the training material requirement information is extracted from a material library based on the training material requirement information; sample materials are used to meet the material requirements of the sample requirement description information and are extracted from the material library; extracting target materials corresponding to the material requirement information from the material library; and generating the target generated content based on the target materials.

[0007] Another aspect of this application provides an apparatus for generating content, comprising: a requirement description acquisition module configured to acquire requirement description information of target generated content; a requirement description decomposition module configured to decompose the requirement description information into material requirement information using a requirement processing model, wherein the requirement processing model is obtained by training an initial requirement processing model based on the difference between training materials and sample materials, the training materials are obtained by decomposing the sample requirement description information into requirement information using the initial requirement processing model, and then extracting the training material requirement information from a material library based on the training material requirement information, and the sample materials are used to meet the material requirements of the sample requirement description information and are extracted from the material library; a material extraction module configured to extract target materials corresponding to the material requirement information from the material library; and a content generation module configured to generate target generated content based on the target materials.

[0008] In another aspect of this application, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method of generating content as provided above.

[0009] Another aspect of this application provides a computer-readable storage medium having computer program instructions stored thereon, which can be executed by a processor to implement the method of generating content as provided above.

[0010] Another aspect of this application is a computer program product that includes a computer program having computer program instructions stored thereon, which, when executed by a processor, can implement the method for generating content as provided above.

[0011] In the solution provided in this application embodiment, firstly, the requirement description information of the target generated content is obtained; then, the requirement description information is decomposed using a requirement processing model to obtain material requirement information. Specifically, the requirement processing model is trained based on the differences between training materials and sample materials. The training materials are then decomposed based on the sample requirement description information using the initial requirement processing model to obtain training material requirement information, which is then extracted from a material library. Sample materials are used to meet the material requirements of the sample requirement description information and are also extracted from the material library. Next, target materials corresponding to the material requirement information are extracted from the material library. Finally, the target generated content is generated based on the target materials. Therefore, by utilizing a requirement processing model with stronger perception and material selection capabilities of the material library, the material requirement information for obtaining materials needed to meet the user's processing needs can be generated more accurately using content from the material library, reducing model illusion and thus better completing the requirement processing and content generation tasks. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 A flowchart illustrating a content generation process provided in an embodiment of this application; Figure 2 A flowchart illustrating the process of training a demand processing model, as provided in an embodiment of this application; Figure 3 A flowchart illustrating another process for training a demand processing model provided in this application embodiment; Figure 4 The flowchart illustrating the specific implementation of this application in a specific application scenario includes the process of training a problem-solving model and generating content. Figure 5 A schematic diagram of a device for generating content, provided as an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device suitable for implementing the solutions in the embodiments of this application.

[0014] The same or similar reference numerals in the accompanying drawings represent the same or similar parts. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] In a typical configuration of this application, the terminal and the service network devices each include one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0017] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0018] Computer-readable media include permanent and non-permanent, removable and non-removable media, which can store information by any method or technology. Information can be computer program instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, read-only optical disc (CD-ROM), digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0019] As discussed above, how to better accomplish requirements processing and content generation tasks is a noteworthy and urgent need.

[0020] Analysis revealed that, typically, when using large language models to complete processing tasks, the large language model is configured to first generate the task execution process and flow based on the provided requirement description information. Then, the large language model itself, or by scheduling other models or devices, executes the above process and flow to obtain the processing result that satisfies the processing requirements of the requirement description information.

[0021] For example, in scenarios where user-provided requirement descriptions are used to generate user-generated content such as text and videos, the large language model can first define the outline and structure of the final generated text and video, as well as the required materials, based on the requirement descriptions. Then, it can generate the text and video by adding content accordingly. For instance, after determining the outline, text, illustrations, audio, and video are added to various parts of the outline to generate the final video file.

[0022] In this context, in order to improve the generation quality of large language models, in the embodiments of this application, the materials that the large language model can refer to and use can first be "defined" using a material library, so that the large language model can better complete the processing task by using this expert knowledge, and the scope of materials that the large language model can refer to and use can be managed to avoid the large language model using inappropriate materials that are not allowed to be used or may pose risks.

[0023] However, in this approach, considering that the large language model is trained simply by comparing whether the final generated text and video can meet the requirements (e.g., user needs), the large language model often has difficulty perceiving the material library (or more specifically, the materials included in the material library) in the process of determining the required materials and the requirements of the materials. This leads to the fact that the architecture and requirements determined by the large language model are often difficult to match well with the materials in the material library, resulting in low quality of material selection and matching, model illusion, and other problems.

[0024] To address this situation, this application provides a technical solution. First, it obtains the requirement description information of the target generated content. Then, it decomposes the requirement description information using a requirement processing model to obtain material requirement information. The requirement processing model is trained based on the differences between training materials and sample materials. After the training materials decompose the sample requirement description information using the initial requirement processing model to obtain training material requirement information, it extracts the required materials from a material library based on this information. Sample materials are used to meet the material requirements of the sample requirement description information and are also extracted from the material library. Next, it extracts the target materials corresponding to the material requirement information from the material library. Finally, it generates the target generated content based on the target materials. Therefore, by utilizing a requirement processing model with stronger perception and material selection capabilities of the material library, it can more accurately generate material requirement information from the content in the material library to obtain the materials needed to meet the user's processing needs, reducing model illusions and thus better completing the requirement processing and content generation tasks.

[0025] In practical scenarios, the execution entity of this method can be a user device, a device formed by integrating a user device and a network device through a network, or an application running on the aforementioned devices. User devices include, but are not limited to, various terminal devices such as computers, mobile phones, tablets, smartwatches, and wristbands. Network devices include, but are not limited to, network hosts, single network servers, multiple network server sets, or cloud computing-based computer sets. Here, the cloud consists of a large number of hosts or network servers based on cloud computing. Cloud computing is a type of distributed computing, consisting of a virtual computer composed of a group of loosely coupled computer sets.

[0026] When the executing entity is software, it can be installed in the electronic devices listed above. It can be implemented as multiple software programs or software modules, or as a single software program or software module, without specific limitations.

[0027] Figure 1This application illustrates a content generation process 100 provided in an embodiment of the present application. The process 100 includes at least the following processing steps: (Step) S101, Obtain the requirement description information of the target generated content; In the embodiments of this application, users can directly use, or interact with, the executing entity through their terminal devices to provide a description of their desired content to be generated. This description of desired content can be, for example, a user's needs described in natural language. For instance, generating a video to introduce or promote XX, generating text to introduce or promote YY, generating a short video to depict a leisurely afternoon, etc. Accordingly, such videos, texts, etc., can be understood as the aforementioned desired content to be generated.

[0028] S102, The requirement description information is decomposed through the requirement processing model to obtain the material requirement information; In the embodiments of this application, based on the above step S101, the executing entity can use the demand processing model to decompose the demand description information to obtain material demand information.

[0029] In the embodiments of this application, the demand processing model can be trained based on the differences between the training materials and the sample materials to obtain an initial demand processing model. The training materials then perform demand decomposition on the sample demand description information according to the initial demand processing model to obtain the training material demand information. This training material demand information is then extracted from the material library. The sample materials can be used to meet the material requirements of the sample demand description information, and are extracted from the material library.

[0030] To facilitate understanding, we can begin by discussing the process of training the initial requirement processing model to obtain the current requirement processing model. For the same purpose of understanding this process, we can also refer to... Figure 2 Let's have a discussion. Figure 2 This application illustrates a process 200 for training a demand processing model according to an embodiment of the present application.

[0031] Process 200 includes at least the following processing steps: S201, Obtain sample requirement description information, material library and sample materials; Specifically, in this step, the executing entity can first obtain sample demand description information, a material library, and sample materials corresponding to the processing requirements of the sample demand description information for training the initial demand processing model.

[0032] Similar to the discussion above, the sample requirement description information can include "requirements" used as training samples. For example, these "requirements" could be user needs described in natural language. Examples include generating a video to introduce or promote XX, generating text to introduce or promote YY, or generating a short video to depict a leisurely afternoon.

[0033] The resource library can include materials used to complete and satisfy the processing requirements described in the requirements description. For example, the resource library can include video materials, image materials, audio materials, etc., so that different processing requirements can be met later using these materials, or combinations of materials.

[0034] For example, when the sample requirement description information is exemplified as generating a video to promote city A, images of attractions in city A included in the media library, promotional videos of city A, etc., can be called to meet the processing requirements of the sample requirement description information, such as generating a video.

[0035] Accordingly, in this step, the executing entity can also obtain sample materials corresponding to the processing requirements described in the sample requirements information. These sample materials also originate from this material library.

[0036] Typically, sample material can be simply understood as the "most ideal" material in the material library that corresponds to and is used to meet the description information of the sample requirements. For example, in the example above, the images of attractions in city A that correspond to the processing requirement of generating a video to promote city A, and the promotional video for city A.

[0037] The initial requirement processing model can be built upon a large language model and pre-trained, already possessing the ability to process requirement description information and generate material requirement information to meet the processing needs of that requirement description. For example, based on its current capabilities, the initial requirement processing model can already process sample requirement description information and generate material requirement information to meet the processing needs of that sample requirement description.

[0038] Large Language Models (LLMs) are artificial intelligence models designed to understand and generate human language. Based on their understanding, LLMs can perform processing operations to obtain corresponding results. For example, an initial LLM, as discussed above, can at least process sample requirement descriptions and generate material requirement information to meet those requirements. LLMs can be trained on large amounts of text data and can perform a wide range of tasks, including text summarization, translation, sentiment analysis, and more. A key characteristic of LLMs is their large scale; they typically include a large number of parameters to help them learn complex patterns in language data. These models are often based on deep learning architectures, such as transformers, which contribute to their superior performance across various natural language processing tasks.

[0039] Material requirements information can actually express and characterize the needs for materials used to meet processing requirements. For example, for the processing requirement of generating a video promoting city A, the material requirements information could be, as discussed above, the need for materials related to attractions in city A, the need for materials from existing promotional short videos promoting city A, and so on.

[0040] For example, an initial LLM can first generate a content outline for a video promoting city A, and then determine the required materials based on each part of the content outline. For example, an initial LLM can first determine how to "promote city A" from dimensions such as historical background, development process, related attractions, and city culture.

[0041] Then, the initial requirements processing model (or, for an initial requirements processing model built on an LLM, it can also be called the initial LLM) generates an outline that includes information on the historical background, development process, attractions involved, and urban culture.

[0042] Then, based on this content outline, the initial requirements processing model determines its requirements for materials (i.e., material requirements information), such as the need for images associated with historical background, images associated with development processes, images associated with tourist attractions, and so on.

[0043] As discussed above, the initial requirement processing model can be a pre-trained LLM that already possesses the ability to process requirement description information and generate material requirement information to satisfy the processing requirements of the requirement description information. Accordingly, in some embodiments, such an initial LLM can be obtained by training a general LLM using sample requirement description information and sample material requirement information.

[0044] That is, in some embodiments, the executing entity can obtain sample requirement description information and sample material requirement information. Here, the sample requirement description information and the sample requirement description information mentioned above can be the same sample or different.

[0045] The sample material requirements information, on the other hand, is the material requirement information that can be predetermined and corresponds to the processing requirements described in the sample requirements information. For example, the sample material requirements information can record what kind of content outline is needed to meet the sample requirements description information, and what kind of materials are needed for each part of the content outline.

[0046] Then, the executing entity can train the general LLM using sample requirement description information as input and sample material requirement information as output, so that the general LLM can learn the correspondence between requirement description information and material requirement information. After training, the general LLM, as the initial LLM, will have the ability to process subsequently acquired requirement description information (e.g., sample requirement description information) and generate material requirement information to meet the corresponding processing requirements, thus obtaining the aforementioned initial requirement processing model.

[0047] S202, Call the initial demand processing model to decompose the sample demand description information and generate training material demand information; Specifically, in the embodiments of this application, after obtaining the sample requirement description information, the material library, and the sample materials corresponding to the processing requirements of the sample requirement description information based on the above S201, the execution entity can call the initial requirement processing model to process the sample requirement description information in order to generate material requirement information to meet the processing requirements of the sample requirement description information.

[0048] For ease of understanding, since this material requirement information is used in the training process, we can refer to it as "training material requirement information" for now.

[0049] S203, Extract training materials corresponding to the training material requirement information from the material library; Specifically, after generating training material requirement information based on the above S202, the executing entity can extract training materials corresponding to the training material requirement information from the material library based on this information. For example, the executing entity can extract training materials corresponding to the training material requirement information from the material library based on the "description" (semantics) of the materials in the training material requirement information, using content semantic matching.

[0050] For example, if the training material requirements include image materials of building B, the executing entity can match and find those "images" with the highest semantic similarity to building B, or those with a similarity threshold greater than or equal to the threshold, from the material library as training materials.

[0051] S204. Based on the degree of difference between the sample material and the training material, the initial demand processing model is adjusted in the direction of reducing the degree of difference to obtain the demand processing model.

[0052] Specifically, after the training material is extracted in S204 above, the executing entity can compare the degree of difference between the training material and the sample material. For example, in the "building" scenario mentioned above, the executing entity can determine the difference between the two by comparing whether the training material and the sample material both point to the same building.

[0053] For example, if both the training materials and the sample materials point to the same building, the implementing agent can executively assume that there is no difference between them. However, if they point to different buildings, the implementing agent can determine the degree of difference between the two buildings based on differences in style and type.

[0054] For example, these differences can be quantified based on pre-determined comparison dimensions and the corresponding difference scores for each dimension. Accordingly, after identifying the dimensions in which two materials differ, the implementing entity can use the corresponding difference scores to determine the degree of difference.

[0055] For example, in the case of architecture, if the objects being referred to are different but the style is the same, then such a degree of difference can have a lower difference value compared to differences where the objects are different and the styles are also different.

[0056] In some embodiments, for different dimensions, the difference scores can be configured accordingly for each different dimension. For example, these dimensions can be geographical distance (e.g., for the same building, the greater the distance between the two shots, the greater the difference and the higher the difference score), age, whether the objects are of the same type, style, etc.

[0057] Accordingly, in this step, the executing entity can adjust the initial demand processing model based on the degree of difference between the training material and the sample material, in order to reduce the degree of difference, and thus obtain the demand processing model.

[0058] In this step, the executing entity can adjust the parameters of the initial demand processing model to reduce the degree of discrepancy, thus obtaining a new demand processing model. Specifically, the executing entity can adjust the initial demand processing model so that the output material demand information after processing the sample demand description information more clearly points to and relates to the training materials. Correspondingly, the adjusted initial demand processing model, or demand processing model, can output material demand information that more clearly points to the materials in the material library, enabling subsequent requests for materials based on this material demand information to be more accurate and within the scope of the material library.

[0059] In some embodiments, to explicitly indicate this adjustment direction, the executing entity may choose to first construct a reward function based on the degree of difference between the training materials and the sample materials during this step. In this reward function, if the degree of difference between the training materials and the sample materials is low, then the processing process of the initial demand processing model can be evaluated with a higher reward value.

[0060] Then, the executing agent can use the reward function to adjust the initial demand processing model in a reinforcement learning manner, in order to reduce the degree of difference, thus obtaining the demand processing model. For example, the executing agent can adjust the initial demand processing model in a reinforcement learning manner with the goal of maximizing the reward value and the maximum cumulative reward value, so that the initial demand processing model can be adjusted in the direction of reducing the degree of difference, thus obtaining the final demand processing model.

[0061] Reinforcement learning (RL) learns optimal behavioral strategies through interaction with the environment to maximize accumulated rewards. In reinforcement learning training, the initial demand processing model generates material demand information in a way that maximizes reward value in each decision and generation of material demand information.

[0062] In some embodiments, to adapt to different demand scenarios and improve the ability and quality of demand fulfillment, the material demand information generated by the initial demand processing model and the demand processing model can simultaneously require at least two or more materials. This enables the initial demand processing model and the demand processing model to output material demand information that utilizes multiple materials, resulting in higher quality and richer content.

[0063] In cases where at least two sets of materials are required, the implementing entity can respond by constructing a reward function that, in addition to considering the degree of difference between the training materials and the sample materials, also considers the style differences between the individual training materials. That is, the reward value ultimately provided by this function will take into account not only the degree of difference between the training materials and the sample materials, but also the style differences between the individual training materials. For example, for the same degree of difference, if the style differences between the training materials required by the material demand information are smaller, then a larger reward value can be provided to the initial demand processing model.

[0064] In practice, specific styles of source materials can be configured based on the specific materials being used. For example, image-based source materials can be categorized into styles based on dimensions such as cartoon, realistic, and augmented reality. Then, the style distance between each style is determined in advance, and this style distance is used to quantify the aforementioned differences in source material styles.

[0065] Accordingly, in the process of using the reward function to adjust the initial demand processing model in the direction of reducing the degree of difference through reinforcement learning, as described above, the executing entity can further use the reward function to adjust the initial demand processing model in the direction of reducing the degree of difference and reducing the difference in material style, as described above, to obtain the demand processing model.

[0066] In some embodiments, the initial demand processing model can be trained to provide at least two or more material demand information for the same processing demand. This approach not only allows for more efficient training of the initial demand processing model (e.g., training the generation "logic" of multiple different material demand information for the initial demand processing model), but also enables the initial demand processing model to better meet diverse user needs by providing multiple candidate material demand information.

[0067] Accordingly, in this approach, during the construction of the reward function, the executing entity can choose a reward function that simultaneously considers the degree of difference between the training material and the sample material, as well as the amount of information required for the training material.

[0068] For example, in such a reward function, for the same degree of difference, if the initial demand processing model generates more training material demand information at once, it can be assigned a larger reward value.

[0069] Accordingly, in the process of using the reward function to adjust the initial demand processing model in the direction of reducing the degree of difference through reinforcement learning, the executing entity can further use the reward function to adjust the initial demand processing model in the direction of reducing the degree of difference and increasing the amount of material demand information, in order to obtain the demand processing model.

[0070] It should be understood that, in some embodiments, when constructing the reward function, the executing entity may also simultaneously consider the aforementioned degree of difference, material style difference, and quantity of material requirement information, so as to simultaneously refer to both the material style difference and the quantity of material requirement information, thereby enabling better training and adjustment of the initial requirement processing model.

[0071] In some embodiments, depending on different scenarios and training requirements, the executing entity can also refer to the user's preference for training materials when constructing the reward function to complete the construction process, so that the trained demand processing model can have capabilities in these dimensions. For example, the executing entity can also use sample materials as standard user preferences, and after determining the preference level corresponding to the training materials based on the sample preference parameters, provide a corresponding reward value based on the "preference level", so that the demand processing model can at least refer to the user's preferences during the process of generating material demand information.

[0072] Accordingly, in this step, based on different training methods, the executing entity can determine whether to complete the adjustment of the initial demand processing model and obtain the demand processing model based on criteria such as the number of training rounds reaching the required level (e.g., training rounds completed using multiple different sample demand descriptions and sample materials), the maximization of the reward value, or the convergence of the (total) reward value. For example, in the reinforcement learning method using the reward function described above, the executing entity can determine whether to complete the adjustment of the initial demand processing model when, for example, the (total) reward value converges, and use the model obtained and trained in the last round as the demand processing model.

[0073] Therefore, the demand processing model trained in the above manner has a stronger ability to perceive and select materials from the material library. This allows it to more accurately generate material demand information from the content in the library to obtain the materials needed to meet the user's processing needs, reducing model illusions. Consequently, in the process of generating content using this demand processing model, it can better accomplish the tasks of demand processing and content generation.

[0074] In such an embodiment, if the initial demand processing model is built based on an LLM, in order to improve the training efficiency of the initial demand processing model and enable the training of the initial demand processing model to generate material demand information by referring to the material library, the executing entity can require the initial demand processing model to refer to the materials included in the material library when calling the initial demand processing model to generate material demand information to meet the processing requirements.

[0075] For easier understanding, you can also refer to... Figure 3 . Figure 3 The present application illustrates another training requirement processing model process 300, which includes at least the following processing steps: S301, Obtain sample requirement description information, material library and sample materials; This step is similar to the implementation process provided by S201 discussed above. For the same parts, please refer to the content discussed in S201 above. It will not be repeated here.

[0076] S302, Construct requirement decomposition prompts based on sample requirement description information and material library; Specifically, in this step, the executing entity can construct requirement decomposition prompts based on the sample requirement description information and the material library. These prompts instruct the generation of training material requirement information using the material library. For example, the requirement decomposition prompts could be in the form of "Please refer to the input sample requirement description information, separate the requirements, decompose and determine the corresponding processing requirements, and based on these processing requirements, determine the required material requirements that can be provided by the material library."

[0077] Then, the executing entity can execute the next step S303 to call the initial requirement processing model through the requirement decomposition prompt, so as to instruct the initial requirement processing model to decompose the sample requirement description information and generate training material requirement information.

[0078] S303, the initial requirement processing model is invoked through the requirement decomposition prompt to decompose the sample requirement description information and generate training material requirement information; S304, Invoke the initial demand processing model to extract training materials corresponding to the training material demand information from the material library; Specifically, in this step, as discussed above, the executing entity may choose to use the initial demand processing model to extract training materials corresponding to the training material demand information from the material library, so that the process of extracting training materials and subsequent target materials can be integrated into the training process of the initial demand processing model.

[0079] In some embodiments, to simplify the process, the above-mentioned requirement decomposition prompts may also include "instructions" for extracting training and target materials, so that the initial requirement processing model can be instructed to continuously generate (training) material requirement information and extract training and target materials accordingly with a single requirement decomposition prompt.

[0080] S305, Based on the degree of difference between the target material and the sample material, adjust the initial demand processing model in the direction of reducing the degree of difference to obtain the demand processing model.

[0081] This step is similar to the implementation process provided by S204 discussed above. For the same parts, please refer to the content discussed in S204 above. It will not be repeated here.

[0082] In some embodiments, requirement decomposition prompts can be configured in a default manner to omit information. For example, the requirement decomposition prompts for generating material requirement information to meet processing needs using a material library, or further, for extracting target material requirement information corresponding to the generated material requirement information from the material library, can all be configured in a default manner. Accordingly, in such cases, after receiving requirement description information, such as sample requirement description information, the executing entity can, based on this default configuration, naturally "understand" the material requirement information that needs to be met by utilizing the material library and perform such actions. This makes it easier to utilize the initial requirement processing model or the subsequently trained requirement processing model.

[0083] In some embodiments, to further enhance the processing capability of the demand processing model, the executing entity can, when training the initial demand processing model, also use sample context information associated with the sample demand description information as input, so that the initial demand processing model not only has the ability to "understand" the demand description information, but also has the ability to "understand" it by referring to the context of the demand description information. This enables the trained demand processing model to better understand the demand description information and generate material demand information by utilizing the context.

[0084] Accordingly, in such a case, the executing entity can choose to obtain sample context information associated with the sample requirement description information. Then, during the process of calling the initial requirement processing model to process the sample requirement description information and generate material requirement information to meet the processing requirements, the entity can choose to call the initial requirement processing model to process both the sample requirement description information and the sample context information to generate material requirement information to meet the processing requirements.

[0085] In some embodiments, if the demand processing model is also trained and configured with the ability to extract target material (training material), or even the ability to generate the final processing result file as discussed below, then in such a scenario, the initial demand processing model can also generate prompt information based on the reliability of its extraction of training material during the training process. For example, after extracting training materials, the (initial) demand processing model can evaluate the degree of matching between the training materials and the sample materials' demand information (e.g., determining the degree of matching based on the difference between the specific demand and the image semantics, text semantics, etc. of the training materials). If the degree of matching is less than or equal to a preset matching threshold, the (initial) demand processing model can "consider" it to be due to a lack of materials in the material library (or, in this case, it can be simply understood that even the most suitable and matching target material in the material library may not meet the requirements of the material demand information). In such a case, the (initial) demand processing model can generate prompts and feedback information to indicate and respond to the situation, so that the executing entity, or the user of the initial demand processing model, or the user of the trained demand processing model, can understand the situation, adjust the target material used, and update the materials in the material library.

[0086] S103, Extract the target material corresponding to the material requirement information from the material library; In the embodiments of this application, the executing entity can extract target materials that meet the material requirement information from the material library based on the material requirement information obtained in S102 above. For example, the executing entity can extract target materials that meet the material requirement information from the material library using methods such as semantic matching discussed above, which will not be repeated here.

[0087] S104, Generate target content based on target material.

[0088] In the embodiments of this application, after obtaining the target materials based on S103 described above, the executing entity can use these target materials to generate processing results. For example, the executing entity can combine these target materials based on the "outline" corresponding to the material requirement information by referring to the material requirement information to obtain the final processing result file.

[0089] In some embodiments, since such an "outline" is provided by the demand processing model, the ability to utilize these target materials can be further trained and encapsulated into the demand processing model during the training process described above. For example, after training the demand processing model to enable it to extract qualified target materials, the executing entity can further instruct the demand processing model to combine these target materials using material demand information through newly constructed prompts to obtain the corresponding processing result file.

[0090] For example, in the same scenario of generating a video to promote city A, the demand processing model can, based on its further trained and encapsulated generation capabilities, refer to the material demand information to insert various target materials (e.g., images of attractions in city A) and the generated text used to introduce the attractions, in order to generate the final processing result file, such as a promotional video for city A.

[0091] Therefore, in such a case, this step can actually be completed by the requirement processing model. Accordingly, in some embodiments, after the executing entity receives the requirement description information, it can choose to directly provide it to the requirement processing model, so that by calling and utilizing the requirement processing model, a processing result file corresponding to the requirement description information can be directly generated. In this way, the trained requirement processing model can comprehensively complete the process from processing the requirement description information to finally generating a processing result file that satisfies the requirement description information, enabling the requirement processing model to more conveniently and comprehensively meet the user's processing needs, such as those provided by the requirement description information.

[0092] The content generation method provided in this application first obtains the requirement description information of the target content to be generated; then, it decomposes the requirement description information into material requirement information through a requirement processing model. Specifically, the requirement processing model is trained based on the differences between training materials and sample materials. The training materials are then decomposed from the sample requirement description information using the initial requirement processing model to obtain training material requirement information, which is then extracted from a material library. Sample materials are used to meet the material requirements of the sample requirement description information and are also extracted from the material library. Next, target materials corresponding to the material requirement information are extracted from the material library. Finally, the target content to be generated is generated based on the target materials. Therefore, by utilizing a requirement processing model with stronger perception and material selection capabilities of the material library, it is possible to more accurately generate material requirement information from the content in the material library to obtain the materials needed to meet the user's processing needs, reducing model illusions and thus better completing the requirement processing and content generation tasks.

[0093] Based on any of the above embodiments, if the user or the maintainer of the demand processing model is satisfied or not satisfied with the target generated content, they can use the target device to return or provide preference information to the execution entity, and instruct the execution entity to make further adjustments to the demand processing model based on and using the preference information, so as to further improve the capability of the demand processing model according to their needs and preferences.

[0094] For example, after generating the target content, the executing entity can provide it to the target device used by the content requester (or user). Then, these users can use the target device to provide feedback on their preferences to the executing entity.

[0095] In some embodiments, preference information can be a user's description in natural language of their degree of liking and reasons for liking the overall or partial content of the target generated content. For example, when the target generated content is in the form of video, the user can indicate which video frames they like or dislike by defining a time range and selecting video frames. Accordingly, during this indication process, the user can also indicate their reasons for liking or disliking the content by providing specific reasons for liking or disliking, and describing the degree of liking or disliking.

[0096] Then, the executing entity can determine the preferred target material based on these "video frames". That is, the executing entity can use the material used to generate the video frame as the preferred target material, and use these preferred target materials to adjust the demand processing model.

[0097] For example, after acquiring preference information for content generated targeting a specific goal and determining the preferred target materials, the executing entity can determine the content that the user likes and / or dislikes based on the indicated content, and adjust the demand processing model accordingly. For instance, a new loss function can be constructed based on the content the user likes and the reasons for liking it, and this loss function can be used to iteratively train the demand processing model. This allows the demand processing model to be more inclined to generate target materials that utilize the user's preferred preferences, and the model can be iteratively optimized (or, in other words, optimized to reduce or avoid using target materials that the user dislikes or does not prefer). In this way, the demand processing model can be made more dynamic and flexible in aligning with the user's content generation preferences, improving its content generation capabilities.

[0098] Correspondingly, this approach also allows the demand processing model to be further adjusted in a preference-aligned manner. For example, adjustments can be made using the aforementioned preference target materials to ensure that the demand processing model has a better ability to decompose demands and generate material demand information for different users and different generation preferences.

[0099] To deepen understanding, this application also incorporates a specific application scenario, through... Figure 4 A flowchart of process 400 is provided for a specific implementation in this application scenario, including the training requirement processing model process and the content generation process.

[0100] To better illustrate the implementation of the various embodiments of this application, in this specific application scenario, we can discuss the case where an LLM is used as the initial demand processing model to train a demand processing model for extracting target materials.

[0101] In process 400, for LLM 410 (which can be understood as the “initial requirement processing model” mentioned above), after obtaining and being provided with sample requirement description information 421, it can perform requirement decomposition on the sample requirement description information 421 by executing S401 to generate training material requirement information 423 to meet the processing requirements of the sample requirement description information 421.

[0102] Then, in LLM 410, S402 can be further executed to extract training material 431 corresponding to training material requirement information 423 from material library 422.

[0103] Next, the execution entity, such as the server, which can call and train LLM 410 as discussed above, can execute S403 to adjust LLM 410 in the direction of reducing the degree of difference between training material 431 and sample material 424, so as to obtain the final demand processing model 415.

[0104] Then, after obtaining the demand processing model 415, if, for example, the execution entity of the server obtains the demand description information 440, it can execute S404 to decompose the demand description information 440 through the demand processing model 415 to obtain the material demand information 445.

[0105] Then, for example, the execution entity of the server can further extract the target material 450 corresponding to the material requirement information 445 from the material library 422 by executing S405 (it should be understood that there may be multiple target materials here, but for the sake of simplicity, only one target material is shown here, which does not mean that this application intends to limit the number generated).

[0106] Finally, for example, the server's execution entity can further generate target generated content 455 based on target material 450 by executing S406.

[0107] This application also provides an apparatus for generating content, the structure of which is as follows: Figure 5 The apparatus 500 shown includes: a requirement description acquisition module 510, configured to acquire requirement description information of the target generated content; a requirement description decomposition module 520, configured to decompose the requirement description information into material requirement information using a requirement processing model, wherein the requirement processing model is obtained by training an initial requirement processing model based on the difference between training materials and sample materials; the training materials are obtained by decomposing the sample requirement description information into requirement information using the initial requirement processing model, and then extracted from a material library based on the training material requirement information; sample materials are used to meet the material requirements of the sample requirement description information and are extracted from the material library; a material extraction module 530, configured to extract target materials corresponding to the material requirement information from the material library; and a content generation module 540, configured to generate target generated content based on the target materials.

[0108] In some embodiments, the apparatus 500 further includes: a preference information acquisition module configured to acquire preference information for target generated content; and a model preference adjustment module configured to adjust the demand processing model based on the preference information.

[0109] In some embodiments, the model preference adjustment module includes a preference material determination submodule, which is configured to determine the preference target material based on the semantic matching result between the target material and preference information; and the model preference adjustment submodule is configured to adjust the requirement processing model based on the preference target material.

[0110] In some embodiments, the apparatus 500 further includes: a sample acquisition module configured to acquire sample requirement description information, a material library, and sample materials; a training material requirement information generation module configured to call an initial requirement processing model to decompose the sample requirement description information into requirements and generate training material requirement information; a training material determination module configured to extract training materials corresponding to the training material requirement information from the material library; and an initial model adjustment module configured to adjust the initial requirement processing model based on the degree of difference between the sample materials and the training materials, in the direction of reducing the degree of difference, to obtain a requirement processing model.

[0111] In some embodiments, the training material requirement information generation module includes: a prompt word construction submodule, configured to construct requirement decomposition prompt words based on sample requirement description information and a material library, wherein the requirement decomposition prompt words indicate the use of the material library to generate training material requirement information; and a training material requirement information generation submodule, configured to call the initial requirement processing model through the requirement decomposition prompt words to decompose the sample requirement description information and generate training material requirement information.

[0112] In some embodiments, the apparatus 500 further includes: a sample context information acquisition module, configured to acquire sample context information associated with sample requirement description information; and a training material requirement information generation module, further configured to invoke an initial requirement processing model to perform requirement decomposition based on the sample context information and sample requirement description information to generate training material requirement information.

[0113] In some embodiments, the initial model adjustment module includes: a reward function construction submodule, configured to construct a reward function for the degree of difference between training materials and sample materials; and a reinforcement learning submodule, configured to use the reward function to adjust the initial demand processing model in a reinforcement learning manner in the direction of reducing the degree of difference, thereby obtaining a demand processing model.

[0114] In some embodiments, the reward function construction submodule is further configured to, in response to the requirement of training material demand information requiring at least two training materials, construct a reward function for the degree of difference between the training materials and the sample materials, and the degree of difference in material style between each training material; and the reinforcement learning submodule is further configured to, using the reward function, adjust the initial demand processing model in a reinforcement learning manner in the direction of reducing the degree of difference and the degree of difference in material style, to obtain a demand processing model.

[0115] In some embodiments, the reward function construction submodule is further configured to construct a reward function for the degree of difference between training materials and sample materials, and the amount of training material demand information; and the reinforcement learning submodule is further configured to use the reward function to adjust the initial demand processing model in a reinforcement learning manner in the direction of reducing the degree of difference and increasing the amount of training material demand information, thereby obtaining a demand processing model.

[0116] This device embodiment corresponds to the method embodiment shown in the figures above. This device can utilize a demand processing model with a stronger perception and selection capability of the material library to more accurately generate material demand information from the content of the material library to obtain the materials needed to meet the user's processing needs, reduce model illusion, and thus better complete the demand processing and content generation tasks.

[0117] Based on the same inventive concept, this application also provides an electronic device, a readable storage medium, and a computer program product. The method corresponding to the electronic device can be the content generation method in the foregoing embodiments, and its problem-solving principle is similar to that method. The electronic device provided in this application includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the methods and / or technical solutions of the foregoing embodiments of this application.

[0118] Electronic devices can be user devices, or devices composed of user devices and network devices integrated through a network, or applications running on the aforementioned devices. User devices include, but are not limited to, various terminal devices such as computers, mobile phones, tablets, smartwatches, and wristbands. Network devices include, but are not limited to, network hosts, single network servers, multiple network server sets, or cloud computing-based computer sets, and can be used to implement some processing functions when setting an alarm clock. Here, the cloud consists of a large number of hosts or network servers based on cloud computing. Cloud computing is a type of distributed computing, consisting of a virtual computer composed of a group of loosely coupled computer sets.

[0119] Figure 6 The diagram illustrates the structure of an electronic device 600 suitable for implementing the methods and / or technical solutions in the embodiments of this application. The electronic device 600 includes a Central Processing Unit (CPU) 601, which can perform various appropriate actions and processes based on a program stored in a Read Only Memory (ROM) 602 or a program loaded from a storage portion 608 into a Random Access Memory (RAM) 603. The RAM 603 also stores various programs and data required for system operation. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An Input / Output (I / O) interface 605 is also connected to the bus 604.

[0120] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, touchscreen, microphone, infrared sensor, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), LED display, OLED display, etc., and speakers, etc.; a storage section 608 including one or more computer-readable media such as hard disk, optical disk, magnetic disk, semiconductor memory, etc.; and a communication section 609 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet.

[0121] In particular, the methods and / or embodiments in this application can be implemented as computer software programs. For example, the embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. When the computer program is executed by a central processing unit (CPU) 601, it performs the functions defined in the methods of this application.

[0122] Another embodiment of this application provides a computer-readable storage medium and a computer program product having computer program instructions stored thereon, which can be executed by a processor to implement the methods and / or technical solutions of any one or more embodiments of this application described above.

[0123] Specifically, this embodiment may employ any combination of one or more computer-readable media. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example, a system, apparatus, or device that is, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0124] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0125] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0126] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0127] The flowcharts or block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-specific system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0128] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0129] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules and units is only a logical functional division, and in actual implementation, there may be other division methods. Taking units as examples, multiple units or page components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0130] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0131] Furthermore, the functional modules and units in the various embodiments of this application can be integrated into one processing module or unit, or each module or unit can exist physically separately, or two or more units can be integrated into one module or unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules and units.

[0132] The integrated modules and units implemented as software functional modules and units described above can be stored in a computer-readable storage medium. These software functional modules and units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

[0134] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a device claim may also be implemented by a single unit or device through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.

Claims

1. A method for generating content, characterized in that, include: Obtain the requirement description information for the target generated content; The requirement description information is decomposed using a requirement processing model to obtain material requirement information. The requirement processing model is trained based on the difference between training materials and sample materials to obtain an initial requirement processing model. The training materials are then used to decompose the sample requirement description information using the initial requirement processing model to obtain training material requirement information. The training material requirement information is then extracted from the material library based on the training material requirement information. The sample materials are used to meet the material requirements of the sample requirement description information and are extracted from the material library. Extract the target material corresponding to the material requirement information from the material library; Based on the target material, the target generated content is generated.

2. The method according to claim 1, characterized in that, Also includes: Obtain preference information for content generated for the target; Based on the preference information, the demand processing model is adjusted.

3. The method according to claim 2, characterized in that, The adjustment of the demand processing model based on the preference information includes: Based on the semantic matching results between the target material and the preference information, the preferred target material is determined; Based on the preferred target materials, adjust the demand processing model.

4. The method according to claim 1, characterized in that, The demand processing model was trained in the following manner: Obtain the sample requirement description information, the material library, and the sample materials; The initial requirement processing model is invoked to decompose the sample requirement description information to generate the training material requirement information. Extract training materials corresponding to the training material requirement information from the material library; Based on the degree of difference between the sample material and the training material, the initial demand processing model is adjusted in the direction of reducing the degree of difference to obtain the demand processing model.

5. The method according to claim 4, characterized in that, The step of calling the initial demand processing model to decompose the sample demand description information and generate the training material demand information includes: Based on the sample requirement description information and the material library, requirement decomposition prompt words are constructed, wherein the requirement decomposition prompt words indicate that the training material requirement information is generated using the material library; The initial requirement processing model is invoked by the requirement decomposition prompt to decompose the sample requirement description information and generate the training material requirement information.

6. The method according to claim 4, characterized in that, The method further includes: Obtain sample context information associated with the sample requirement description information; and The step of calling the initial demand processing model to decompose the sample demand description information and generate the training material demand information includes: The initial requirement processing model is invoked to decompose the requirements based on the sample context information and the sample requirement description information, thereby generating the training material requirement information.

7. The method according to claim 4, characterized in that, The step of adjusting the initial demand processing model based on the degree of difference between the sample material and the training material, in order to reduce the degree of difference, to obtain the demand processing model includes: Construct a reward function that reflects the degree of difference between the training material and the sample material; Using the reward function, the initial demand processing model is adjusted in a reinforcement learning manner to reduce the degree of difference, thus obtaining the demand processing model.

8. The method according to claim 7, characterized in that, The construction of the reward function based on the degree of difference between the training material and the sample material includes: In response to the requirement that at least two training materials are needed, a reward function is constructed based on the degree of difference between the training materials and the sample materials, and the degree of style difference between the training materials; and The step of using the reward function to adjust the initial demand processing model in a reinforcement learning manner to reduce the degree of difference, thereby obtaining the demand processing model, includes: Using the reward function, the initial demand processing model is adjusted in a reinforcement learning manner to reduce the degree of difference and the degree of difference in material style, thus obtaining the demand processing model.

9. The method according to claim 7, characterized in that, The construction of the reward function based on the degree of difference between the training material and the sample material includes: Construct a reward function that considers the degree of difference between the training material and the sample material, and the amount of information required by the training material; and The step of using the reward function to adjust the initial demand processing model in a reinforcement learning manner to reduce the degree of difference, thereby obtaining the demand processing model, includes: Using the reward function, the initial demand processing model is adjusted in a reinforcement learning manner to reduce the degree of difference and increase the amount of demand information in the training materials, thereby obtaining the demand processing model.

10. An apparatus for generating content, characterized in that, include: The requirement description acquisition module is configured to acquire the requirement description information of the target generated content. The requirement description decomposition module is configured to decompose the requirement description information into material requirement information using a requirement processing model. The requirement processing model is trained by creating an initial requirement processing model based on the differences between training materials and sample materials. The training materials are then used to decompose the sample requirement description information using the initial requirement processing model to obtain training material requirement information, which is then extracted from a material library. The sample materials are used to meet the material requirements of the sample requirement description information and are extracted from the material library. The material extraction module is configured to extract target materials corresponding to the material requirement information from the material library; The content generation module is configured to generate the target content based on the target material.

11. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 9.

12. A computer-readable medium, characterized in that, It stores computer program instructions that can be executed by a processor to implement the method as described in any one of claims 1 to 9.

13. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 9.

Images