Content generation optimization method and device, electronic equipment and storage medium
By dynamically adjusting the generation model through the generation and evaluation models in the collaborative architecture, the problem of low content generation quality is solved, thereby improving content quality.
Patent Information
- Application Number
- CN202511042715.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-28
- Publication Date
- 2025-11-25
AI Technical Summary
In existing content generation systems, the generation model cannot effectively adapt and adjust based on feedback from the evaluation model, resulting in low-quality generated content.
By using the generation and evaluation models in the collaborative architecture, conditional information is obtained to generate content and its evaluation value is evaluated. When the value exceeds the set range, it is converted into correction parameters to adjust the generation model until the evaluation value is within the range, thereby improving the content quality.
It effectively improves the content quality of the content generation system by dynamically adjusting the generation model to meet user needs and set standards.
Smart Images
Figure CN121009945A_ABST
Abstract
Description
Technical Field
[0001] The technical field of this disclosure, more specifically, relates to a content generation optimization method, apparatus, electronic device, and storage medium. Background Technology
[0002] With the rapid development of artificial intelligence technology, content generation technology is widely used in the creation of various media formats such as text, images, and videos. In particular, deep learning-based generative models can automatically generate content that meets user requirements based on the conditions provided by the user, thus improving the efficiency of content creation.
[0003] Currently, in existing technologies, content generation systems typically consist of two core components: a generation model and an evaluation model. The former is responsible for creating the content, while the latter is responsible for evaluating the quality of the generated content. However, the evaluation results of the generated content are difficult to effectively feed back into the generation model, preventing the generation model from adaptively adjusting based on the evaluation results. This results in lower quality content generated by the generation model. Summary of the Invention
[0004] One objective of this disclosure is to provide a new technical solution for optimizing content generation.
[0005] According to a first aspect of this disclosure, a content generation optimization method is provided, the method comprising:
[0006] Obtain conditional information for the target content generated by the user output; wherein, the conditional information includes a content type indicating the target content;
[0007] Using the first part of the model in the pre-set collaborative architecture, with the condition information as the generation target, the first generated content corresponding to the content type and the first evaluation value of the first generated content are obtained.
[0008] If the first evaluation value exceeds the set range, the first evaluation value is converted into a correction parameter and input into the first part of the model. Then, the first part of the model in the preset collaborative architecture is used as the generation target to obtain the first generated content corresponding to the content type and the first evaluation value of the first generated content, until the second evaluation value of the second generated content output by the first part of the model is within the set range.
[0009] The second generated content is sent to the user.
[0010] Optionally, the first part of the model includes a generation model and an evaluation model; the step of obtaining the first generated content corresponding to the content type and the first evaluation value of the first generated content by using the condition information as the generation target through the first part of the model in the preset collaborative architecture includes:
[0011] Using the aforementioned generation model and the conditional information as the generation target, the first generated content corresponding to the aforementioned content type is obtained;
[0012] The first generated content is input into the evaluation model to obtain the first evaluation value of the first generated content.
[0013] Optionally, the content type includes text, image, and video types; the step of obtaining the first generated content corresponding to the content type through the generation model, using the conditional information as the generation target, includes:
[0014] In the generative model, prompt words are extracted from the conditional information;
[0015] The content type corresponding to the generation target of the generation model is determined to be at least one of text type, image type and video type, and a first generated content corresponding to the content type is generated.
[0016] Optionally, inputting the first generated content into the evaluation model to obtain a first evaluation value for the first generated content includes:
[0017] In the evaluation model, the visual features of the first generated content are extracted to determine the aesthetics of the first generated content.
[0018] Extract harmful elements from the first generated content and determine the safety level of the first generated content;
[0019] Based on the weights set for aesthetics and security, a first evaluation value for the first generated content is obtained.
[0020] Optionally, the collaborative architecture further includes a control model; the step of converting the first evaluation value into correction parameters and inputting them into the first part of the model includes:
[0021] The first evaluation value is input into the control model to obtain the probability distribution parameters for adjusting the generation process of the generation model, the priority parameters for adjusting the internal modules of the generation model, and the guidance parameters for optimizing the input layer of the generation model. The probability distribution parameters, the priority parameters, and the guidance parameters are used as correction parameters.
[0022] The correction parameters are input into the generated model in the first part of the model.
[0023] Optionally, the collaborative architecture further includes a preference model; after obtaining the first generated content corresponding to the content type and the first evaluation value of the first generated content by using the condition information as the generation target through the first part of the model in the preset collaborative architecture, the method further includes:
[0024] By using the preference model, the user's behavioral data is obtained to determine the user's application scenario preferences and style preferences for the target content;
[0025] The first generated content is set to conform to the application scenario preference and the style preference to obtain the first generated content after setting.
[0026] Optionally, the collaborative architecture further includes an evolutionary model; after setting the first generated content to conform to the application scenario preferences and the style preferences, and obtaining the set first generated content, the method further includes:
[0027] The evolutionary model is used to obtain the user's historical generated tasks and the historical optimization paths of those tasks.
[0028] Based on the historical optimization paths, construct a general optimization path corresponding to the content type;
[0029] Based on the general optimization path, the first generated content is optimized to obtain the optimized first generated content.
[0030] According to a second aspect of this disclosure, a content generation optimization apparatus is also provided, the apparatus comprising:
[0031] The acquisition module is used to acquire conditional information of the target content generated by the user; wherein, the conditional information includes a content type indicating the target content;
[0032] The module is used to obtain, through the first part of the model in the preset collaborative architecture, with the condition information as the generation target, the first generated content corresponding to the content type and the first evaluation value of the first generated content;
[0033] The execution module is used to convert the first evaluation value into a correction parameter and input it into the first part of the model when the first evaluation value exceeds the set range, and continue to execute the steps of obtaining the first generated content corresponding to the content type and the first evaluation value of the first generated content by using the condition information as the generation target through the first part of the model in the preset collaborative architecture, until the second evaluation value of the second generated content output by the first part of the model is within the set range.
[0034] The sending module is used to send the second generated content to the user.
[0035] According to a third aspect of this disclosure, an electronic device is also provided, including a memory and a processor, the memory being used to store a computer program; the processor being used to execute the computer program to implement the method according to a first aspect of this disclosure.
[0036] According to a fourth aspect of this disclosure, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the method described according to a first aspect of this disclosure.
[0037] According to a fifth aspect of this disclosure, a computer program product is also provided, including a computer program that, when executed by a processor, implements the method described according to a first aspect of this disclosure.
[0038] One beneficial effect of this disclosure is that the content generation optimization method provided by the present invention can obtain the generated content of the required content type and the evaluation value of the generated content through the first part model in the collaborative architecture. When the evaluation value is low, it can be converted into corresponding correction parameters to correct the first part model, so that the evaluation value of the generated content subsequently output by the first part model can gradually be within the set range, so as to realize that the evaluation value acts on the first part model and effectively improve the content quality of the collaborative architecture.
[0039] Other features and advantages of the embodiments of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description
[0040] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments of the present disclosure and, together with their description, serve to explain the principles of the embodiments of the present disclosure.
[0041] Figure 1 This is a flowchart illustrating a content generation optimization method based on one embodiment;
[0042] Figure 2 This is a block diagram of an optimization device generated based on the content of one embodiment;
[0043] Figure 3 This is a schematic diagram of the hardware structure of an electronic device according to one embodiment. Detailed Implementation
[0044] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values of the parts and steps set forth in these embodiments do not limit the scope of the invention.
[0045] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.
[0046] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.
[0047] In all the examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.
[0048] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.
[0049] <Method Implementation>
[0050] Figure 1 This is a flowchart illustrating a content generation optimization method based on one embodiment. The implementing entity can be a personal computer, mobile phone, or server, etc., and is not limited thereto.
[0051] like Figure 2 As shown, the content generation optimization method of this embodiment may include the following steps S110 to S140:
[0052] Step S110: Obtain condition information of the target content generated by the user output; wherein, the condition information includes a content type indicating the target content.
[0053] In this embodiment, the content type can be text, image, or video. The condition information may include the user's specific needs, style preferences, and application scenarios, and may also include a Prompt instruction.
[0054] Step S120: Using the first part of the model in the preset collaborative architecture, and taking the condition information as the generation target, obtain the first generated content corresponding to the content type and the first evaluation value of the first generated content.
[0055] In this embodiment, the first part of the model may consist of one or more models. For example, the first part of the model may consist of a generation model and an evaluation model. The generation model is used to output the first generated content corresponding to the content type, and the evaluation model can evaluate the first generated content and obtain a first evaluation value.
[0056] In some embodiments, the first part of the model includes a generator (G) and an evaluator (E); step S120 may include the following steps S210 and S220:
[0057] Step S210: Using the generation model and the condition information as the generation target, the first generated content corresponding to the content type is obtained.
[0058] In this embodiment, the generative model can be configured with a generative artificial intelligence model (AIGC), such as a text generation model (e.g., GPT), an image generation model (e.g., GAN, Stable Diffusion), or a video generation model.
[0059] Step S220: Input the first generated content into the evaluation model to obtain the first evaluation value of the first generated content.
[0060] In this embodiment, the evaluation model can assess various dimensions of the generated content through deep learning and classification models to output a specific score, namely the first evaluation value of the first generated content.
[0061] In some embodiments, the content type includes text type, image type, and video type; step S220 may include the following steps S310 and S320:
[0062] Step S310: Extract prompt words from the condition information in the generated model.
[0063] Step S320: Determine that the content type corresponding to the generation target of the generation model is at least one of text type, image type and video type, and generate first generated content corresponding to the content type.
[0064] In this embodiment, when the prompts in the conditional information are plain text or contain text descriptions, the generation model can generate text. When the prompts in the conditional information are visual content (e.g., scene, object, or style), the generation model generates an image. When the prompts in the conditional information are a time sequence or video scene, the generation model generates a video.
[0065] In some examples, when the conditional information includes a requirement to combine images and text or video with text, the generative model can generate first generated content containing the corresponding content type. For example, if the conditional information includes instructions to generate an image and text of "white rabbit," the first generated content produced by the generative model could be an image and text of "white rabbit."
[0066] In this embodiment, by setting a generation model, it is possible to generate content that combines images and text or videos with text, effectively broadening the applicable scenarios of the collaborative architecture.
[0067] In some embodiments, step S220 may include the following steps S410 to S430:
[0068] Step S410: In the evaluation model, extract the visual features of the first generated content and determine the aesthetics of the first generated content.
[0069] In this embodiment, the evaluation model assesses the aesthetics of an image or video by extracting visual features (such as hue, composition, and details) of the first generated content. This evaluation model can be configured with models such as Convolutional Networks (CNNs) or Generative Adversarial Networks (GANs) to provide an aesthetics score based on the visual appeal of the first generated content.
[0070] Step S420: Extract harmful elements from the first generated content and determine the security level of the first generated content.
[0071] In this embodiment, the evaluation model can be configured with a content detection model (e.g., a Transformer model), which can analyze harmful elements (e.g., violence, pornography) in text, images, or videos. If the first generated content meets the safety standards, a high safety score is given; otherwise, a low safety score is given.
[0072] Step S430: Based on the weights set for aesthetics and security, the first evaluation value of the first generated content is obtained.
[0073] In this embodiment, the weights of aesthetics and security can both be 0.5, or the weight of aesthetics can be 0.6 and the weight of security can be 0.4. No limitation is made here.
[0074] In this embodiment, the first evaluation value = aesthetic score × aesthetic weight + security score × security weight. In other words, by setting an evaluation model to evaluate both aesthetics and security, the applicable scenarios of the collaborative architecture can be further broadened.
[0075] Step S130: If the first evaluation value exceeds the set range, the first evaluation value is converted into a correction parameter and input into the first part of the model. Then, the first part of the model in the preset collaborative architecture is used as the generation target to obtain the first generated content corresponding to the content type and the first evaluation value of the first generated content, until the second evaluation value of the second generated content output by the first part of the model is within the set range.
[0076] In this embodiment, the setting range can be set manually, and is not limited here.
[0077] In some embodiments, the collaborative architecture further includes a control model (Controller, C); step S130 may include the following steps S510 and S520:
[0078] Step S510: Input the first evaluation value into the control model to obtain the probability distribution parameters for adjusting the generation process of the generation model, the priority parameters for adjusting the internal modules of the generation model, and the guidance parameters for optimizing the input layer of the generation model, and use the probability distribution parameters, the priority parameters, and the guidance parameters as correction parameters.
[0079] In this embodiment, the control model can dynamically adjust the generation strategy based on the first evaluation value output by the evaluation model. Specifically, it adjusts the probability selection rules of the generation model's generation process to generate probability distribution parameters that balance the diversity (avoiding repetition) and fidelity (fitting the input logic) of the generated content. The control model can also determine the weights, or priority parameters, allocated to different functional modules within the generation model based on the first evaluation value output by the evaluation model, enabling the generation model to prioritize and strengthen specific features (such as style, semantics, and syntax). Furthermore, the control model can generate guidance parameters based on the first evaluation value output by the evaluation model to adjust the structure or logic of input prompts, thereby guiding the generation model to generate content that meets specific requirements.
[0080] Step S520: Input the correction parameters into the generated model in the first part of the model.
[0081] In this embodiment, by feeding back the correction parameters such as probability distribution parameters, priority parameters, and guidance parameters into the generative model, a personalized and adjustable feedback path is achieved, thereby making the generated content output by the generative model more in line with the user's needs.
[0082] In some embodiments, the collaborative architecture further includes a preference learner (P); after step S120, the method further includes the following steps S610 and S620:
[0083] Step S610: Obtain the user's behavioral data through the preference model, and determine the user's application scenario preferences and style preferences for the target content.
[0084] Step S620: Set the first generated content to conform to the application scenario preference and the style preference to obtain the set first generated content.
[0085] In this embodiment, the application scenarios are, for example, video creation, music editing, and advertising production. The application scenario preference can be one or more application scenarios that users frequently choose based on user behavior (such as viewing history, feedback, clicks, etc.).
[0086] In this embodiment, style preference can be one or more styles frequently selected by the user based on user behavior (such as viewing history, feedback, clicks, etc.), such as humorous style, lighthearted style, etc.
[0087] In some examples, when the first generated content is a short drama script, the user's preferred style is humor, and the preference model can set the first generated content to include more humorous elements in the short drama script.
[0088] In some examples, by setting up a preference model, it is possible to learn the user's application scenario preferences and style preferences for the target content, and provide personalized optimization references, so that the generated content output by the generation model is more in line with the user's needs.
[0089] In some embodiments, the collaborative architecture further includes an evolutionary module (M); after step S620, the method further includes steps S710 to S730:
[0090] Step S710: Obtain the user's historical generated tasks and the historical optimization paths of the historical generated tasks through the evolution model.
[0091] Step S720: Based on the historical optimization paths, construct a general optimization path corresponding to the content type.
[0092] Step S730: Based on the general optimization path, optimize the first generated content to obtain the optimized first generated content.
[0093] In some examples, when the evolutionary model handles multiple video generation tasks, each involving different styles of generated content, the model stores parameter adjustments and optimization trajectories from these different style generation tasks and constructs a general optimization path corresponding to the video type. When faced with a new style of generation task, the evolutionary model can quickly transfer previous optimization paths, thereby generating high-quality videos that conform to the new style in a short time.
[0094] In some examples, evolutionary models can store parameter tuning and optimization trajectories for different image enhancement tasks (such as denoising, super-resolution enhancement, etc.) and construct general optimization paths corresponding to image types. In new image enhancement tasks, the evolutionary model can quickly adjust the parameters of the generative model using these general optimization paths, effectively improving the quality of the generated images.
[0095] In this embodiment, the evolutionary model acquires the user's historical generation tasks and their optimized paths. These optimized paths can include information such as optimization trajectories and parameter adjustments. These optimized paths contain information on how to adjust the parameters and strategies of the generation model in different generation tasks. Based on these optimized paths, a general optimized path is obtained, enabling rapid adjustment and effective optimization when the collaborative architecture handles similar tasks.
[0096] Step S140: Send the second generated content to the user.
[0097] In this embodiment, the generated content of the required content type and the evaluation value of the generated content are obtained through the first part model in the collaborative architecture. When the evaluation value is low, it can be converted into corresponding correction parameters to correct the first part model, so that the evaluation value of the generated content output by the first part model can gradually be within the set range, so that the evaluation value can act on the first part model and effectively improve the content quality of the collaborative architecture.
[0098] <Equipment Example 1>
[0099] Figure 2 This is a schematic diagram of the optimized device generated based on the content of one embodiment. For example... Figure 2 As shown, the content generation optimization device 200 may include:
[0100] The acquisition module 210 is used to acquire conditional information of the target content generated by the user; wherein, the conditional information includes a content type indicating the target content;
[0101] The module 220 is used to obtain, through the first part of the model in the preset collaborative architecture, with the condition information as the generation target, the first generated content corresponding to the content type and the first evaluation value of the first generated content.
[0102] The execution module 230 is used to convert the first evaluation value into a correction parameter and input it into the first part of the model when the first evaluation value exceeds the set range, and continue to execute the steps of obtaining the first generated content corresponding to the content type and the first evaluation value of the first generated content by using the condition information as the generation target through the first part of the model in the preset collaborative architecture, until the second evaluation value of the second generated content output by the first part of the model is within the set range.
[0103] The sending module 240 is used to send the second generated content to the user.
[0104] In some embodiments, the obtaining module 220 is further configured to obtain a first generated content corresponding to the content type by using the condition information as the generation target through the generation model; and input the first generated content into the evaluation model to obtain a first evaluation value of the first generated content.
[0105] In some embodiments, the obtaining module 220 is further configured to extract prompt words from the condition information in the generation model; determine that the content type corresponding to the generation target of the generation model is at least one of text type, image type and video type, and generate first generated content corresponding to the content type.
[0106] In some embodiments, the obtaining module 220 is further configured to extract visual features of the first generated content in the evaluation model to determine the aesthetics of the first generated content; extract harmful elements of the first generated content to determine the safety of the first generated content; and obtain a first evaluation value of the first generated content according to the weights set for the aesthetics and the safety, respectively.
[0107] In some embodiments, the execution module 230 is further configured to input the first evaluation value into the control model to obtain the probability distribution parameters for adjusting the generation process of the generation model, the priority parameters for adjusting the internal modules of the generation model, and the guidance parameters for optimizing the input layer of the generation model, and to use the probability distribution parameters, the priority parameters, and the guidance parameters as correction parameters; and to input the correction parameters into the generation model in the first part of the model.
[0108] In some embodiments, the content generation optimization device 200 further includes a setting module, which is used to obtain the user's behavioral data through the preference model, determine the user's application scenario preference and style preference for the target content, and set the first generated content to conform to the application scenario preference and style preference to obtain the set first generated content.
[0109] In some embodiments, the content generation optimization device 200 further includes an optimization module, configured to obtain the user's historical generation tasks and the historical optimization paths of the historical generation tasks through the evolution model; construct a general optimization path corresponding to the content type based on the historical optimization paths; and optimize the first generated content based on the general optimization paths to obtain optimized first generated content.
[0110] <Equipment Example 2>
[0111] Figure 3 This is a schematic diagram of the hardware structure of an electronic device according to another embodiment.
[0112] like Figure 3 As shown, the electronic device 300 includes a processor 310 and a memory 320, the memory 320 being used to store an executable computer program, and the processor 310 being used to execute methods as described in any of the above method embodiments under the control of the computer program.
[0113] The modules of the above-mentioned optimization device 200 can be implemented by the processor 310 in this embodiment executing the computer program stored in the memory 320, or they can be implemented by other structures, which are not limited here.
[0114] This invention can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the invention.
[0115] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.
[0116] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.
[0117] The computer program instructions used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as the "C" language or similar programming languages. The computer-readable program instructions may 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 a remote computer, the remote computer may 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 may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions. This electronic circuitry can execute the computer-readable program instructions to implement various aspects of the invention.
[0118] Various aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.
[0119] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.
[0120] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.
[0121] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive 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-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. It will be known to those skilled in the art that implementation in hardware, implementation in software, and implementation using a combination of software and hardware are equivalent.
[0122] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, and are not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein. The scope of the invention is defined by the appended claims.
Claims
1. A content generation optimization method, characterized in that, The method includes: Obtain conditional information for the target content generated by the user output; wherein, the conditional information includes a content type indicating the target content; Using the first part of the model in the pre-set collaborative architecture, with the condition information as the generation target, the first generated content corresponding to the content type and the first evaluation value of the first generated content are obtained. If the first evaluation value exceeds the set range, the first evaluation value is converted into a correction parameter and input into the first part of the model. Then, the first part of the model in the preset collaborative architecture is used as the generation target to obtain the first generated content corresponding to the content type and the first evaluation value of the first generated content, until the second evaluation value of the second generated content output by the first part of the model is within the set range. The second generated content is sent to the user.
2. The method according to claim 1, characterized in that, The first part of the model includes a generation model and an evaluation model; the step of obtaining the first generated content corresponding to the content type and the first evaluation value of the first generated content by using the condition information as the generation target through the first part of the model in the preset collaborative architecture includes: Using the aforementioned generation model and the conditional information as the generation target, the first generated content corresponding to the aforementioned content type is obtained; The first generated content is input into the evaluation model to obtain the first evaluation value of the first generated content.
3. The method according to claim 2, characterized in that, The content types include text, image, and video types; the step of obtaining the first generated content corresponding to the content type through the generation model, using the conditional information as the generation target, includes: In the generative model, prompt words are extracted from the conditional information; The content type corresponding to the generation target of the generation model is determined to be at least one of text type, image type and video type, and a first generated content corresponding to the content type is generated.
4. The method according to claim 2, characterized in that, The step of inputting the first generated content into the evaluation model to obtain a first evaluation value for the first generated content includes: In the evaluation model, the visual features of the first generated content are extracted to determine the aesthetics of the first generated content. Extract harmful elements from the first generated content and determine the safety level of the first generated content; Based on the weights set for aesthetics and security, a first evaluation value for the first generated content is obtained.
5. The method according to claim 2, characterized in that, The collaborative architecture further includes a control model; the step of converting the first evaluation value into correction parameters and inputting them into the first part of the model includes: The first evaluation value is input into the control model to obtain the probability distribution parameters for adjusting the generation process of the generation model, the priority parameters for adjusting the internal modules of the generation model, and the guidance parameters for optimizing the input layer of the generation model. The probability distribution parameters, the priority parameters, and the guidance parameters are used as correction parameters. The correction parameters are input into the generated model in the first part of the model.
6. The method according to claim 1, characterized in that, The collaborative architecture further includes a preference model; after obtaining the first generated content corresponding to the content type and the first evaluation value of the first generated content by using the condition information as the generation target through the first part of the model in the preset collaborative architecture, the method further includes: By using the preference model, the user's behavioral data is obtained to determine the user's application scenario preferences and style preferences for the target content; The first generated content is set to conform to the application scenario preference and the style preference to obtain the first generated content after setting.
7. The method according to claim 6, characterized in that, The collaborative architecture also includes an evolutionary model; after setting the first generated content to conform to the application scenario preferences and the style preferences, and obtaining the set first generated content, the method further includes: The evolutionary model is used to obtain the user's historical generated tasks and the historical optimization paths of those tasks. Based on the historical optimization paths, construct a general optimization path corresponding to the content type; Based on the general optimization path, the first generated content is optimized to obtain the optimized first generated content.
8. A content generation optimization apparatus, characterized in that, The device includes: An acquisition module is used to acquire conditional information of the target content generated by the user; wherein, the conditional information includes a content type indicating the target content; The module is used to obtain, through the first part of the model in the preset collaborative architecture, with the condition information as the generation target, the first generated content corresponding to the content type and the first evaluation value of the first generated content; The execution module is used to convert the first evaluation value into a correction parameter and input it into the first part of the model when the first evaluation value exceeds the set range, and continue to execute the steps of obtaining the first generated content corresponding to the content type and the first evaluation value of the first generated content by using the condition information as the generation target through the first part of the model in the preset collaborative architecture, until the second evaluation value of the second generated content output by the first part of the model is within the set range. The sending module is used to send the second generated content to the user.
9. An electronic device, characterized in that, The system includes a memory and a processor, the memory being used to store a computer program; the processor being used to execute the computer program to implement the method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the method according to any one of claims 1 to 7.