Content generation method and related equipment

Abstract

The invention provides a content generation method and related equipment, and the method comprises the steps: a server determines the type of a task according to request information; splitting the task into a plurality of task flows according to the type of the task; according to the execution sequence of the multiple task processes, based on models matched with all the task processes in a model combination, the multiple task processes are executed, content corresponding to the request information is generated, and the model combination corresponds to the types of the tasks. In the embodiment of the invention, each task process is executed based on the matched model, so that the content conforming to the user expectation can be generated, and the quality of the generated content is improved.

Description

Technical Field

[0001] This application relates to the technical field of servers, and in particular, to a content generation method and related devices. Background Art

[0002] Automated content generation technology refers to the technology of automatically generating various types of content such as text, images, and videos using artificial intelligence technology (AI).

[0003] In the traditional automated content generation process, content can be generated based on a single general model according to user requirements. However, due to the limited capabilities of a single general model, when relying on a single general model for content generation currently, it is easily restricted by its inherent capabilities, resulting in the generated content being difficult to meet the actual needs of users and having poor quality.

[0004] Based on this, there is an urgent need for a solution to solve the above technical problems. Summary of the Invention

[0005] A content generation method and related devices provided by an embodiment of this application can utilize multiple models to collaborate in processing tasks and improve task processing efficiency.

[0006] The embodiments of this application adopt the following technical solutions:

[0007] First aspect: An embodiment of this application provides a content generation method. By determining the type of a task according to request information; splitting the task into multiple task processes according to the type of the task; and executing the multiple task processes based on the models in the model combination that match each task process in the execution order of the multiple task processes to generate content corresponding to the request information, where the model combination corresponds to the type of the task.

[0008] In the embodiment of this application, the server can split a task into multiple task processes and execute each task process based on a model combination that matches the type of the task to generate content corresponding to the request information. Since in the embodiment of this application, each task process is executed based on a model that matches it, content that meets user expectations can be generated, improving the quality of the generated content. In the embodiment of this application, the method of using multiple models to collaborate in executing tasks to generate content corresponding to the request information can effectively improve the efficiency of content generation.

[0009] In a possible implementation manner, the steps for generating the model combination include: determining the models that match each task process based on the multiple task processes corresponding to the type of the task; and performing an integration process on the models that match each task process to obtain the model combination.

[0010] In the embodiments of the present application, integrating the models matching each task process can obtain a model combination. Based on the models in the model combination collaborating to execute tasks and generating the content corresponding to the request information, the quality of the generated content can be improved, and at the same time, the efficiency of content generation can be effectively provided.

[0011] In a possible implementation manner, when the type of the task is code writing, the model combination is a code writing model combination. According to the type of the task, the task is split into multiple task processes, including: splitting the task into a requirements analysis process, a function design process, a code writing process, a code testing process, a code correction process, and a document writing process according to the type of the task.

[0012] According to the execution order of the multiple task processes, based on the models in the model combination that match each task process, execute the multiple task processes to generate the content corresponding to the request information, including: according to the execution order of the requirements analysis process, the function design process, the code writing process, the code testing process, the code correction process, and the document writing process, based on the requirements analysis model, the function design model, the code writing model, the code testing model, the code correction model, and the document writing model in the code writing model combination, respectively execute the requirements analysis process, the function design process, the code writing process, the code testing process, the code correction process, and the document writing process to generate the code file corresponding to the request information.

[0013] In the embodiments of the present application, based on the models corresponding to each of the multiple task processes in the task of code writing, each task process is executed, improving the quality of the generated content and enabling the obtained code file to meet the user's expectations.

[0014] In a possible implementation manner, according to the execution order of the requirement analysis process, function design process, code writing process, code testing process, code correction process, and document writing process, based on the requirement analysis model, function design model, code writing model, code testing model, code correction model, and document writing model in the code writing model combination, respectively execute the requirement analysis process, function design process, code writing process, code testing process, code correction process, and document writing process to generate a code file corresponding to the request information, including: according to the execution order of the requirement analysis process, function design process, code writing process, code testing process, code correction process, and document writing process, based on the requirement analysis model in the code writing model combination, execute the requirement analysis process, perform data analysis on the request information to obtain an analysis result; input the analysis result into the function design model to execute the function design process to determine the requirements for the steps of writing code; input the requirements for the steps of writing code into the code writing model to execute the code writing process to obtain code; based on the code testing model, execute the code testing process to perform code testing on the code to obtain a test result; input the test result into the code correction model to execute the code correction process to obtain the corrected code; input the corrected code into the document writing model to execute the document writing process to annotate and summarize the corrected code to generate a code file corresponding to the request information.

[0015] In the embodiment of the present application, each task process is executed based on the models corresponding to the requirement analysis process, function design process, code writing process, code testing process, code correction process, and document writing process respectively, improving the quality of the generated content and enabling the obtained code file to meet the user's expectations.

[0016] In a possible implementation manner, when the type of the task is file recognition, the model combination is a file recognition model combination. According to the type of the task, the task is split into multiple task processes, including: according to the type of the task, the task is split into a security detection process and a label recognition process.

[0017] According to the execution order of the multiple task processes, based on the models in the model combination that match each task process, execute the multiple task processes to generate the content corresponding to the request information, including: according to the execution order of the security detection process and the label recognition process, based on the first security detection model and the label recognition model in the file recognition model combination, respectively execute the security detection process and the label recognition process to generate the content corresponding to the request information.

[0018] In the embodiment of the present application, each task process is executed based on the models corresponding to the multiple task processes of the file recognition task, effectively improving the quality of the generated content and enabling the generated content to meet the user's expectations.

[0019] In a possible implementation, according to the execution order of the security detection process and the label recognition process, based on the first security detection model and the label recognition model in the file recognition model combination, the security detection process and the label recognition process are respectively executed to generate the content corresponding to the request information, including: according to the execution order of the security detection process and the label recognition process, based on the first detection model in the first security detection model, the first detection process in the security detection process is executed to perform security detection on the request information to obtain the first security detection result; in the case where the first security detection result is qualified, based on the label recognition model, the label recognition process is executed to perform label recognition on the request information to obtain the recognition result; based on the second detection model in the first security detection model, the second detection process in the security detection process is executed to perform security detection on the recognition result to obtain the second security detection result; in the case where the second security detection result is qualified, the recognition result is used as the content corresponding to the request information.

[0020] In the embodiments of the present application, each task process is executed based on the models corresponding to the security detection process and the label recognition process respectively, effectively improving the quality of the generated content, so that the generated content can meet the user's expectations. At the same time, the security of the generated content can be improved through the security detection model.

[0021] In a possible implementation, based on the label recognition model, the label recognition process is executed to perform label recognition on the request information to obtain the recognition result, including: based on the first recognition model in the label recognition model, the first recognition process in the label recognition process is executed to recognize the sending unit in the request information to determine the sending unit; based on the second recognition model in the label recognition model, the second recognition process in the label recognition process is executed to recognize the main recipient and / or carbon copy recipient unit in the request information to determine the main recipient and / or carbon copy recipient unit; based on the third recognition model in the label recognition model, the third recognition process in the label recognition process is executed to recognize the case to be handled and / or the document to be read in the request information to determine the case to be handled and / or the document to be read; based on the fourth recognition model in the label recognition model, the fourth recognition process in the label recognition process is executed to recognize the business classification department in the request information to determine the business classification department; at least one of the sending unit, the main recipient and / or carbon copy recipient unit, the case to be handled and / or the document to be read, and the business classification department is used as the recognition result.

[0022] In the embodiments of the present application, each task process is executed based on the models corresponding to the security detection process and the label recognition process respectively, effectively improving the quality of the generated content, so that the generated content can meet the user's expectations.

[0023] In a possible implementation, when the type of the task is text generation, the model combination is a text generation model combination. According to the type of the task, the task is split into multiple task processes, including: splitting the task into a security detection process and a text generation process according to the type of the task.

[0024] According to the execution order of the multiple task processes, based on the models in the model combination that match each task process, execute the multiple task processes to generate the content corresponding to the request information, including: according to the execution order of the security detection process and the text generation process, based on the second security detection model and the text generation model in the text generation model combination, execute the security detection process and the text generation process respectively to generate the text content corresponding to the request information.

[0025] In the embodiments of the present application, each task process is executed based on the models corresponding to the multiple task processes in the text generation task, effectively improving the quality of the generated content and enabling the generated content to meet the user's expectations. At the same time, the security of the generated text content can be improved through the security detection model.

[0026] In a possible implementation, according to the execution order of the security detection process and the text generation process, based on the second security detection model and the text generation model in the text generation model combination, execute the security detection process and the text generation process respectively to generate the text content corresponding to the request information, including: according to the execution order of the security detection process and the text generation process, based on the third detection model in the second security detection model in the text generation model combination, execute the third detection process in the security detection process to perform security detection on the request information to obtain a third security detection result; in the case where the third security detection result is qualified, execute the text generation process based on the text generation model to generate initial text content; based on the fourth detection model in the second security detection model in the text generation model combination, execute the fourth detection process in the security detection process to perform security detection on the initial text content to obtain a fourth security detection result; in the case where the fourth security detection result is qualified, use the initial text content as the text content corresponding to the request information.

[0027] In the embodiments of the present application, each task process is executed based on the models corresponding to the security detection process and the text generation process respectively, effectively improving the quality of the generated content and enabling the generated content to meet the user's expectations.

[0028] Second aspect: The embodiments of the present application provide a content generation device, including: a determination unit, a splitting unit, and a generation unit;

[0029] Among them, the determination unit is used to determine the type of the task by according to the request information;

[0030] A splitting unit, configured to split the task into multiple task processes according to the type of the task;

[0031] A generating unit, configured to execute the multiple task processes according to the execution order of the multiple task processes, based on the models in the model combination that match the respective task processes, and generate the content corresponding to the request information, where the model combination corresponds to the type of the task.

[0032] In the device provided in the embodiment of the present application, each task process is executed based on the model that matches it, and content that meets the user's expectations can be generated, improving the quality of the generated content. At the same time, in the embodiment of the present application, the method of using multiple models to cooperate to execute tasks and generate the content corresponding to the request information can effectively improve the efficiency of content generation.

[0033] In a third aspect: An embodiment of the present application provides a computing device, where the computing device includes: a processor and a memory;

[0034] The memory is configured to store program code and transmit the program code to the processor;

[0035] The processor is configured to execute the steps of a content generation method as described in the first aspect above according to the instructions in the program code.

[0036] In a fourth aspect: An embodiment of the present application provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of a content generation method as described in the first aspect above are implemented. Description of the Drawings

[0037] Figure 1 It is a schematic diagram of an application scenario of a content generation method provided by an embodiment of the present application;

[0038] Figure 2 It is a flowchart of a content generation method provided by an embodiment of the present application;

[0039] Figure 3 It is a flowchart of a method for generating a code file provided by an embodiment of the present application;

[0040] Figure 4 It is a flowchart of a method for file recognition provided by an embodiment of the present application;

[0041] Figure 5 It is a flowchart of a method for text generation provided by an embodiment of the present application;

[0042] Figure 6 It is a schematic diagram of a content generation method based on a model combination provided by an embodiment of the present application;

[0043] Figure 7 Schematic structural diagram of a content generation device provided by an embodiment of the present application;

[0044] Figure 8 Schematic structural diagram of a computing device provided by an embodiment of the present application. Detailed implementation manners

[0045] Terms such as "first", "second", and "third" in the specification, claims, and accompanying drawings of the present application are used to distinguish different objects, rather than to limit a specific order.

[0046] In the embodiments of the present application, words such as "exemplary" or "for example" are used to represent examples, illustrations, or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary" or "for example" is intended to present relevant concepts in a specific manner.

[0047] The embodiments of the present application provide a content generation method. By according to request information, the type of a task can be determined, and according to the type of the task, the task is split into multiple task processes. After determining the multiple task processes, in accordance with the execution order of the multiple task processes, based on the models in the model combination that match each task process, the multiple task processes are executed, and the content corresponding to the request information can be generated. Among them, the model combination corresponds to the type of the task. In the embodiments of the present application, since each task process is executed based on the model that matches it, content that meets the user's expectations can be generated, improving the quality of the generated content.

[0048] The following combines Figure 1 to introduce the application scenario of a content generation method provided by the present application.

[0049] Figure 1 Schematic diagram of an application scenario of a content generation method provided by an embodiment of the present application. In the figure, the interaction between server 1101 and client 1102 is taken as an example for introduction.

[0050] It should be noted that the client 1102 can be deployed in the server 1101 or in other computing devices. The embodiments of the present application do not make a specific limitation on the deployment location of the client 1102. The server 1101 can be a cabinet server, a rack server, a blade server, etc., or a general-purpose server, a GPU server, a DPU server, etc., and is not limited herein.

[0051] In the embodiments of the present application, the server 1101 can pre-generate a corresponding model combination for various types of tasks, so that subsequently, the multiple task processes obtained by splitting can be executed based on the model combination corresponding to the type of the task.

[0052] Exemplarily, in the process of generating a model combination, for each type of task, the server 1101 may determine models matching each task process based on multiple task processes corresponding to the type of the task, and obtain a model combination corresponding to the type of each task by integrally processing the models matching each task process.

[0053] On this basis, the user can input request information by interacting with the client 1102. After obtaining the request information, the client 1102 sends the request information to the server 1101.

[0054] After receiving the request information, the server 1101 may determine the type of the task according to the request information; split the task into multiple task processes according to the type of the task; and execute the multiple task processes based on the models matching the respective task processes in the model combination in the execution order of the multiple task processes to generate the content corresponding to the request information. Among them, the model combination corresponds to the type of the task.

[0055] After the server 1101 generates the content corresponding to the request information based on the model combination, it may send the content corresponding to the request information to the client 1102. After receiving the content corresponding to the request information, the client 1102 may display it for the user to view or use.

[0056] Next, a content generation method provided in an embodiment of the present application will be introduced.

[0057] As Figure 2 shown, the figure is a flowchart of a content generation method provided in an embodiment of the present application, including S201 - S203.

[0058] S201. The server determines the type of the task according to the request information.

[0059] In an embodiment of the present application, the request information may include the requirement information proposed by the user, or include the requirement information proposed by the user and the type of the task.

[0060] In an example, the request information includes the requirement information proposed by the user. For example, the requirement information proposed by the user may include requirement information for writing code, requirement information for file recognition, requirement information for text generation, etc.

[0061] When the request information is the requirement information for writing code, the server can determine the type of the task as writing code according to the requirement information for writing code; when the request information is the requirement information for file recognition, the server can determine the type of the task as file recognition according to the requirement information for file recognition; when the request information is the requirement information for text generation, the server can determine the type of the task as text generation according to the requirement information for text generation.

[0062] In one example, the request information includes the requirement information proposed by the user and the type of the task. In this case, the server can directly determine the type of the task based on the type of the task in the request information.

[0063] For example, the user can input the requirement information and the type of the task through the client. The client packages the requirement information and the type of the task input by the user into the request information and sends it to the server. After receiving the request information sent by the client, the server can directly obtain the type of the task by parsing the request information.

[0064] It can be understood that the specific content and form of the request information in the embodiments of the present application are not specifically limited. For example, the form of the request information includes voice form, text form, etc.

[0065] S202. The server splits the task into multiple task processes according to the type of the task.

[0066] In the embodiments of the present application, for tasks of different types, the server can split the task into multiple task processes according to the execution logic of the task.

[0067] Exemplarily, when the type of the task is writing code, the server can split the task into a requirement analysis process, a function design process, a code writing process, a code testing process, a code correction process, and a document writing process; when the type of the task is file recognition, the server can split the task into a security detection process and a label recognition process; when the type of the task is text generation, the server can split the task into a security detection process and a text generation process.

[0068] S203. The server executes multiple task processes based on the models in the model combination that match each task process according to the execution order of the multiple task processes, and generates the content corresponding to the request information.

[0069] In the embodiments of the present application, the model combination corresponds to the type of the task, and the model combination is obtained by integrating the models that match each task process. Thus, each task process of the task can be executed by the model that matches it.

[0070] Among them, each model can be called an agent. In the embodiments of the present application, each task process is executed based on the corresponding agent, and the agents matching each task process coordinate and cooperate with each other to jointly complete the task and generate the content corresponding to the request information.

[0071] In the embodiments of the present application, the server can pre-generate various model combinations corresponding to the types of each task.

[0072] In the process of generating various model combinations, the server can determine the models matching each task process based on multiple task processes corresponding to the type of the task. On this basis, the models matching each task process are integrated to obtain a model combination.

[0073] In a possible implementation manner, when the task type is code writing, the task processes corresponding to the task include a requirements analysis process, a function design process, a code writing process, a code testing process, a code correction process, and a document writing process. According to the capabilities of various models and the characteristics of each task process, a large semantic recognition model with strong text writing ability can be used to execute the requirements analysis process, the function design process, and the document writing process, and a large code model with strong code writing ability can be used to execute the code writing process, the code testing process, and the code correction process.

[0074] In an example, at least two of the requirements analysis process, the function design process, and the document writing process can be executed using the same model.

[0075] In another example, to improve the task execution effect, the requirements analysis process, the function design process, and the document writing process can be executed using different large models with strong text writing ability respectively.

[0076] Similarly, at least two of the code writing process, the code testing process, and the code correction process can be executed using the same model; or, at least two of the code writing process, the code testing process, and the code correction process can be executed using different large models with strong code writing ability respectively.

[0077] Thus, when the task type is code writing, the model combination corresponding to this task type is a code writing model combination, which includes models for executing each task process. For example, a requirements analysis model for executing the requirements analysis process, a function design model for executing the function design process, a code writing model for executing the code writing process, a code testing model for executing the code testing process, a code correction model for executing the code correction process, and a document writing model for executing the document writing process.

[0078] The server sequentially executes each task process based on the code writing model combination in the order of the requirement analysis process, function design process, code writing process, code testing process, code correction process, and document writing process, and generates a code file corresponding to the request information.

[0079] In a possible implementation manner, when the task type is file recognition, the task process corresponding to the task includes a security detection process and a label recognition process. According to the capabilities of various models and the characteristics of each task process, a model with strong security detection capabilities can be used to execute the security detection process, and a model with strong label recognition capabilities can be used to execute the label recognition process.

[0080] Thus, when the task type is file recognition, the model combination corresponding to this task type is a file recognition model combination, which includes models for executing each task process. For example, a first security detection model for executing the security detection process and a label recognition model for executing the label recognition process.

[0081] The server executes each task process based on the file recognition model combination in the execution order of the security detection process and the label recognition process, and generates the content corresponding to the request information.

[0082] In a possible implementation manner, when the task type is text generation, the task process corresponding to the task includes a security detection process and a text generation process. According to the capabilities of various models and the characteristics of each task process, a model with strong security detection capabilities can be used to execute the security detection process, and a model with strong text generation capabilities can be used to execute the text generation process.

[0083] Thus, when the task type is text generation, the model combination corresponding to this task type is a text generation model combination, which includes models for executing each task process. For example, a second security detection model for executing the security detection process and a text generation model for executing the text generation process.

[0084] The server executes each task process based on the text generation model combination in the execution order of the security detection process and the text generation process, and generates the text content corresponding to the request information.

[0085] In summary, in the embodiments of the present application, the server can split a task into multiple task processes, execute each task process based on a model combination matching the type of the task, and generate the content corresponding to the request information. Since in the embodiments of the present application, each task process is executed based on a model matching it, content meeting the user's expectations can be generated, and the quality of the generated content can be improved. At the same time, in the embodiments of the present application, the method of using multiple models to cooperate to execute tasks and generate the content corresponding to the request information can effectively improve the efficiency of content generation.

[0086] For ease of understanding, the content generation method provided in the embodiments of the present application will be introduced below by taking the types of tasks as code writing, document recognition, and text generation as examples respectively.

[0087] A content generation method provided in the embodiments of the present application can generate a code file that meets the user's requirements based on a code writing model combination, such as Figure 3 shown, this figure is a flowchart of a method for generating a code file provided in the embodiments of the present application.

[0088] S301. The server obtains the request information input by the user.

[0089] In the embodiments of the present application, the generated code file may include, but is not limited to, a web page code file, a game code file, etc. Here, only the code file of the web page file is taken as an example for introduction.

[0090] Exemplarily, in the case where the user expects to write a business trip approval page, the request information input by the user is as follows:

[0091] "Create a business trip approval page with the following functions:

[0092] 1. The page can fill in the name, department, and project name;

[0093] 2. The page has a new button, supporting the addition of the departure place, destination, departure time, etc.;

[0094] 3. The page contains elements such as approvers and carbon copy recipients;

[0095] Please provide the core logic code of the page and necessary annotations."

[0096] It should be noted that only the business trip approval page is taken as an example for introduction here, and the above request information input by the user is only an example.

[0097] S302. The server determines that the type of the task is code writing according to the request information.

[0098] S303. The server splits the task into a requirements analysis process, a function design process, a code writing process, a code testing process, a code correction process, and a document writing process according to the type of the task.

[0099] S304. The server executes the requirements analysis process based on the requirements analysis model in the code writing model combination, performs data analysis on the request information, and obtains an analysis result.

[0100] In the embodiments of the present application, the server can interpret and analyze the request information input by the user based on the requirement analysis model and give the analysis result. For example, the page requirements include name, department, and project name; custom addition of departure location, destination, and departure time is supported; functions such as selecting approvers and carbon copy recipients are available.

[0101] Among them, the requirement analysis model can include but is not limited to the Qwen2.5-7B-Instruct model. The Qwen2.5-7B-Instruct model is an open-source model with powerful natural language processing capabilities, having capabilities in aspects such as language understanding, task adaptability, and multilingual support. At the same time, it also has a certain long text processing ability and is applicable to various natural language processing tasks. By adopting a suitable inference acceleration framework and fine-tuning technology, the performance and accuracy of the model can be further improved.

[0102] S305. The server inputs the analysis result into the function design model to execute the function design process and determine the requirements for the steps of writing code.

[0103] In the embodiments of the present application, when the server inputs the analysis result into the function design model, it can interpret the analysis result based on the function design model and output the requirements for the steps of writing code, such as the relevant function implementation steps and relevant running dependencies and environments.

[0104] Exemplarily, based on the function design model, it can be determined to implement using at least one coding language such as HyperText Markup Language (HTML), Cascading Style Sheets (CSS), and JavaScript (JS). The different page functions are split, for example, elements such as name, department, and project name are designed in the basic information area; the departure location, destination, and departure time are designed in the itinerary area; the approvers and carbon copy recipients are designed in the approval area, and each area is assembled into a complete business travel approval page.

[0105] Among them, the function design model can include but is not limited to the Qwen2.5-7B-Instruct model.

[0106] S306. The server inputs the requirements for the steps of writing code into the code writing model to execute the code writing process and obtains the code.

[0107] In the embodiments of the present application, when the server inputs the requirements for the steps of writing code into the code writing model, the code writing model can successively write the complete code to implement the web page function.

[0108] Among them, the code can be in Markdown format. Markdown is a lightweight markup language with the characteristics of being lightweight, easy to read and write, etc.

[0109] The code writing model can include but is not limited to the Qwen2.5-Coder-7B-Instruct model, which has strong code writing capabilities.

[0110] S307. The server executes the code testing process based on the code testing model, conducts code testing on the code, and obtains the test results.

[0111] In the embodiments of the present application, after obtaining the code, in order to determine whether the code can run normally, the server can input the code into the code testing model to execute the code testing process and conduct code testing on the code.

[0112] Exemplarily, the code testing model can conduct test evaluation on the code, evaluate from aspects such as coding specifications, executability, redundancy, etc., and generate test results. The generated test results can include but are not limited to modification suggestions or existing problems, such as incorrect syntax, un-referenced variables, possible defects, and other problems.

[0113] Among them, the code testing model can include but is not limited to the Qwen2.5-Coder-7B-Instruct model.

[0114] S308. The server inputs the test results into the code correction model to execute the code correction process and obtains the corrected code.

[0115] In the embodiments of the present application, when the test results are input into the code correction model, the code correction model can correct the code based on the problems or modification opinions in the test results to obtain the corrected code that meets the specifications and can be executed smoothly.

[0116] Among them, the code correction model can include but is not limited to the Qwen2.5-Coder-7B-Instruct model.

[0117] S309. The server inputs the corrected code into the document writing model to execute the document writing process, annotates and summarizes the corrected code, and generates the code file corresponding to the request information.

[0118] In the embodiments of the present application, the server inputs the corrected code into the document writing model, and the document writing model can annotate and summarize the corrected code to generate the code file corresponding to the request information.

[0119] Exemplarily, the code file corresponding to the request information can include but is not limited to the code with annotations and the document description. Among them, the document description includes a summary description of the code function, which is convenient for users to understand.

[0120] Among them, the document writing model may include, but is not limited to, the Qwen2.5-7B-Instruct model.

[0121] S310. The server sends the code file corresponding to the request information to the client.

[0122] The server sends the code file corresponding to the request information to the client so that the user can obtain the complete annotated code and documentation instructions through the client.

[0123] In summary, the method provided by the embodiment of the present application can automatically generate a code file that serves the user's needs based on the code writing model combination. In the above example, based on this code writing model, the front-end and back-end codes related to the business trip approval page can be automatically generated. The business trip approval page corresponding to this code file supports filling in basic personnel information, business trip details, main / carbon copy recipients, etc. This method is suitable for the rapid development of internal office automation (OA) pages in the company.

[0124] A content generation method provided by an embodiment of the present application can identify the elements of the user's needs in the file based on the file recognition model combination to obtain a recognition result. As Figure 4 shown, this figure is a flowchart of a file recognition method provided by an embodiment of the present application.

[0125] S401. The server obtains the request information input by the user.

[0126] In the embodiment of the present application, the request information input by the user may be a file, such as an official document, etc.

[0127] S402. The server determines that the type of the task is file recognition according to the request information.

[0128] S403. The server splits the task into a security detection process and a tag recognition process according to the type of the task.

[0129] In the embodiment of the present application, the security detection process may include at least one of a first detection process and a second detection process. The tag recognition process may include at least one of, but is not limited to, a first recognition process, a second recognition model, a third recognition process, and a fourth recognition process.

[0130] Among them, the first detection process is the security detection process before the label recognition process, and the second detection process is the security detection process after the label recognition process. The first recognition process is the process of recognizing the sender unit in the request information; the second recognition process is the process of recognizing the main recipient and / or carbon copy units in the request information; the third recognition process is the process of recognizing the document handling and / or reading in the request information; the fourth recognition process is the process of recognizing the business affiliated department in the request information.

[0131] Subsequently, only taking the security detection process including the first detection process and the second detection process as an example for introduction.

[0132] S404. The server executes the first detection process in the security detection process based on the first detection model in the first security detection model, performs security detection on the request information, and obtains the first security detection result.

[0133] Exemplarily, during the process of the server executing the first detection process based on the first detection model, security detection can be performed on the text in the request information. For example, detecting whether there are risk words in the text, whether there are non-compliant terms, etc.

[0134] Among them, the first detection model can be obtained based on the Pangu security detection model, and the Pangu security detection model is a security detection model based on deep learning and natural language processing technologies.

[0135] It should be noted that in the embodiments of the present application, the risk words and non-compliant terms, etc. are not specifically limited, and they can all be set according to requirements.

[0136] In the case where the first security detection result is unqualified, for example, the server detects at least one of risk words in the text or non-compliant terms, etc., the server can execute S405; in the case where the first security detection result is qualified, for example, the server does not detect risk words and non-compliant terms in the text, the server can execute S406.

[0137] S405. The server returns the first security detection result to the client.

[0138] In the case where the first security detection result is unqualified, the server returns the first security detection result to the client so that the user can obtain the information that the first security detection result is unqualified.

[0139] In this case, the user can modify the text in the request information based on the first security detection result, for example, deleting or replacing the risk words in the text, modifying non-compliant terms, etc., to obtain the modified request information.

[0140] By sending the modified request information to the server, the server can execute S401 again.

[0141] S406. The server executes the first recognition process in the label recognition process based on the first recognition model in the file recognition model combination to recognize the sender unit in the request information and determine the sender unit.

[0142] When the first security detection result is qualified, the server can execute the label recognition process based on the label recognition model to perform label recognition on the request information and obtain the recognition result.

[0143] Exemplarily, the first recognition model can be obtained based on the Pangu semantic recognition model. The Pangu semantic recognition model is an AI large model based on natural language processing (NLP) technology and shows powerful capabilities and advantages in processing complex Chinese language tasks.

[0144] Based on the first recognition model, the server can recognize the sender unit in the request information and determine the sender unit.

[0145] S407. The server executes the second recognition process in the label recognition process based on the second recognition model in the label recognition model to recognize the main recipient and / or carbon copy unit in the request information and determine the main recipient and / or carbon copy unit.

[0146] Exemplarily, the second recognition model can be obtained based on the Pangu semantic recognition model.

[0147] S408. The server executes the third recognition process in the label recognition process based on the third recognition model in the label recognition model to recognize the case to be handled and / or the document to be read in the request information and determine the case to be handled and / or the document to be read.

[0148] Exemplarily, the third recognition model can be obtained based on the Pangu semantic recognition model.

[0149] S409. The server executes the fourth recognition process in the label recognition process based on the fourth recognition model in the label recognition model to recognize the business affiliated department in the request information and determine the business affiliated department.

[0150] Exemplarily, the fourth recognition model can be obtained based on the Pangu semantic recognition model.

[0151] It should be noted that the execution order of S406 - S409 in the embodiments of this application is not specifically limited. Exemplarily, S406 - S409 can be executed in a random order, or S406 - S409 can be executed simultaneously.

[0152] S410. The server uses at least one of the incoming document unit, the main recipient and / or carbon copy recipient units, the document to be processed and / or the document to be read, and the business affiliated department as the recognition result.

[0153] It can be understood that in the embodiments of the present application, other information in the request information can be recognized as needed. Here, only the recognition of at least one of the incoming document unit, the main recipient and / or carbon copy recipient units, the document to be processed and / or the document to be read, and the business affiliated department is taken as an example for introduction.

[0154] S411. The server executes the second detection process in the security detection process based on the second detection model in the first security detection model to perform a security detection on the recognition result and obtains a second security detection result.

[0155] Exemplarily, during the process of the server executing the second detection process based on the second detection model, it can perform a security detection on the text in the detection result. For example, it can detect whether there are risk words in the text, whether there are non-compliant terms, etc.

[0156] Among them, the second detection model can be obtained based on the Pangu security detection model.

[0157] It should be noted that in the embodiments of the present application, the risk words and non-compliant terms are not specifically limited, and they can all be set according to requirements.

[0158] In the case where the second security detection result is unqualified, for example, the server detects at least one of the presence of risk words in the text or the presence of non-compliant terms, etc., the server can execute S412; in the case where the second security detection result is qualified, for example, the server does not detect the presence of risk words and non-compliant terms in the text, the server can execute S413.

[0159] S412. The server returns the second security detection result to the client.

[0160] In the case where the second security detection result is unqualified, the server returns the second security detection result to the client so that the user can obtain the information that the second security detection result is unqualified.

[0161] In this case, the user can modify the text in the request information based on the second security detection result to obtain a modified request information.

[0162] The client sends the modified request information to the server, and the server can execute S401 again.

[0163] S413. The server uses the recognition result as the content corresponding to the request information.

[0164] When the second security detection result is qualified, the server summarizes information such as the incoming unit, the main recipient and / or carbon copy recipient, the document handling and / or reading, and the business affiliated department obtained from each model in the label recognition model to obtain an identification result, and uses this identification result as the content corresponding to the request information.

[0165] S414. The server sends the content corresponding to the request information to the client.

[0166] The server sends the content corresponding to the request information to the client so that the user can obtain this identification result through the client. Based on this identification result, the file can be sent to the corresponding department to improve efficiency.

[0167] In summary, the method provided by the embodiment of the present application can automatically identify the main elements in the file based on the file recognition model combination, such as the incoming unit, the main recipient / carbon copy recipient, the document handling / reading, the business affiliated department, etc., so that the file can be sent to the corresponding department based on the identification result subsequently.

[0168] A content generation method provided by an embodiment of the present application can generate text content that meets user needs based on a text generation model combination, improving the work efficiency of users. As Figure 5 shown, this figure is a flowchart of a text generation method provided by an embodiment of the present application.

[0169] S501. The server obtains the request information input by the user.

[0170] In the embodiment of the present application, the request information input by the user may include information about the text that the user needs to generate.

[0171] Taking the example that the user needs to generate a speech draft, the information of this text may include but is not limited to the style, type, etc. of the text.

[0172] It should be noted that in the embodiment of the present application, the format of the request information is not specifically limited. Exemplarily, the user can input the request information in text form or in voice form.

[0173] S502. The server determines that the type of the task is text generation according to the request information.

[0174] S503. The server splits the task into a security detection process and a text generation process according to the type of the task.

[0175] In the embodiment of the present application, the security detection process may include at least one of a third detection process and a fourth detection process. Among them, the third detection process is the security detection process before executing the text generation process, and the fourth detection process is the security detection process after executing the text generation process.

[0176] Subsequently, only the security detection process including the third detection process and the fourth detection process will be taken as an example for introduction.

[0177] S504. The server executes the third detection process in the security detection process based on the third detection model in the second security detection model to perform security detection on the request information and obtains the third security detection result.

[0178] Exemplarily, when the server executes the third detection process based on the second detection model, it can perform security detection on the request information. For example, it can detect whether there are risk words in the request information, whether there are non-compliant terms, etc.

[0179] Among them, the third detection model can be obtained based on the Pangu security detection model.

[0180] It should be noted that in the embodiments of the present application, the risk words, non-compliant terms, etc. are not specifically limited, and they can all be set according to requirements.

[0181] In the case where the third security detection result is unqualified, for example, the server detects at least one of the existence of risk words or non-compliant terms in the request information, the server can execute S505; in the case where the third security detection result is qualified, for example, the server does not detect the existence of risk words and non-compliant terms in the request information, the server can execute S506.

[0182] S505. The server returns the third security detection result to the client.

[0183] In the case where the third security detection result is unqualified, the server returns the third security detection result to the client so that the user can obtain the information that the third security detection result is unqualified.

[0184] In this case, the user can modify the request information based on the third security detection result. For example, delete or replace the risk words in the text, modify the non-compliant terms, etc., to obtain the modified request information.

[0185] The client sends the modified request information to the server, and the server can execute S401 again.

[0186] S506. The server executes the text generation process based on the text generation model in the text generation model combination to generate the initial text content.

[0187] In the embodiments of the present application, in the case where the third security detection result is qualified, the server executes the text generation process based on the text generation model in the text generation model combination to generate the initial text content that conforms to the style and type of the text required by the user.

[0188] Exemplarily, the text generation model can be obtained based on the Pangu text generation model.

[0189] S507. The server executes the fourth detection process in the security detection process based on the fourth detection model in the second security detection model to perform security detection on the initial text content and obtains the fourth security detection result.

[0190] Exemplarily, during the process of the server executing the fourth detection process based on the fourth detection model, security detection can be performed on the initial text content. For example, it can be detected whether there are risk words or non-compliant terms in the initial text content, etc.

[0191] Among them, the fourth detection model can be obtained based on the Pangu security detection model.

[0192] It should be noted that in the embodiments of the present application, no specific limitations are imposed on risk words and non-compliant terms, etc., and they can all be set according to requirements.

[0193] In the case where the fourth security detection result is unqualified, for example, the server detects at least one of the existence of risk words or non-compliant terms in the initial text content, the server can execute S508; in the case where the fourth security detection result is qualified, for example, the server does not detect the existence of risk words and non-compliant terms in the initial text content, the server can execute S509.

[0194] S508. The server returns the fourth security detection result to the client.

[0195] In the case where the fourth security detection result is unqualified, the server returns the fourth security detection result to the client so that the user can obtain the information that the fourth security detection result is unqualified.

[0196] In this case, the user can modify the request information based on the fourth security detection result to obtain the modified request information.

[0197] The client sends the modified request information to the server, and the server can execute S501 again.

[0198] S509. The server uses the initial text content as the text content corresponding to the request information.

[0199] In the case where the fourth security detection result is qualified, the server uses the initial text content as the text content corresponding to the request information.

[0200] S510. The server sends the text content corresponding to the request information to the client.

[0201] Exemplarily, in the case where the user needs to generate a speech draft, the server can output a speech draft that meets the user's requirements in terms of style and type based on the text generation model combination.

[0202] In summary, the method provided by the embodiments of the present application is based on the text generation model combination, and can automatically generate speech draft content that meets the user's requirements, such as speeches, reports, hosting, speeches, etc., improving the user's work efficiency.

[0203] It should be noted that the method provided by the embodiments of the present application can be used to execute various types of tasks. Here, only the types of tasks being writing code, file recognition, and text generation are taken as examples for introduction. As Figure 6 shown, this figure is a schematic diagram of a content generation method based on model combination provided by the embodiments of the present application.

[0204] In summary, in the embodiments of the present application, since each task process is executed based on a model that matches it, content that meets the user's expectations can be generated, improving the quality of the generated content. At the same time, in the embodiments of the present application, the method of using multiple models to cooperate to execute tasks and generate the content corresponding to the request information can effectively improve the efficiency of content generation.

[0205] The embodiments of the present application also provide a content generation device. As Figure 7 shown, this figure is a schematic structural diagram of a content generation device provided by the embodiments of the present application. Its specific implementation manner is the same as the implementation manner and the achieved technical effects recorded in the embodiments of the above method, and some contents will not be elaborated.

[0206] A content generation device 7100 provided by the embodiments of the present application includes: a determination unit 7101, a splitting unit 7102, and a generation unit 7103;

[0207] The determination unit 7101 is configured to determine the type of the task according to the request information;

[0208] The splitting unit 7102 is configured to split the task into multiple task processes according to the type of the task;

[0209] The generation unit 7103 is configured to execute the multiple task processes based on the models in the model combination that match the respective task processes in the execution order of the multiple task processes, and generate the content corresponding to the request information. The model combination corresponds to the type of the task.

[0210] In a possible implementation manner, the device includes a model combination generation unit, which is specifically configured to:

[0211] Determine the models that match each task process based on the multiple task processes corresponding to the type of the task;

[0212] Integrate the models matching each task process to obtain the model combination.

[0213] In a possible implementation, when the type of the task is code writing, the model combination is the code writing model combination, and the splitting unit is specifically used for:

[0214] According to the type of the task, split the task into a requirements analysis process, a function design process, a code writing process, a code testing process, a code correction process, and a document writing process;

[0215] The generating unit is specifically used for:

[0216] According to the execution order of the requirements analysis process, the function design process, the code writing process, the code testing process, the code correction process, and the document writing process, based on the requirements analysis model, the function design model, the code writing model, the code testing model, the code correction model, and the document writing model in the code writing model combination, respectively execute the requirements analysis process, the function design process, the code writing process, the code testing process, the code correction process, and the document writing process to generate the code file corresponding to the request information.

[0217] In a possible implementation, the generating unit is specifically used for:

[0218] According to the execution order of the requirements analysis process, the function design process, the code writing process, the code testing process, the code correction process, and the document writing process, execute the requirements analysis process based on the requirements analysis model in the code writing model combination, perform data analysis on the request information to obtain an analysis result;

[0219] Input the analysis result into the function design model to execute the function design process to determine the requirements for the steps of writing code;

[0220] Input the requirements for the steps of writing code into the code writing model to execute the code writing process to obtain code;

[0221] Based on the code testing model, execute the code testing process to perform code testing on the code to obtain a test result;

[0222] Input the test result into the code correction model to execute the code correction process to obtain the corrected code;

[0223] Input the corrected code into the document writing model to execute the document writing process, annotate and summarize the corrected code, and generate the code file corresponding to the request information.

[0224] In a possible implementation, when the type of the task is file recognition, the model combination is a file recognition model combination, and the splitting unit is specifically configured to:

[0225] Split the task into a security detection process and a label recognition process according to the type of the task;

[0226] The generating unit is specifically configured to:

[0227] Execute the security detection process and the label recognition process respectively based on the first security detection model and the label recognition model in the file recognition model combination according to the execution order of the security detection process and the label recognition process, and generate the content corresponding to the request information.

[0228] In a possible implementation, the generating unit is specifically configured to:

[0229] Execute the first detection process in the security detection process based on the first detection model in the first security detection model according to the execution order of the security detection process and the label recognition process, perform security detection on the request information, and obtain a first security detection result;

[0230] In the case where the first security detection result is qualified, execute the label recognition process based on the label recognition model, perform label recognition on the request information, and obtain a recognition result;

[0231] Execute the second detection process in the security detection process based on the second detection model in the first security detection model, perform security detection on the recognition result, and obtain a second security detection result;

[0232] In the case where the second security detection result is qualified, use the recognition result as the content corresponding to the request information.

[0233] In a possible implementation, the generating unit is specifically configured to:

[0234] Execute the first recognition process in the label recognition process based on the first recognition model in the label recognition model, recognize the sender unit in the request information, and determine the sender unit;

[0235] Execute the second recognition process in the label recognition process based on the second recognition model in the label recognition model, recognize the main recipient and / or carbon copy unit in the request information, and determine the main recipient and / or carbon copy unit;

[0236] Execute the third recognition process in the label recognition process based on the third recognition model in the label recognition model, recognize the document handling and / or reading item in the request information, and determine the document handling and / or reading item;

[0237] Execute the fourth recognition process in the label recognition process based on the fourth recognition model in the label recognition model to recognize the business attribution department in the request information and determine the business attribution department;

[0238] Use at least one of the sender unit, the main recipient and / or carbon copy recipients, the document to be processed and / or the document to be read, and the business attribution department as the recognition result.

[0239] In a possible implementation, when the type of the task is text generation, the model combination is a text generation model combination, and the splitting unit is specifically used for:

[0240] Split the task into a security detection process and a text generation process according to the type of the task;

[0241] The generating unit is specifically used for:

[0242] Based on the second security detection model and the text generation model in the text generation model combination, execute the security detection process and the text generation process respectively according to the execution order of the security detection process and the text generation process, and generate the text content corresponding to the request information.

[0243] In a possible implementation, the generating unit is specifically used for:

[0244] According to the execution order of the security detection process and the text generation process, execute the third detection process in the security detection process based on the third detection model in the second security detection model in the text generation model combination to perform security detection on the request information and obtain the third security detection result;

[0245] When the third security detection result is qualified, execute the text generation process based on the text generation model to generate the initial text content;

[0246] Execute the fourth detection process in the security detection process based on the fourth detection model in the second security detection model in the text generation model combination to perform security detection on the initial text content and obtain the fourth security detection result;

[0247] When the fourth security detection result is qualified, use the initial text content as the text content corresponding to the request information.

[0248] In summary, the device provided by the embodiments of the present application can split a task into multiple task processes, and execute each task process based on a model combination that matches the type of the task, so as to generate the content corresponding to the request information. Since each task process is executed based on a model that matches it, content that meets the user's expectations can be generated, improving the quality of the generated content. At the same time, in the embodiments of the present application, a method of using multiple models to cooperate to execute tasks and generate the content corresponding to the request information can effectively improve the efficiency of content generation.

[0249] As Figure 8 shown, this figure is a schematic structural diagram of a computing device provided by an embodiment of the present application. In this embodiment, the computing device may be a server, and the server may include, but is not limited to, a cabinet server, a rack server, a blade server, etc., or a general-purpose server, a GPU server, a DPU server, etc., without limitation here.

[0250] The computing device includes a memory 8101, a processor 8102, and a communication interface 8103; wherein, the memory 8101 stores computer instructions, and the processor 8102 is configured to execute the computer instructions, so that the computing device executes a content generation method as shown above.

[0251] In some embodiments, the processor 8102 may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0252] In some embodiments, the memory 8101 may be a volatile memory or a non-volatile memory, such as a register, etc. Specifically, a volatile memory refers to a memory in which the stored data will be lost after the power supply is interrupted. Among them, the volatile memory is mainly a random access memory (RAM), including a static random access memory (SRAM) and a dynamic random access memory (DRAM). A non-volatile memory refers to a memory in which the stored data will not be lost even if the power supply is interrupted. Common non-volatile memories include a read only memory (ROM), an optical disc, a magnetic disk, a solid state drive, and various memory cards based on flash memory technology, etc.

[0253] In some embodiments, the memory 8101 has executable code, and the processor 8102 executes this code. The communication interface 8103 can be used to implement communication between the client and the server.

[0254] The bus may be a Peripheral Component Interconnect (PCI) bus, or an extended industry standard architecture (eisa) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of understanding, Figure 8 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0255] As described above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A content generation method, characterized in that, Including: Determine the type of the task according to the request information; Split the task into multiple task processes according to the type of the task; Execute the multiple task processes in the execution order of the multiple task processes, based on the models in the model combination that match each task process, to generate the content corresponding to the request information, where the model combination corresponds to the type of the task.

2. The method according to claim 1, wherein The generation steps of the model combination include: Determine the models that match each task process based on the multiple task processes corresponding to the type of the task; Integrate the models that match each task process to obtain the model combination.

3. The method according to claim 1 or 2, characterized in that, When the type of the task is code writing, the model combination is the code writing model combination, and the splitting of the task into multiple task processes according to the type of the task includes: Split the task into a requirements analysis process, a function design process, a code writing process, a code testing process, a code correction process, and a document writing process according to the type of the task; The execution of the multiple task processes in the execution order of the multiple task processes, based on the models in the model combination that match each task process, to generate the content corresponding to the request information includes: Execute the requirements analysis process, the function design process, the code writing process, the code testing process, the code correction process, and the document writing process in the execution order of the requirements analysis process, the function design process, the code writing process, the code testing process, the code correction process, and the document writing process, based on the requirements analysis model, the function design model, the code writing model, the code testing model, the code correction model, and the document writing model in the code writing model combination, respectively, to generate the code file corresponding to the request information.

4. The method according to claim 3, characterized in that, The execution of the requirements analysis process, the function design process, the code writing process, the code testing process, the code correction process, and the document writing process in the execution order of the requirements analysis process, the function design process, the code writing process, the code testing process, the code correction process, and the document writing process, based on the requirements analysis model, the function design model, the code writing model, the code testing model, the code correction model, and the document writing model in the code writing model combination, respectively, to generate the code file corresponding to the request information includes: Execute the requirements analysis process based on the requirements analysis model in the code writing model combination in the execution order of the requirements analysis process, the function design process, the code writing process, the code testing process, the code correction process, and the document writing process, perform data analysis on the request information, and obtain an analysis result; Input the analysis result into the function design model to execute the function design process and determine the requirements for the steps of code writing; Input the requirements for the steps of code writing into the code writing model to execute the code writing process and obtain the code; Execute the code testing process based on the code testing model, perform code testing on the code, and obtain a test result; Input the test result into the code correction model to execute the code correction process and obtain the corrected code; Input the corrected code into the document writing model to execute the document writing process, annotate and summarize the corrected code, and generate a code file corresponding to the request information.

5. The method according to claim 1 or 2, characterized in that When the type of the task is file recognition, the model combination is a file recognition model combination. According to the type of the task, the task is split into multiple task processes, including: According to the type of the task, split the task into a security detection process and a label recognition process; In accordance with the execution order of the multiple task processes, based on the models in the model combination that match each task process, execute the multiple task processes to generate the content corresponding to the request information, including: In accordance with the execution order of the security detection process and the label recognition process, based on the first security detection model and the label recognition model in the file recognition model combination, execute the security detection process and the label recognition process respectively to generate the content corresponding to the request information.

6. The method according to claim 5, wherein In accordance with the execution order of the security detection process and the label recognition process, based on the first security detection model and the label recognition model in the file recognition model combination, execute the security detection process and the label recognition process respectively to generate the content corresponding to the request information, including: In accordance with the execution order of the security detection process and the label recognition process, based on the first detection model in the first security detection model, execute the first detection process in the security detection process to perform a security detection on the request information and obtain a first security detection result; When the first security detection result is qualified, based on the label recognition model, execute the label recognition process to perform a label recognition on the request information and obtain a recognition result; Based on the second detection model in the first security detection model, execute the second detection process in the security detection process to perform a security detection on the recognition result and obtain a second security detection result; When the second security detection result is qualified, use the recognition result as the content corresponding to the request information.

7. The method according to claim 6, characterized in that, The step of, based on the label recognition model, executing the label recognition process to perform a label recognition on the request information and obtain a recognition result includes: Based on the first recognition model in the label recognition model, execute the first recognition process in the label recognition process to recognize the sender unit in the request information and determine the sender unit; Based on the second recognition model in the label recognition model, execute the second recognition process in the label recognition process to recognize the main recipient and / or carbon copy unit in the request information and determine the main recipient and / or carbon copy unit; Based on the third recognition model in the label recognition model, execute the third recognition process in the label recognition process to recognize the case to be handled and / or document to be read in the request information and determine the case to be handled and / or document to be read; Based on the fourth recognition model in the label recognition model, execute the fourth recognition process in the label recognition process to recognize the business unit in charge in the request information and determine the business unit in charge; Use at least one of the sender unit, the main recipient and / or carbon copy unit, the case to be handled and / or document to be read, and the business unit in charge as the recognition result.

8. The method according to claim 1 or 2, characterized in that, When the type of the task is text generation, the model combination is a text generation model combination. Splitting the task into multiple task processes according to the type of the task includes: Splitting the task into a security detection process and a text generation process according to the type of the task; Executing the multiple task processes according to the execution order of the multiple task processes, based on the models in the model combination that match each task process, to generate the content corresponding to the request information, including: Executing the security detection process and the text generation process respectively according to the execution order of the security detection process and the text generation process, based on the second security detection model and the text generation model in the text generation model combination, to generate the text content corresponding to the request information.

9. The method according to claim 8, wherein Executing the security detection process and the text generation process respectively according to the execution order of the security detection process and the text generation process, based on the second security detection model and the text generation model in the text generation model combination, to generate the text content corresponding to the request information, including: Executing the third detection process in the security detection process based on the third detection model in the second security detection model in the text generation model combination according to the execution order of the security detection process and the text generation process, to perform security detection on the request information and obtain a third security detection result; When the third security detection result is qualified, executing the text generation process based on the text generation model to generate the initial text content; Executing the fourth detection process in the security detection process based on the fourth detection model in the second security detection model in the text generation model combination, to perform security detection on the initial text content and obtain a fourth security detection result; When the fourth security detection result is qualified, using the initial text content as the text content corresponding to the request information.

10. A computing device, characterized in that, The computing device includes: a processor and a memory; The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the steps of a content generation method according to any one of claims 1-9 based on the instructions in the program code.

Images