Content generation method and electronic equipment

Abstract

The embodiment of the invention discloses a content generation method and electronic equipment, and the method comprises the steps: receiving a content generation request of a user, and determining the content generation demand information inputted by the user; generating first prompt information, and calling the first AI generation model to perform preliminary generation of contents; performing quality inspection on the preliminarily generated content, if a target quality inspection item does not pass, generating second prompt information, and calling a second AI generation model to perform content regeneration to obtain target content; the second AI generation model is a special model for the target quality inspection item. Through the embodiment of the invention, high-quality content generation can be completed under the condition of not depending on the understanding of the user on the prompt word engineering.

Description

(19) State Intellectual Property Office (12) Invention Patent Application (10) Application Publication Number (43) Application Publication Date (21) Application Number 202511384191.0 (22) Application Date 2025.09.25 (71) Applicant Hangzhou Alibaba Overseas Internet Industry Co., Ltd. Address 310056, Room 303, 3rd Floor, Building 5, No. 699, Wangshang Road, Changhe Street, Binjiang District, Hangzhou City, Zhejiang Province (72) Inventors Xiao Wo, Dai Lixia, Qi Yajun, An Yang, Dong Jie (74) Patent Agency Beijing Zhongda Dequan Intellectual Property Agency Co., Ltd. 11570 Patent Attorney Nan Haiyan (51) Int.Cl. G06F 40 / 58 (2020.01) G06F 16 / 335 (2019.01) G06F 16 / 334 (2025.01) G06N 20 / 00 (2019.01) (54) Invention Title: Content Generation Method and Electronic Device (57) Abstract: This application discloses a content generation method and electronic device, including: receiving a user's content generation request and determining the user's input content generation requirement information; generating a first prompt message and calling a first AI generation model to perform preliminary content generation; performing a quality check on the preliminary generated content; if the target quality check item fails, generating a second prompt message and calling a second AI generation model to regenerate the content and obtain the target content; the second AI generation model is a dedicated model for the target quality check item. Through this application embodiment, high-quality content generation can be completed without relying on the user's understanding of the prompt word engineering. Claims (2 pages), Description (13 pages), Drawings (3 pages), CN 121351846 A, 2026.01.16, CN 1 21 35 18 46 A. 1. A content generation method, characterized in that it includes: receiving a user's content generation request and determining content generation requirement information input by the user, the content generation requirement information including content description text, the required content type, and at least one required target language information; generating first prompt information and calling a first AI generation model to perform preliminary content generation; wherein, the first prompt information includes the user-input content generation requirement information and content generation guide information extracted from a database based on the content type and the target language information; the content generation guide is generated based on the professional knowledge provided by expert users corresponding to the target language for the content type and is pre-saved in the database; performing a quality check on the preliminarily generated content, and if the target quality check item fails, generating a second prompt information and calling a second AI generation model to regenerate the content to obtain the target content; itsIn the method described in claim 1, the second AI generation model is a dedicated model for the target quality inspection item. The second prompt information includes the content to be rewritten, the content generation requirement information input by the user, and information related to the target quality inspection item in the content generation guide information. The content to be rewritten includes the initially generated content or the content generated after the previous quality inspection step. 2. The method according to claim 1, characterized in that, after receiving the user's content generation request and determining the content generation requirement information input by the user, it further includes: selecting a corresponding generation path according to the content type, the generation path being used to define multiple steps included in content generation and the execution order between the multiple steps, the multiple steps including a preliminary generation step and at least one quality inspection step, so that the first AI generation model is called during the execution of the preliminary generation step according to the generation path, and the second AI generation model is called during the execution of the quality inspection step according to the generation path after the preliminary generation of the content is completed. 3. The method according to claim 2, characterized in that, different content types correspond to different generation paths, and the number and / or order of quality inspection steps are different in different generation paths. 4. The method according to claim 1, characterized in that, if multiple target languages ​​are required, the target content corresponding to the first language is generated first, and then the third AI generation model corresponding to other languages ​​is called to perform preliminary generation of content corresponding to other languages, and a quality check step is performed according to the generation path. If the target quality check item fails, the target content corresponding to other languages ​​is regenerated through the second AI generation model; the third AI generation model is multiple, each corresponding to a different target language; the third prompt information corresponding to the third AI generation model includes the target content corresponding to the first language, the content generation requirement information input by the user, and the content generation guide information corresponding to other languages. 5. The method according to claim 1 or 4, characterized in that, the target quality check item includes a character count check; the second AI generation model is specifically used to, if the character count of the preliminarily generated content exceeds a threshold, or the character count of the content generated after the previous quality check step exceeds a threshold, then according to the second prompt information, with the goal of reducing the character count, the preliminarily generated content or the content generated after the previous quality check step is rewritten to generate content that is qualified in terms of character count. 6. The method according to claim 1 or 4, characterized in that, the target quality inspection items include sensitive word filtering;The second AI generation model is specifically used to, if the initially generated content contains sensitive words, or if the content generated after the previous quality check step contains sensitive words, rewrite the initially generated content or the content generated after the previous quality check step according to the second prompt information, with the goal of replacing the sensitive words, to generate content that is qualified in terms of sensitive words. 7. The method according to claim 1, wherein, during the initial generation of content, the generated prompt information further includes the content type and example information corresponding to the target language in the sample library. 8. A content generation method, comprising: determining the user's content generation requirement information through a target interface, the content generation requirement information including content description text, the required content type, and at least one required target language information; after receiving a content generation request from the user, submitting the request and the content generation requirement information to a server, so that the server generates target content according to the method according to any one of claims 1 to 7; and displaying the target content generated by the server. 9. A computer-readable storage medium storing a computer program thereon, wherein, when executed by a processor, the program implements the steps of the method according to any one of claims 1 to 8. 10. An electronic device, characterized in that it comprises: one or more processors; and a memory associated with the one or more processors, the memory being used to store program instructions, which, when read and executed by the one or more processors, perform the steps of the method according to any one of claims 1 to 8. 11. A computer program product, comprising a computer program / computer executable instructions, characterized in that, when the computer program / computer executable instructions are executed by a processor in an electronic device, they implement the steps of the method according to any one of claims 1 to 8. Claims 2 / 2 Page 3 CN 121351846 A Content Generation Method and Electronic Device Technical Field

[0001] This application relates to the field of content generation technology, and in particular to content generation methods and electronic devices. Background Art

[0002] In some cross-border scenarios, there is a huge demand for the production of high-quality multilingual creative content. For example, in cross-border commodity information service systems, it is often necessary to design banner images (also known as banner ads, which are banner-type images used to display brand, event, or product information on web pages or mobile interfaces, usually located in prominent positions such as the top of the page, serving to attract attention, convey core information, and guide users to click) on pages, which need to display title-type text content; or, it is necessary to design the theme of a promotional event, and so on. This content not only needs to be sufficiently attractive in its expression, but also requires compliance control in terms of length, filtering of prohibited words, etc.

[0003] In traditional systems targeting domestic users, the above content is mainly generated manually. However, in cross-border scenarios, the same page needs to be displayed to users in different countries / regions in different languages. When expressing content in different languages, the text length will be different, and the restrictions on prohibited words may also be different in different countries / regions, etc. Therefore, in the process of multilingual expression, it is not as simple as translation. Therefore, if the content production in multilingual scenarios is still done manually, the workload is huge, the efficiency is very low, and the knowledge requirements of the human personnel are also very high. If the generation is done using ordinary AI (Artificial Intelligence) models, it also requires users to have a strong understanding of the prompt word engineering of AI models, and also requires users to have a deep understanding of the languages ​​and cultures of multiple different countries / regions, otherwise it is difficult to generate content that meets the requirements and complies with regulations. However, since personnel with such capabilities are scarce, there are also problems such as high cost and low efficiency. Summary of the Invention

[0004] This application provides a content generation method and electronic device that can complete high-quality content generation without relying on the user's understanding of prompt word engineering.

[0005] This application provides the following solution:

[0006] A content generation method, comprising:

[0007] receiving a user's content generation request and determining the content generation requirement information input by the user, wherein the content generation requirement information includes content description text, the required content type, and at least one target language information;

[0008] generating a first prompt message and calling a first AI generation model to perform preliminary content generation; wherein the first prompt message includes the content generation requirement information input by the user, and content generation guide information extracted from a database according to the content type and the target language information; the content generation guide is generated based on the professional knowledge provided by expert users corresponding to the target language for the content type, and is pre-saved in the database;

[0009] performing a quality check on the preliminary generated content; if the target quality check item fails, generating a second prompt message and calling a second AI generation model to regenerate the content and obtain the target content; The second AI generation model is a dedicated model for the target quality inspection item. The second prompt information includes the content to be rewritten, the content generation requirement information input by the user, and information related to the target quality inspection item in the content generation guide information (page 1 / 13, CN 121351846 A). The content to be rewritten includes the initially generated content or the content generated after the previous quality inspection step.

[0010] The method further includes, after receiving the user's content generation request and determining the user's input content generation requirements, the following:

[0011] Selecting a corresponding generation path based on the content type. The generation path defines multiple steps involved in content generation and the execution order of these steps. The multiple steps include a preliminary generation step and at least one quality check step. This allows the first AI generation model to be invoked during the execution of the preliminary generation step according to the generation path, and the second AI generation model to be invoked during the execution of the quality check step according to the generation path after the preliminary content generation is completed.

[0012] Different content types correspond to different generation paths, and the number and / or order of quality check steps vary in different generation paths.

[0013] Wherein, if multiple target languages ​​are required, the target content corresponding to the first language is generated first, and then the third AI generation model corresponding to other languages ​​is called to perform preliminary generation of the content corresponding to other languages, and a quality check step is performed according to the generation path. If the target quality check item fails, the target content corresponding to other languages ​​is regenerated through the second AI generation model; there are multiple third AI generation models, each corresponding to a different target language;

[0014] The third prompt information corresponding to the third AI generation model includes the target content corresponding to the first language, the content generation requirement information input by the user, and the content generation guide information corresponding to other languages.

[0015] Wherein, the target quality check item includes character count check;

[0016] The second AI generation model is specifically used to, if the number of characters in the preliminarily generated content exceeds a threshold, or the number of characters in the content generated after the previous quality check step exceeds a threshold, then according to the second prompt information, with the goal of reducing the number of characters, rewrite the preliminarily generated content or the content generated after the previous quality check step to generate content that is qualified in terms of character count.

[0017] Wherein, the target quality inspection item includes sensitive word filtering;

[0018] The second AI generation model is specifically used to, if the initially generated content includes sensitive words, or the content generated after the previous quality inspection step includes sensitive words, according to the second prompt information, rewrite the initially generated content or the content generated after the previous quality inspection step with the goal of replacing the sensitive words, so as to generate content that is qualified in terms of sensitive words.

[0019] Wherein, during the initial generation of content, the generated prompt information also includes the content type and example information corresponding to the target language in the sample library.

[0020] A content generation method, comprising:

[0021] The user's content generation requirements are determined through the target interface. The content generation requirements include content description text, the type of content to be generated, and at least one target language information required.

[0022] After receiving a content generation request from the user, the request and the content generation requirements are submitted to the server so that the server can generate the target content according to the aforementioned method.

[0023] The target content generated by the server is displayed.

[0024] A computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method described in any of the preceding claims.

[0025] An electronic device, comprising:

[0026] one or more processors; and

[0027] a memory associated with the one or more processors, the memory storing program instructions which, when read and executed by the one or more processors, perform the steps of the method described in any of the preceding claims.

[0028] A computer program product includes a computer program / computer-executable instructions, which, when executed by a processor in an electronic device, implement the steps of any of the methods described above.

[0029] According to specific embodiments provided in this application, the following technical effects are disclosed:

[0030] Through embodiments of this application, when a user needs to generate content, they can input content generation requirement information, which may include content description text, the type of content to be generated, and at least one target language information. After receiving the generation request, the system can first generate a first prompt message and call a first AI generation model to perform preliminary content generation; wherein, the first prompt message includes the content generation requirement information input by the user, and content generation guide information extracted from the database according to the content type and the target language information; the content generation guide is generated based on the professional knowledge provided by expert users corresponding to the target language for the content type, and is pre-saved in the database. After the initial content generation is completed, if the target quality inspection item fails the quality inspection, a second prompt message can be generated, and a second AI generation model can be invoked to regenerate the content and obtain the target content. The second AI generation model is a dedicated model for the target quality inspection item. The second prompt message includes the content to be rewritten, the user-inputted content generation requirement information, and information related to the target quality inspection item from the content generation guide information. The content to be rewritten includes the initially generated content or the content generated after the previous quality inspection step. In this way, the content generation process can be divided into content...The system includes an initial generation stage and a quality check stage. Different prompts can be generated at different stages, and different AI generation models can be used to generate or rewrite the content. In addition, content generation guidelines provided by expert users can be injected into the prompts. Therefore, high-quality content generation can be completed without relying on the user's understanding of the prompt word engineering.

[0031] In a preferred embodiment, a corresponding generation path can be specified in advance for a specific content type. The specific generation path can be used to define multiple steps included in content generation and the execution order between the multiple steps. The multiple steps include an initial generation step and at least one quality check step. In order to call the first AI generation model to perform the initial generation of content during the execution of the initial generation step according to the generation path; after the initial generation of content is completed, the second AI generation model is called to perform quality check during the execution of the quality check step according to the generation path. Since different content types may have different requirements for the items, order, etc. of quality checks, this method of specifying corresponding generation paths for different content types in advance can better meet the generation and quality check needs of various different content types.

[0032] Of course, implementing any product of this application does not necessarily require achieving all the advantages described above simultaneously. Brief Description of the Drawings

[0033] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 is a schematic diagram of the system architecture provided in the embodiment of this application;

[0035] Figure 2 is a flowchart of the server method provided in the embodiment of this application; Specification 3 / 13 pages 6 CN 121351846 A

[0036] Figure 3 is a schematic diagram of the interface provided in the embodiment of this application;

[0037] Figure 4 is a schematic diagram of another interface provided in the embodiment of this application;

[0038] Figure 5 is a flowchart of the client method provided in the embodiment of this application;

[0039] Figure 6 is a schematic diagram of the electronic device provided in the embodiment of this application. Detailed Description of Embodiments

[0040] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art are within the scope of protection of this application.

[0041] First, it should be noted that even with advanced AI models, there are still many problems when generating content in multiple content types and multilingual scenarios. For example:

[0042] First, the content to be generated, such as the title of a webpage banner or the title of an event venue, is usually not simply generated. It also requires quality checks for compliance issues, including limitations on character length in some scenarios, or the presence of sensitive or prohibited words. Furthermore, different countries / regions have different regulations regarding the use of sensitive / prohibited words. Additionally, the quality check does not simply result in a pass / fail conclusion; if the check fails, the content needs to be regenerated. In this situation, if the same AI model is used to complete all the aforementioned quality checks during multilingual content generation, the quality of the generated results is usually very poor, at least in some languages, the generated results are not directly applicable.

[0043] Second, the quality check of the generated content usually includes multiple check items. However, for different content types (e.g., the title in a banner image, the title of an event venue, etc., are different content types), the required check items and the order of importance of different check items may be different. In this case, if a general path is used to complete the content generation and quality check tasks, the content generation quality may be low, making it difficult to meet the content quality requirements in specific application scenarios.

[0044] In view of the above situation, the embodiments of this application provide corresponding solutions to meet the content generation needs of multiple different content types (usually the content that needs to be generated under multiple different needs in the same application or the same product) and multiple languages ​​with the same solution. Specifically, firstly, the specific content generation process can be divided into two stages: preliminary content generation and quality check. Different AI models can be used to complete the tasks at different stages. Especially in the case of quality problems, the AI ​​model can be used to regenerate the content in the quality check stage. Furthermore, the quality inspection stage may typically involve multiple quality inspection items. Therefore, different AI models can be provided for different quality inspection items to complete the quality inspection tasks for different items.

[0045] In the initial content generation and quality inspection stages, some professional knowledge is required. The specific knowledge varies depending on the content type and language. Although the AI ​​model's own knowledge base contains some knowledge, it is usually far from sufficient when facing specific types, languages, and other professional scenarios. Another approach is to provide this knowledge when the user inputs content requirements.Knowledge, however, as mentioned above, in the context of multiple content types and multiple languages, it is almost impossible for ordinary users to have accurate knowledge when facing multiple different content types and multiple different languages ​​(it requires users to be very proficient in the requirements or restrictions of various content types and languages, etc., as detailed in the instruction manual 4 / 13 page 7 CN 121351846 A).

[0046] Therefore, in the embodiments of this application, a dedicated knowledge database can also be provided. Specifically, for various specific content types, experts in multiple languages ​​can be consulted and collaborated with to provide specific content generation guidelines for various content types, including specific precautions, words that need to be blocked, etc. These content generation guidelines can be saved in the database in a structured manner. In this way, for a specific user's content generation request, after determining the specific content type and target language to be generated, the corresponding content generation guide can be extracted from the database and injected into the prompt information. Then, the AI ​​model can be invoked based on this prompt information, providing the AI ​​model with relevant knowledge for content generation based on the specific content type and language, thereby reducing the reliance on the AI ​​model's own knowledge base and user input information.

[0047] In addition, this application embodiment also provides the concept of a content generation path, and different content types can correspond to different generation paths. A specific generation path can be used to define multiple steps included in content generation, and the execution order between these steps. Specific steps can include the aforementioned preliminary generation step, and at least one quality check step. The reason for describing it as "at least one quality check step" is that quality checks are necessary for different content types, but the specific items checked may vary. For example, some content types may require character length checks, while others do not, etc. Therefore, the number of quality check items corresponding to different content types may be different. Furthermore, even when checking items, the order between different items may differ. For example, for a certain content type, character length needs to be checked first, followed by sensitive words, while for other content types, sensitive words may need to be checked first, followed by checks on other items, and so on. These differences in quality check items and order can be reflected by defining different generation paths.

[0048] After completing the above preparations, the specific generation process can begin. Specifically, referring to Figure 1, this application embodiment can provide a content generation system for users. The client of this system can provide an interactive interface for users, through which users can submit specific content generation requirements, which may include specific content description text (e.g., ...).Information such as a 50% discount on a certain product category during a specific holiday, content type, and target language are then generated. The main program on the server side can then determine the corresponding generation path based on the specific content type and generate the content accordingly. During the initial generation, the main program can automatically generate a prompt message (referred to as the "first prompt message" for easy distinction from other prompt messages in subsequent processes). This first prompt message can include the user-inputted content generation requirements and content generation guidelines extracted from the database based on the content type and target language information, which are then injected into the prompt message. Next, an AI generation model (referred to as the "first AI generation model" for easy distinction from other AI models) can be invoked based on this prompt message to perform initial content generation. Afterward, the main program can execute quality check steps according to the generation path. If a target quality check item fails at a certain step, a second prompt message can be generated, which can include the aforementioned initially generated content, the user-inputted content requirements, and the content generation guidelines related to that quality check item. Afterwards, the second AI generation model can be called to regenerate the target content for the project. Other projects can also be processed similarly. After multiple projects have completed the check and passed, the final generation result can be obtained.

[0049] In addition, in specific implementation, for multilingual cases, in order to ensure the generation quality of various languages, different AI generation models can be provided for different languages. However, since different languages ​​are used differently, the probability of obtaining more accurate generation results is higher for more universal languages, and it will also have certain reference value for other languages. Therefore, the solution provided by the embodiments of this application can be that for languages ​​with international universality such as English (which can be called "first languages"), priority can be given to generation. That is, in the generation process specification on page 5 / 13 of CN 121351846 A mentioned above, the target content in the first language version can be generated first. Afterwards, for other languages, generation can be carried out based on the target content in the first language version, through translation, rewriting, etc. At this point, the specific prompts given to the AI ​​model may include the target content in the first language version, the content requirements input by the user, the specific content type, and the content generation guidelines for the corresponding other language. These are then input into the AI ​​model for the other language to perform initial content generation in that language. Afterwards, a quality check step can be performed to obtain the target content for the other language.

[0050] The specific implementation scheme provided by the embodiments of this application will be described in detail below.

[0051] First, this embodiment provides a content generation method from the perspective of the server side of the aforementioned generation system. Referring to Figure 2, the method may specifically include:

[0052] S201: Receiving a user's content generation request and determining the content generation requirement information input by the user. The content generation requirement information includes content description text, the type of content to be generated, and at least one target language information required.

[0053] Specifically, a client can provide the user with an interface for initiating a content generation request. The user can input specific content generation requirement information through the above interface, and the client can submit the request to the server after receiving it. Among them, the content generation requirement information may specifically include content description text, the type of content to be generated, and at least one target language information required. Regarding the content type and target language information, the above interface can provide the user with a variety of optional options. For example, as shown in Figure 3, it is an interface for submitting a content generation request under a specific implementation, including the content type option shown at 31 (where, assuming the user selected "banner", i.e., Banner image), the target language option shown at 32 (where, assuming the user selected "Spanish" and "English"), etc.

[0054] For the target language option, a list of possible languages ​​can be provided. Regarding the content type option, some content types can be predefined, and users can choose one of these types to generate content. The specific content types defined can be determined based on the needs of the specific application. For example, the content types can be defined based on the specific content that the application has corresponding multilingual generation requirements. Of course, new content types can also be added. However, to ensure the system supports new content types, relevant configurations can be made for the new content types, including configurations for content generation guidelines. That is, experts in multiple languages ​​can be invited to provide content generation guidelines for the new content type, which are then saved in the system's database. Additionally, a specific generation path can be selected or configured for the content type. Afterwards, the new content type can be added to the front-end interface for users to choose from, and so on.

[0055] Regarding the content description text, it can be entered by the user through an input box. For example, in the input box shown at position 33 in Figure 3, the user enters the content description text as: "Black Friday mobile phone 50% off," etc. In other words, the user wants the generated content to reflect "50% off mobile phones on Black Friday," but it needs to be described in a format suitable for a "banner image" title, and it also needs to be expressed in multiple languages. Therefore, the content generation needs to be performed using the system provided in this application embodiment.

[0056] Of course, in specific implementations, more detailed content generation requirements can be set. For example, when selecting specific...When selecting the content type, users can also choose the specific fields to include, such as whether to generate only the main title, or generate both the main title and subtitle, etc. Alternatively, they can limit the number of characters generated. That is to say, although there may be a limit on the number of characters in the specific content generation guide, users can also set a limit. If the user sets a character limit threshold, the generation can be based on the user's settings; otherwise, the generation can be based on the character limit in the specific content generation guide, etc. Manual 6 / 13 Page 9 CN 121351846 A

[0057] After completing the input of the specific content generation requirements information, the user can initiate a specific content generation request by clicking the "Generate" option as shown in Figure 3. The client can submit the request to the server.

[0058] After receiving the content generation request, the server can extract the user's content generation requirements information. In specific implementation, it can also perform format verification and metadata extraction operations. Metadata extraction can specifically include the aforementioned content type and target language information. Additionally, it can extract metadata with specific business semantics from the content description text, such as "event name," "product category," and "discount." In the example shown in Figure 3, the data in each of these metadata fields are "Black Friday," "mobile phone," and "50% off," respectively. After metadata extraction, the user's generation requirements can be input into the AI ​​generation model in this metadata form for content generation, facilitating more efficient content generation.

[0059] Furthermore, after receiving the content generation request, the server-side main program can, optionally, select the corresponding generation path based on the required content type. Specifically, the generation paths corresponding to various content types can be pre-configured and stored on the server side. Different content types can correspond to different generation paths, and the number and / or order of quality check steps differ in different generation paths. Of course, the relationship between content type and generation path does not have to be one-to-one; multiple content types can correspond to the same generation path, and so on.

[0060] After the user issues a specific content generation request, the content generation request will include the required content type information. Therefore, the generation path corresponding to the specific content type can be retrieved from the configuration information. The specific generation path defines which steps need to be performed when generating content of that content type, and the order of these steps. These specific steps may include a preliminary generation step and at least one quality check step.

[0061] S202: Generate a first prompt message and call the first AI generation model to perform preliminary content generation.The first prompt information includes the content generation requirement information input by the user, and the content generation guide information extracted from the database according to the content type and the target language information; the content generation guide is generated based on the professional knowledge provided by expert users corresponding to the target language for the content type, and is pre-saved in the database.

[0062] After determining the user's content generation type, content generation can be performed. If a generation path is also determined, the main program can perform specific generation according to the steps included in the generation path. In the preliminary generation step, the first prompt information can be provided to the first AI generation model. Specifically, the content generation requirement information input by the user can be injected into the first prompt information as user prompt information. In addition, system prompt information can also be injected into the first prompt information. Specifically, the system prompt information can be the content generation guide information provided to the first AI generation model. Since the specific content generation guide has been pre-saved in the database, the main program can extract the corresponding content generation guide information from the database according to the specific required content type and target language information, and then inject this content generation guide information into the first prompt information. The specific content generation guidelines can include restrictions (including character length restrictions, prohibited word restrictions, etc.) or precautions when generating content for specific content types and target languages. For example, country A may have the custom of celebrating a certain festival on a certain date, but another country B does not have the custom of celebrating that festival. In this case, when generating content related to the festival for country B, it may be necessary to change the festival to another festival that country B will celebrate during the corresponding time period, and so on.

[0063] In addition, in an optional manner, a sample library can be provided in advance, which can include some example content that has been generated for various content types in various target languages. In this case, the specific first prompt information can also be injected with the specific content type and the example information corresponding to the target language in the sample library to help the first AI generation model generate higher quality content. Specification 7 / 13 pages 10 CN 121351846 A

[0064] After the first prompt information is generated, the first AI generation model can be called with the first prompt information as input. The first AI generation model can then perform preliminary content generation according to the first prompt information. It should be noted that, in this embodiment, different AI generation models can be provided for multiple target languages. Specifically, in the preliminary generation step, different first prompts can be generated for the AI ​​generation models corresponding to different target languages, and preliminary generation for multiple languages ​​can be performed separately. Alternatively, as mentioned above, to improve efficiency, also...The target content corresponding to a more general first language such as English can be generated first. Then, the content in other languages ​​can be generated by translating or rewriting the target content in the first language. In this case, the first AI generation model in this step can specifically refer to the AI ​​generation model corresponding to the first language. The specific initially generated content can be the initially generated content in the first language.

[0065] S204: Perform a quality check on the initially generated content. If the target quality check item fails, generate a second prompt message and call the second AI generation model to regenerate the content to obtain the target content. The second AI generation model is a dedicated model for the target quality check item. The second prompt message includes the content to be rewritten, the content generation requirement information input by the user, and the information related to the target quality check item in the content generation guide information. The content to be rewritten includes the initially generated content or the content generated after the previous quality check step.

[0066] After the initial generation is completed, the main program can perform a quality check on the initially generated content. Specifically, if the generation path is predetermined, the quality checks on multiple items can be performed sequentially according to the order of the quality check steps in the generation path. After each project's quality check is completed, it can be determined whether the check has passed. If it has passed, the next project's quality check can proceed; otherwise, if it has failed, a second prompt message can be generated, and the second AI generation model corresponding to the project can be called to regenerate the content. This regeneration can be a rewrite-style generation to improve efficiency. Specifically, when generating the second prompt message, in addition to the content generation requirement information input by the user, the second prompt message can also include the initially generated content, or the content after the previous quality check step (which may have been rewritten in the previous step). In addition, regarding the system prompt message, content generation guide information related to the current specific project can be injected into the second prompt message. That is to say, since the content generation guide usually includes multiple aspects, when it is necessary to rewrite or rewrite the content for a certain project during the quality check process, only the content generation guide related to the project can be provided to the second AI generation model, so that the second AI generation model can focus more on the rewriting or rewriting task for that project.

[0067] The specific target quality check item can include character count (i.e., character length) check. The specific check process can be completed by the code in the main program. The second AI generation model can be used to address situations where the number of characters in the initially generated content exceeds a threshold, or the number of characters in the content generated after the previous quality check step exceeds a certain threshold.If the threshold is reached, the content generated initially or after the previous quality check step can be rewritten according to the second prompt information with the goal of reducing the number of characters, so as to generate content that complies with the number of characters.

[0068] If the target quality check item is sensitive word filtering, the second AI generation model can be used to rewrite the content generated initially or after the previous quality check step with the goal of replacing the sensitive words, according to the second prompt information, if the content generated initially or after the previous quality check step contains sensitive words, so as to generate content that complies with the number of sensitive words.

[0069] In addition, specific quality checks may also include punctuation standard checks (e.g., whether there is a period at the end). Among them, since the punctuation check task is relatively simple, even if the check fails, it is easy to add, delete, or rewrite the punctuation, so it does not need to be completed by the AI ​​model. That is to say, not all quality check items may need to use the AI ​​model, and it can be used only when necessary.

[0070] After completing the quality check and passing all quality checks, the target content can be obtained, that is, content that meets the conditions in terms of character length, prohibited words, etc.

[0071] As mentioned above, if multiple target languages ​​are required, the target content corresponding to the first language such as English can be generated first in the aforementioned manner, and then the third AI generation model corresponding to other languages ​​can be called to perform the preliminary generation of the content corresponding to other languages. The quality check steps are executed according to the specific generation path (since the content type remains unchanged, the generation path also remains unchanged). If the target quality check item fails, the target content corresponding to other languages ​​can be regenerated through the second AI generation model to obtain the target content corresponding to other languages. There can be multiple third AI generation models, each corresponding to a different target language. That is, the content in different languages ​​can be generated separately. The third prompt information corresponding to the third AI generation model can include the target content generation result corresponding to the first language, the content generation requirement information input by the user, and the content generation guide information corresponding to other languages.

[0072] After completing the target content in various languages, it can be returned to the client for display. For example, in the example shown in Figure 3, after generating English and Spanish content, the returned generation result can be as shown in Figure 4.

[0073] In summary, through the embodiments of this application, when a user needs to generate content, they can input content generation requirement information, which may include content description text, the type of content to be generated, and at least one target language information. After receiving the generation request, the system can first generate a first prompt message and call the first AI generation model to...The process involves initial content generation. The first prompt includes the user-inputted content generation requirement and content generation guidelines extracted from the database based on the content type and target language. These guidelines are generated using expert users providing professional knowledge relevant to the content type and are pre-saved in the database. After initial content generation, if the target quality check item fails, a second prompt is generated, and a second AI generation model is invoked to regenerate the content and obtain the target content. The second AI generation model is a dedicated model for the target quality check item. The second prompt includes the content to be rewritten, the user-inputted content generation requirement, and information related to the target quality check item from the content generation guidelines. The content to be rewritten includes the initially generated content or the content generated after the previous quality check step. Through the above method, the content generation process can be divided into a preliminary content generation stage and a quality check stage. Different prompts can be generated at different stages, and different AI generation models can be used to generate or rewrite the content. In addition, content generation guidelines provided by expert users can be injected into the prompts. Therefore, high-quality content generation can be completed without relying on the user's understanding of the prompt word engineering.

[0074] In a preferred embodiment, a corresponding generation path can be specified in advance for a specific content type. The specific generation path can be used to define multiple steps included in content generation and the execution order between the multiple steps. The multiple steps include a preliminary generation step and at least one quality check step, so that the first AI generation model is called during the execution of the preliminary generation step according to the generation path to perform preliminary content generation; after the preliminary content generation is completed, the second AI generation model is called during the execution of the quality check step according to the generation path to perform quality check. Since different content types may have different requirements for the items, order, etc. of quality checks, this method of specifying corresponding generation paths for different content types in advance can better meet the generation and quality check needs of various different content types.

[0075] Embodiment Two

[0076] Embodiment Two provides a content generation method from the client's perspective. Referring to Figure 5, this method, as described on page 9 / 13 of the specification (CN 121351846 A), includes:

[0077] S501: Determining the user's content generation requirement information through the target interface. The content generation requirement information includes content description text, the required content type, and at least one required target language information;

[0078] S502: After receiving a content generation request from a user, the request and the content generation requirement information are submitted to the server so that the server can generate the target content according to the method described in Embodiment 1 or Embodiment 2 above;

[0079] S503: The target content generated by the server is displayed.

[0080] For the parts of Embodiment 2 that are not detailed above, please refer to Embodiment 1 and other parts of this specification. They will not be repeated here.

[0081] It should be noted that the embodiments of this application may involve the use of user data. In practical applications, user-specific personal data can be used in the scheme described herein within the scope permitted by applicable laws and regulations, provided that it complies with the applicable laws and regulations of the country (e.g., the user explicitly consents, the user is effectively notified, etc.).

[0082] Corresponding to Embodiment 1, this application embodiment also provides a content generation apparatus, which may include:

[0083] a request receiving unit, configured to receive a user's content generation request and determine the content generation requirement information input by the user, wherein the content generation requirement information includes content description text, the required content type, and at least one target language information;

[0084] a preliminary generation unit, configured to generate first prompt information and call a first AI generation model to perform preliminary content generation; wherein the first prompt information includes the content generation requirement information input by the user, and content generation guide information extracted from the database according to the content type and the target language information; the content generation guide is generated based on the professional knowledge provided by expert users corresponding to the target language for the content type, and is pre-saved in the database;

[0085] a regeneration unit, configured to perform a quality check on the preliminary generated content, and if the target quality check item fails, generate a second prompt information and call a second AI generation model to regenerate the content to obtain the target content; wherein the second AI generation model is a dedicated model for the target quality check item. The second prompt information includes the content to be rewritten, the content generation requirement information input by the user, and information related to the target quality inspection item in the content generation guide information; the content to be rewritten includes the initially generated content or the content generated after the previous quality inspection step.

[0086] Optionally, the device may further include:

[0087] After receiving the user's content generation request and determining the content generation requirement information input by the user, selecting a corresponding generation path according to the content type, wherein the generation path is used to define the multiple steps included in content generation and the execution order between the multiple steps, wherein the multiple steps include a preliminary generation step and at least one quality inspection step.So that the first AI generation model is called during the preliminary generation step according to the generation path, and the second AI generation model is called during the quality check step according to the generation path after the preliminary generation of the content is completed.

[0088] Different content types may correspond to different generation paths, and the number and / or order of quality check steps may differ in different generation paths.

[0089] If multiple target languages ​​are required, the target content corresponding to the first language is generated first, and then the third AI generation model corresponding to other languages ​​is called to perform the preliminary generation of the content corresponding to other languages. The quality check step is performed according to the generation path. If the target quality check item fails, the target content corresponding to other languages ​​is regenerated through the second AI generation model (page 10 / 13, CN 121351846 A). The third AI generation model has multiple models, each corresponding to a different target language.

[0090] The third prompt information corresponding to the third AI generation model includes the target content corresponding to the first language, the content generation requirement information input by the user, and the content generation guide information corresponding to other languages.

[0091] The target quality inspection item includes character count inspection;

[0092] The second AI generation model is specifically used to, if the number of characters in the initially generated content exceeds a threshold, or the number of characters in the content generated after the previous quality inspection step exceeds a threshold, then according to the second prompt information, with the goal of reducing the number of characters, rewrite the initially generated content or the content generated after the previous quality inspection step to generate content that is qualified in terms of character count.

[0093] Alternatively, the target quality inspection item includes sensitive word filtering;

[0094] The second AI generation model is specifically used to, if the initially generated content includes sensitive words, or the content generated after the previous quality inspection step includes sensitive words, according to the second prompt information, with the goal of replacing the sensitive words, rewrite the initially generated content or the content generated after the previous quality inspection step to generate content that is qualified in terms of sensitive words.

[0095] During the initial generation of content, the generated prompt information also includes the content type and example information corresponding to the target language in the sample library.

[0096] Corresponding to Embodiment 2, this application embodiment also provides a content generation apparatus, which may include:

[0097] a demand information determination unit, configured to determine the user's content generation demand information through a target interface, wherein the content generation demand information includes content description text, the type of content to be generated, and at least one target language information required;

[0098] A request submission unit is used to submit the request and the content generation requirement information to the server after receiving a content generation request from a user, so that the server can generate the target content according to the method described in the aforementioned embodiment 1;

[0099] A display unit is used to display the target content generated by the server.

[0100] In addition, this application embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method described in any one of the aforementioned method embodiments.

[0101] And an electronic device, including:

[0102] one or more processors; and

[0103] a memory associated with the one or more processors, the memory being used to store program instructions, which, when read and executed by the one or more processors, execute the steps of the method described in any one of the aforementioned method embodiments.

[0104] A computer program product, including a computer program / computer-executable instructions, which, when executed by a processor in an electronic device, implement the steps of the method described in the aforementioned method embodiments.

[0105] FIG6 exemplarily illustrates the architecture of an electronic device, which may specifically include a processor 610, a video display adapter 611, a disk drive 612, an input / output interface 613, a network interface 614, and a memory 620. The processor 610, video display adapter 611, disk drive 612, input / output interface 613, network interface 614, and memory 620 can communicate with each other via a communication bus 630.

[0106] The processor 610 may be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., to execute relevant programs to implement the technical solutions provided in this application.

[0107] The memory 620 may be implemented using ROM (Read Only Memory), RAM (Random Access Memory), static storage devices, dynamic storage devices, etc. The memory 620 may store the operating system 621 for controlling the operation of the electronic device 600, and the basic input / output system (BIOS) for controlling the low-level operations of the electronic device 600. Additionally, it may store a web browser 623, a data storage management system 624, and content generation...System 625, etc. The above-mentioned content generation system 625 can be the application program that specifically implements the aforementioned steps in the embodiments of this application. In short, when the technical solution provided by this application is implemented by software or firmware, the relevant program code is stored in memory 620 and is called and executed by processor 610.

[0108] Input / output interface 613 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0109] Network interface 614 is used to connect communication modules (not shown in the figure) to realize communication interaction between this device and other devices. The communication module can realize communication through wired means (e.g., USB, network cable, etc.) or through wireless means (e.g., mobile network, WIFI, Bluetooth, etc.).

[0110] Bus 630 includes a pathway for transmitting information between various components of the device (e.g., processor 610, video display adapter 611, disk drive 612, input / output interface 613, network interface 614, and memory 620).

[0111] It should be noted that although the above-described device only shows processor 610, video display adapter 611, disk drive 612, input / output interface 613, network interface 614, memory 620, bus 630, etc., in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include components necessary for implementing the present application, and not necessarily all components shown in the figures.

[0112] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of this application.

[0113] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the method.The description of the embodiments is only partially provided. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0114] The content generation method and electronic device provided by this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and its core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. In summary, the content of this specification should not be construed as a limitation of this application. Instruction manual, page 13 / 13, 16 CN 121351846 A, Figure 1, Figure 2; Instruction manual, Figure 1 / 3, page 17 CN 121351846 A, Figure 3, Figure 4; Instruction manual, Figure 2 / 3, page 18 CN 121351846 A, Figure 5, Figure 6; Instruction manual, Figure 3 / 3, page 19 CN 121351846 A, Content generation method and electronic equipment. The embodiment of the invention discloses a content generation method and electronic equipment, and the method comprises the steps: receiving a content generation request from a user, and determining the content generation demand information inputted by the user; generating first prompt information, and calling the first AI generation model to perform preliminary generation of contents; performing quality...inspection on the preliminarily generated content, if a target quality inspection item does not pass, generating second prompt information, and calling a second AI generation model to perform content regeneration to obtain target content; the second AI generation model is a special model for the target quality inspection item. Through the embodiment of the invention, high-quality content generation can be completed under the condition of not depending on the understanding of the user on the prompt word engineering. Abstract

Claims

1. A content generation method characterized by, The method comprises: receiving a content generation request of a user, and determining content generation requirement information input by the user, the content generation requirement information comprising content description text, a required content type, and required at least one target language information; generating first prompt information, and calling a first AI generation model to perform preliminary generation of content; wherein the first prompt information comprises the content generation requirement information input by the user, and content generation guideline information extracted from a database according to the content type and the target language information; the content generation guideline is generated according to professional knowledge provided by an expert user corresponding to the target language for the content type, and is pre-stored in the database; performing quality check on the preliminarily generated content, and if a target quality check item fails, generating second prompt information, and calling a second AI generation model to perform re-generation of content to obtain target content; wherein the second AI generation model is a special model for the target quality check item, the second prompt information comprises to-be-revised content, the content generation requirement information input by the user, and information related to the target quality check item in the content generation guideline information; the to-be-revised content comprises the preliminarily generated content or content generated after a previous quality check step.

2. The method of claim 1, further comprising: after receiving the content generation request of the user and determining the content generation requirement information input by the user, selecting a corresponding generation path according to the content type, the generation path being used to define a plurality of steps included in content generation and an execution order between the plurality of steps, the plurality of steps comprising a preliminary generation step and at least one quality check step, so as to call the first AI generation model in the process of executing the preliminary generation step according to the generation path, and call the second AI generation model in the process of executing the quality check step according to the generation path after completing the preliminary generation of content.

3. The method of claim 2, wherein different content types correspond to different generation paths, and the number and / or order of quality check steps are different in different generation paths.

4. The method of claim 1, wherein if the required target language is multiple, the target content corresponding to a first language is generated first, a third AI generation model corresponding to each of the other languages is called to perform preliminary generation of content corresponding to each of the other languages, and a quality check step is executed according to the generation path, if a target quality check item fails, the second AI generation model is used to re-generate to obtain target content corresponding to each of the other languages; the third AI generation model is multiple and corresponds to different target languages respectively; the third prompt information corresponding to the third AI generation model comprises the target content corresponding to the first language, the content generation requirement information input by the user, and content generation guideline information corresponding to the other languages. ​ ​ ​ 5. The method of claim 1 or 4, wherein the target quality check item comprises a character number check. The second AI generation model is specifically configured to, if the character number of the preliminarily generated content exceeds a threshold value, or if the character number of the content generated after the last quality check step exceeds a threshold value, rewrite the preliminarily generated content or the content generated after the last quality check step according to the second prompt information, with the goal of shortening the character number, to generate content that is qualified in terms of character number.

6. The method of claim 1 or 4, wherein the target quality check item comprises sensitive word filtering. The second AI generation model is specifically configured to, if the preliminarily generated content includes a sensitive word, or if the content generated after the last quality check step includes a sensitive word, rewrite the preliminarily generated content or the content generated after the last quality check step according to the second prompt information, with the goal of replacing the sensitive word, to generate content that is qualified in terms of sensitive words.

7. The method of claim 1, wherein in the process of preliminarily generating the content, the generated prompt information further includes example information corresponding to the content type and the target language in the sample library. The method comprises: determining content generation requirement information of a user through a target interface, the content generation requirement information including a content description text, a content type to be generated, and at least one target language information required; after receiving a content generation request issued by the user, submitting the request and the content generation requirement information to a server, so that the server generates target content according to the method of any one of claims 1 to 7; 8. A content generation method characterized by, displaying the target content generated by the server. The program is executed by a processor to implement the steps of the method of any one of claims 1 to 8. The method comprises: one or more processors; 9. A computer-readable storage medium having stored thereon a computer program, characterized in that, and 10. An electronic device, comprising: a memory associated with the one or more processors, the memory being configured to store program instructions, the program instructions being executed by the one or more processors to perform the steps of the method of any one of claims 1 to 8. The computer program / computer executable instructions are executed by a processor in an electronic device to implement the steps of the method of any one of claims 1 to 8. ​ ​ 11. A computer program product comprising computer program / computer executable instructions, characterized in that, ​

Images