Cboth case generation method and device and program product
By obtaining topics and copywriting materials, using AI models to generate copywriting, and performing quality inspection and adjustment, the problem of poor copywriting quality of AI models is solved, and the quality and accuracy of copywriting is improved.
Patent Information
- Application Number
- CN202510127627.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-27
- Publication Date
- 2025-05-27
AI Technical Summary
Existing AI models may have poor quality problems when generating copywriting, such as AI illusion, where the generated copywriting is inconsistent with real-world facts or user input, and the generated copywriting quality is uneven.
By obtaining the generated topic and determining the first-level tag, obtaining the corresponding copy material, using the AI model to generate copy, and detecting the generation quality through methods such as dependent syntax analysis and similarity comparison. If the conditions do not meet the requirements, the copy will be adjusted.
It improves the quality and accuracy of copywriting generation, reduces the occurrence of AI hallucinations, and ensures that the generated copywriting meets quality standards.
Smart Images
Figure CN120045703A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computers, and in particular, to a copywriting generation method, apparatus, and program product. Background Art
[0002] Currently, with the continuous development of artificial intelligence (AI) technology, more and more industries or fields are beginning to use AI technology to improve the level of intelligence. Natural language processing (NLP), as an important application direction of AI technology, can understand, process, or generate human natural language.
[0003] Generally, a user can input a prompt to an AI model with NLP capabilities, and thus can obtain a natural language output result that conforms to the content of the prompt. For example, applied to the insurance industry, for the user's need to write an insurance promotion copy, the user can input the theme of the copy to be generated into the AI model, so as to use the AI model to generate an insurance promotion copy that conforms to the theme.
[0004] However, in the actual application process, the copy generated by using the AI model may have poor quality. For example, the AI model may have AI hallucinations, that is, the copy generated by the AI model is inconsistent with the facts of the real world or the user's input. Summary of the Invention
[0005] Embodiments of this application provide a copywriting generation method to improve the quality of copy generated by using an AI model. In addition, embodiments of this application also provide a corresponding copywriting generation apparatus, computer program product, computing device, and computer-readable storage medium.
[0006] In a first aspect, embodiments of this application provide a copywriting generation method, including: obtaining a generation theme; determining a first first-level tag according to the generation theme, where the first first-level tag is one of multiple first-level tags included in a database; obtaining a first copywriting material corresponding to the first first-level tag according to the database; generating a first copy according to the first copywriting material by using a first artificial intelligence (AI) model; determining whether the generation quality of the first copy meets a quality condition; and if the generation quality does not meet the quality condition, adjusting the first copy.
[0007] In a possible implementation manner, determining whether the generation quality of the first copywriting meets the quality condition includes: using a second AI model to perform dependency syntax analysis on a first sentence, where the dependency syntax analysis is used to analyze the dependency relationships between words in the sentence, and the first sentence is a sentence in the first copywriting; determining whether the dependency relationships between words in the first sentence meet the quality condition; if the generation quality does not meet the quality condition, adjusting the first copywriting, including: if the dependency relationships between words in the first sentence do not meet the quality condition, using the second AI model to adjust the first sentence in the first copywriting.
[0008] In a possible implementation manner, determining whether the generation quality of the first copywriting meets the quality condition includes: using the second AI model to compare the similarity between a second sentence and a third sentence, where the second sentence is a sentence in the first copywriting and the third sentence is a sentence in the first copywriting material; determining whether the similarity meets the quality condition, where the quality condition includes that the value of the similarity satisfies a first value range; if the generation quality does not meet the quality condition, adjusting the first copywriting, including: if the value of the similarity does not satisfy the first value range, using the second AI model to adjust the second sentence in the first copywriting.
[0009] In a possible implementation manner, determining whether the generation quality of the first copywriting meets the quality condition includes: determining the occurrence frequency of a first character included in the first copywriting in the first copywriting; determining whether the occurrence frequency meets the quality condition, where the quality condition includes that the occurrence frequency satisfies a second value range; if the generation quality does not meet the quality condition, adjusting the first copywriting, including: if the occurrence frequency does not satisfy the second value range, using the first AI model to adjust the first copywriting.
[0010] In a possible implementation manner, the first first-level label corresponds to N second-level labels, the first copywriting material includes multiple copywriting paragraph samples respectively corresponding to each of the N second-level labels, the first copywriting includes M copywriting paragraphs, and the copywriting paragraphs correspond to one of the N second-level labels, where N and M are positive integers; generating the first copywriting according to the first copywriting material using a first AI model includes: using the first AI model to generate the M copywriting paragraphs, and the copywriting paragraphs are generated based on the multiple copywriting paragraph samples corresponding to the second-level label corresponding to the copywriting paragraphs.
[0011] In a possible implementation, the method further includes: determining a second first-level tag according to the generated topic, where the second first-level tag is one of the multiple first-level tags included in the database; obtaining second copywriting materials corresponding to the second first-level tag according to the database; and generating a second copywriting according to the first copywriting and the second copywriting materials by using a first AI model, where the second copywriting is after the first copywriting.
[0012] In a possible implementation, the method further includes: obtaining a generation style, where the generation style is used to indicate the content style of the first copywriting; and the step of generating the first copywriting according to the first copywriting materials by using the first AI model includes: generating the first copywriting by using the first AI model according to the generation style and the first copywriting materials.
[0013] In a possible implementation, the method further includes: obtaining a target article; parsing the target article to obtain at least one copywriting material; and updating the database according to the at least one copywriting material.
[0014] In a second aspect, an embodiment of the present application provides a copywriting generation device, including: an obtaining module, configured to obtain a generation topic; the obtaining module is further configured to obtain first copywriting materials corresponding to a first first-level tag according to a database, where the first first-level tag is one of the multiple first-level tags included in the database; a determining module, configured to determine the first first-level tag according to the generation topic; the determining module is further configured to determine whether the generation quality of the first copywriting meets a quality condition; a generating module, configured to generate a first copywriting by using a first artificial intelligence (AI) model according to the first copywriting materials; and an adjusting module, configured to adjust the first copywriting if the generation quality does not meet the quality condition.
[0015] In a possible implementation, the determining module is specifically configured to: perform dependency syntactic analysis on a first sentence by using a second AI model, where the dependency syntactic analysis is used to analyze the dependency relationship between words in the sentence, and the first sentence is a sentence in the first copywriting; and determine whether the dependency relationship between words in the first sentence meets the quality condition; the adjusting module is specifically configured to: if the dependency relationship between words in the first sentence does not meet the quality condition, adjust the first sentence in the first copywriting by using the second AI model.
[0016] In a possible implementation manner, the determining module is specifically configured to: use the second AI model to compare the similarity between a second sentence and a third sentence, where the second sentence is a sentence in the first copywriting, and the third sentence is a sentence in the first copywriting material; determine whether the similarity meets the quality condition, where the quality condition includes that the value of the similarity satisfies a first value range; the adjusting module is specifically configured to: if the value of the similarity does not satisfy the first value range, use the second AI model to adjust the second sentence in the first copywriting.
[0017] In a possible implementation manner, the determining module is specifically configured to: determine the occurrence frequency of the first characters included in the first copywriting in the first copywriting; determine whether the occurrence frequency meets the quality condition, where the quality condition includes that the occurrence frequency satisfies a second value range; the adjusting module is specifically configured to: if the occurrence frequency does not satisfy the second value range, use the first AI model to adjust the first copywriting.
[0018] In a possible implementation manner, the first first-level tag corresponds to N second-level tags, the first copywriting material includes multiple copywriting paragraph samples respectively corresponding to each of the N second-level tags, the first copywriting includes M copywriting paragraphs, and the copywriting paragraphs correspond to one of the N second-level tags, where N and M are positive integers; the generating module is specifically configured to: use the first AI model to generate the M copywriting paragraphs, and the copywriting paragraphs are generated based on the multiple copywriting paragraph samples corresponding to the second-level tag corresponding to the copywriting paragraph.
[0019] In a possible implementation manner, the determining module is further configured to determine a second first-level tag according to the generation theme, where the second first-level tag is one of the multiple first-level tags included in the database; the obtaining module is further configured to obtain, according to the database, the second copywriting material corresponding to the second first-level tag; the generating module is further configured to generate a second copywriting after the first copywriting according to the first copywriting and the second copywriting material by using the first AI model.
[0020] In a possible implementation manner, the obtaining module is further configured to obtain a generation style, where the generation style is used to indicate the content style of the first copywriting; the generating module is specifically configured to: generate the first copywriting by using the first AI model according to the generation style and the first copywriting material.
[0021] In a possible implementation manner, the obtaining module is further configured to obtain a target article; the apparatus further includes an analysis module, and the analysis module is configured to analyze the target article to obtain at least one copywriting material; the apparatus further includes an updating module, and the updating module is configured to update the database according to the at least one copywriting material.
[0022] In a third aspect, an embodiment of the present application further provides a computing device, and the computing device may include a processor and a memory:
[0023] The memory is used to store a computer program;
[0024] The processor is configured to execute the method described in the first aspect and any implementation manner in the first aspect according to the computer program.
[0025] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, and the computer-readable storage medium is used to store a computer program, and the computer program is used to execute the method described in the first aspect and any implementation manner in the first aspect.
[0026] In a fifth aspect, an embodiment of the present application further provides a computer program product including instructions, and when it runs on a computing device, it causes the computing device to execute the method described in the first aspect and any implementation manner in the first aspect.
[0027] In the above implementation manner of the embodiment of the present application, a generation theme is obtained, and according to the generation theme, a first first-level tag is determined, where the first first-level tag is one of the multiple first-level tags included in the database. Subsequently, according to the database, the first copywriting material corresponding to the first first-level tag is obtained, and further, according to the first copywriting material, a first copywriting is generated by using a first AI model. On this basis, it is determined whether the generation quality of the first copywriting meets the quality condition. If the generation quality does not meet the quality condition, the first copywriting is adjusted. In this way, the first first-level tag is determined according to the generation theme of the user, and then the first copywriting can be generated by using the first AI model based on the first copywriting material corresponding to the first first-level tag. Compared with directly generating a copywriting based on the generation theme, the first copywriting material, as a reference material for generating the first copywriting, expands the breadth and comprehensiveness of the information that the first AI model can access during the process of generating the first copywriting, improves the generation quality and accuracy of the generated first copywriting, and reduces the occurrence of the AI hallucination phenomenon. In addition, after the first copywriting is generated, it is determined whether the generation quality of the first copywriting meets the quality condition, that is, the quality of the first copywriting is detected. And when the generation quality does not meet the quality condition, the first copywriting is adjusted, effectively avoiding problems such as poor quality of the first copywriting and non-compliance of the generated first copywriting, and further ensuring the generation quality of the first copywriting. Brief Description of the Drawings
[0028] To more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments recorded in the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.
[0029] Figure 1 It is a schematic structural diagram of a copywriting generation system in an embodiment of the present application;
[0030] Figure 2 It is a schematic flowchart of a copywriting generation method in an embodiment of the present application;
[0031] Figure 3 It is a schematic structural diagram of a copywriting generation device in an embodiment of the present application;
[0032] Figure 4 It is a schematic hardware structure diagram of a computing device in an embodiment of the present application. Detailed Description of the Embodiments
[0033] Refer to Figure 1 , which is a schematic diagram of a copywriting generation system provided by the present application. As Figure 1 shown, the copywriting generation system 10 includes a processor and a memory. Figure 1 Taking the processor 100 and the memory 201 as an example, in the actual application scenario, the copywriting generation system 10 may include a larger number of memories or processors.
[0034] The processor 100 and the memory 201 can be connected through a bus. This bus can be, for example, a peripheral component interconnect express (PCIE) bus. Or, the processor 100 and the memory 201 can also be connected through a network. This network can be, for example, a local area network (LAN), a wide area network (WAN), etc. in terms of coverage, and can be a wired network or a wireless network, etc. in terms of connection method.
[0035] Among them, the processor 100 refers to a device with data processing and control capabilities. For example, the processor 100 can be a graphics processing unit (GPU), a central processing unit (CPU), etc., and no limitation is imposed on this.
[0036] The memory 201 refers to a device with file storage capabilities, such as a semiconductor memory, a magnetic memory, an optical memory, etc., and there is no limitation on this. Additionally, a database can be stored on the memory 201 for data management and data query, and there is no limitation on this.
[0037] Under normal circumstances, users can use a pre-trained AI model to write copywriting. For example, in the insurance industry, users can input the theme of the insurance promotion copywriting to be generated into the AI model to obtain a copywriting output result that matches the copywriting theme.
[0038] However, in the actual application process, the copywriting generated by the AI model may be of poor quality. For example, relying solely on the theme of the copywriting, the amount of information that the AI model can obtain may be low. Furthermore, the AI model cannot correctly respond to the user's demand for copywriting generation, resulting in the copywriting generated by the AI model being inconsistent with the facts of the real world or the user's input, that is, the AI model has hallucinations. Another example is that even when faced with the same input request, the copywriting results output by the AI model may be different each time, and the quality of the copywriting is correspondingly uneven.
[0039] Based on this, the processor 100 obtains the generation theme and determines the first first-level label according to the generation theme, where the first first-level label is one of the multiple first-level labels included in the database. Subsequently, the processor 100 obtains the first copywriting material corresponding to the first first-level label according to the database, and then, according to the first copywriting material, uses the first AI model to generate the first copywriting. On this basis, the processor 100 determines whether the generation quality of the first copywriting meets the quality condition. If the generation quality does not meet the quality condition, the first copywriting is adjusted. In this way, the processor 100 determines the first first-level label according to the user's generation theme, and then can generate the first copywriting using the first AI model based on the first copywriting material corresponding to the first first-level label. Compared with directly generating copywriting based on the generation theme, the first copywriting material, as a reference material for generating the first copywriting, expands the breadth and comprehensiveness of the information that the first AI model can access during the process of generating the first copywriting, improves the generation quality and accuracy of the generated first copywriting, and reduces the occurrence of the AI hallucination phenomenon. In addition, after generating the first copywriting, it is determined whether the generation quality of the first copywriting meets the quality condition, that is, the quality of the first copywriting is detected. And in the case where the generation quality does not meet the quality condition, the first copywriting is adjusted, effectively avoiding problems such as poor quality of the first copywriting and non-compliance of the generated first copywriting, and further ensuring the generation quality of the first copywriting.
[0040] To make the above objects, features, and advantages of the present application more obvious and understandable, the following will exemplarily illustrate various non-limiting embodiments in the embodiments of the present application with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.
[0041] Refer to Figure 2 , Figure 2 which shows a schematic flowchart of a copywriting generation method in an embodiment of the present application. This method can be applied to Figure 1 the copywriting generation system 10 shown in Figure 1 or can be applied to other applicable copywriting generation systems. Hereinafter, taking the copywriting generation in the copywriting generation system 10 shown in Figure 2 as an example for illustration. As
[0042] S201: The processor 100 obtains a generation theme.
[0043] The processor 100 first obtains a generation theme to generate a copywriting that conforms to the generation theme according to the generation theme. This copywriting can be, for example, an insurance promotion copywriting applied to the insurance industry, or other copywriting applied to other industries, which is not limited herein.
[0044] Exemplarily, the generation theme is used to generally describe the content requirements of the user for the target-generated copywriting. For example, the generation theme can include one or more pieces of information such as the title of the copywriting, the content summary of the copywriting, the product name associated with the copywriting, etc., which is not limited herein.
[0045] For example, applied to the insurance industry, in order to generate a promotion copywriting for the target insurance product, the generation theme can include one or more of the type of the target insurance product, the name of the target insurance product, the content type of the copywriting, etc. Among them, the content type of the copywriting can specifically be one content type among multiple predefined content types, for example, it can be an evaluation introduction of a single insurance product.
[0046] S202: According to the generation theme, the processor 100 determines a first first-level tag, and the first first-level tag is one of the multiple first-level tags included in the database.
[0047] Based on the obtained generation theme, the processor 100 can determine a first-level tag that matches the generation theme, so that the processor 100 expands the amount of information according to the first-level tag to better match the needs of the user-generated copywriting. Among them, a plurality of first-level tags are stored in a database (for example, the database in the memory 201), and the first-level tag is one of the plurality of first-level tags included in the database.
[0048] Among them, the first-level tag, as a first-level tag that matches the generation theme, can be understood as an aspect or an angle of the copywriting that needs to be generated for the above generation theme. For example, when applied to the insurance industry, the first-level tag can be the basic introduction of the insurance product indicated by the generation theme, or it can be the specific advantages of the insurance product indicated by the generation theme. In other embodiments, the first-level tag can also be replaced by other defined terms such as the first-level outline, which is not limited herein.
[0049] It should be noted that in the database, the generation theme can also match more first-level tags. For example, the generation theme can match a second-level tag. Furthermore, the processor 100 can also determine the second-level tag according to the generation theme.
[0050] Furthermore, the correspondence between the generation theme and the first-level tag (or other more first-level tags) can be pre-configured in the processor 100 or pre-stored in the memory 201. For example, when applied to the insurance industry, the first-level tag can be the basic introduction of the insurance product, and then, the processor 100 can determine the first-level tag according to the correspondence between the generation theme and the basic introduction of the insurance product indicated by the generation theme.
[0051] Or, the generation theme can include the identifier of the first-level tag (or the identifiers of other more first-level tags), that is, the user can customize the first-level tag so that the generated target copywriting covers multiple different aspects indicated by the user. Furthermore, the processor 100 can determine the first-level tag according to the identifier of the first-level tag.
[0052] It should be noted that the above implementation example of the processor 100 determining the first-level tag according to the generation theme is only for illustrative purposes. In actual applications, the processor 100 can also determine the first-level tag based on other methods, which is not limited herein.
[0053] S203: According to the database, the processor 100 obtains the first copywriting material corresponding to the first-level tag.
[0054] Each first-level tag included in the database can correspond to a copywriting material. For example, the first first-level tag can correspond to the first copywriting material. Furthermore, the processor 100 can obtain the first copywriting material corresponding to the first first-level tag in the database, with a view to generating a target copy using the first copywriting material.
[0055] Specifically, the copywriting material can be a reference material for generating the target copy. Each copywriting material can include one or more reference copies. The reference copies included in the copywriting material can be high-quality copies pre-written in association with the first-level tag. Furthermore, the processor 100 can generate the target copy using the reference copies included in the copywriting material, that is, use the pre-written high-quality copy that matches the generation theme as a reference for generating the target copy, thereby significantly improving the quality of the generated target copy.
[0056] The following introduces an implementation example of the processor 100 constructing or updating the database according to the target article.
[0057] The processor 100 can obtain the target article, and then by parsing the target article, obtain at least one copywriting material included in the target article. For example, applied to the insurance industry, the target article can include the content of the basic introduction of a certain insurance product and the content of the advantages of a certain insurance product. The processor 100 parses the target article and can obtain the above two different copywriting materials.
[0058] Furthermore, the processor 100 can construct or update the database according to the at least one copywriting material obtained by parsing.
[0059] Exemplarily, if there are multiple first-level tags and the corresponding copywriting materials in the database, the processor 100 can update the database. Specifically, the processor 100 can determine the similarity between each parsed copywriting material and each first-level tag in the database. For example, the processor 100 can determine the above similarity through Embedding technology. If the similarity between a parsed copywriting material and a certain first-level tag in the database meets the similarity condition (the similarity condition can be, for example, that the similarity is greater than the similarity threshold), the database can be updated, that is, the processor 100 aggregates the copywriting material corresponding to the first-level tag and the parsed copywriting material, so that in the updated database, the first-level tag corresponds to the aggregated copywriting material. If the similarity between a parsed copywriting material and any first-level tag in the database does not meet the similarity condition, the processor 100 can determine a new first-level tag, and the new first-level tag corresponds to the parsed copywriting material. Alternatively, the processor 100 can cluster multiple copywriting materials that do not meet the similarity condition to obtain one or more new first-level tags, and each new first-level tag corresponds to an aggregated copywriting material.
[0060] Referring to the above implementation method, the processor 100 can construct a database according to at least one parsed copywriting material. For example, the processor 100 can determine a new first-level tag, and the new first-level tag corresponds to the parsed copywriting material. Then, the processor 100 stores the corresponding relationship between the first-level tag and the copywriting material in the database. Alternatively, the processor 100 can cluster multiple copywriting materials to obtain one or more new first-level tags, and each new first-level tag corresponds to an aggregated copywriting material. Then, the processor 100 stores the above corresponding relationship in the database.
[0061] It should be noted that the description of the processor 100 constructing or updating the database according to the target article is an exemplary illustration. In actual application, the database can also be constructed or updated based on other methods, or the database can be updated or constructed by other processors, or the database can be pre-configured by developers, and this is not limited.
[0062] Optionally, the first first-level tag can correspond to N second-level tags, where N is a positive integer. Furthermore, the first copywriting material corresponding to the first first-level tag can include multiple copywriting paragraph samples respectively corresponding to the N second-level tags. In this way, by further dividing the first first-level tag into N second-level tags, in the process of generating the target copywriting, the target copywriting can be gradually generated according to the second-level tags, reducing the granularity of the target copywriting generation and also improving the quality of the target copywriting. It can be understood that the copywriting paragraph sample is a reference sample for each copywriting paragraph used to generate the target copywriting.
[0063] For example, when applied to the insurance industry, for the first-level label of the advantages of a certain insurance product, it can include multiple second-level labels, and each second-level label can be the advantages of the insurance product in different dimensions (such as price, benefits, etc.).
[0064] Correspondingly, the processor 100 can also parse and obtain multiple copywriting paragraph samples included in the target article with reference to the above description, and update or construct a database based on the parsed copywriting paragraph samples, which will not be elaborated here.
[0065] S204: According to the first copywriting material, the processor 100 uses the first AI model to generate the first copywriting.
[0066] According to the first copywriting material corresponding to the first first-level label, the processor 100 can use the first AI model to generate the first copywriting, that is, the target copywriting. Specifically, the processor 100 can input the first copywriting material as a prompt into the first AI model, so that the first AI model generates and outputs the first copywriting with reference to the first copywriting material.
[0067] In this way, compared with directly generating the first copywriting based on the generation theme, the first copywriting material, as the reference material for generating the first copywriting, expands the breadth and comprehensiveness of the information that the first AI model can access during the process of generating the first copywriting, improves the generation quality and accuracy of the generated first copywriting, and reduces the occurrence of the AI hallucination phenomenon.
[0068] Among them, the first AI model can be an AI model with NLP capabilities. Therefore, the first AI model can also be called an NLP model. In addition, the first AI model can also be an AI model trained based on a large amount of natural language data, that is, an AI large model or an NLP large model, which is not limited here.
[0069] It should be noted that the prompt input into the first AI model can include not only the first copywriting material, but also content such as the generation theme and the first first-level label, which is not limited here.
[0070] Optionally, the processor 100 can also obtain the generation style, where the generation style can indicate the content style of the generated first copywriting. For example, when applied to the insurance industry, the generation style can include the character setting, tone of the person in the content of the first copywriting, the media platform to which the first copywriting is applied, and the user portrait of the audience group of the first copywriting. Furthermore, the processor 100 can generate the first copywriting according to the generation style and the first copywriting material.
[0071] According to the description of step S202, the generated topic can also match more first-level tags. That is, the processor 100 can also determine a second first-level tag according to the generated topic (or can also determine more first-level tags). The following makes an exemplary description of the content where the processor 100 uses the first AI model to generate the first copywriting and the second copywriting when there are at least two first-level tags that match the generated topic.
[0072] Similar to step S203, the processor 100 can obtain the second copywriting material corresponding to the second first-level tag according to the database. Furthermore, the processor 100 can use the first AI model to generate the second copywriting according to the second copywriting material.
[0073] Furthermore, if the second copywriting is after the first copywriting in terms of content order (similarly when the first copywriting is after the second copywriting), then when generating the second copywriting, the processor 100 can use the generated first copywriting and the second copywriting material, and use the first AI model to generate the second copywriting. In this way, since the first copywriting before the second copywriting is considered, the quality of the connection of the content of the second copywriting with the first copywriting will be better.
[0074] Optionally, when generating the first copywriting, the processor 100 can also generate the first copywriting based on the second first-level tag and the first copywriting material. In this way, the quality of the connection of the content of the first copywriting with the second copywriting will be better. Or, after the second copywriting is generated, the processor 100 can adjust the first copywriting based on the second copywriting to make the connection between the first copywriting and the second copywriting smoother.
[0075] Optionally, before generating the first copywriting and the second copywriting, the first copywriting material and the second copywriting material can be input into the first AI model in advance, so that the first AI model can have a higher-level understanding of the joint task of generating the first copywriting and the second copywriting in advance. Furthermore, during the process of generating the first copywriting or the second copywriting, the content of the two different copywritings will not be fragmented and the connection will be more fluent.
[0076] Optionally, the generated topic can also match more first-level tags. Furthermore, the order of the content between the copywritings corresponding to different first-level tags can be indicated by the generated topic in advance. That is, the generated topic can indicate a copywriting framework, and the copywriting framework indicates the order of the copywritings corresponding to different first-level tags. In addition, this copywriting framework can be user-defined or determined by other content in the generated topic, and this is not limited.
[0077] It should be noted that the content of the above-mentioned processor 100 generating the first copywriting and the second copywriting using the first AI model is only for exemplary illustration. In actual applications, the processor can also generate the first copywriting and the second copywriting according to other implementation methods, and this is not limited herein.
[0078] According to the description of step S202, the first-level label can correspond to N second-level labels, and the first copywriting material can include multiple copywriting paragraph samples respectively corresponding to each of the N second-level labels. Furthermore, the first copywriting generated by the processor 100 using the first AI model can include M copywriting paragraphs, where M is a positive integer. Moreover, each of the M copywriting paragraphs in the M copywriting paragraphs corresponds to one of the N second-level labels, and the copywriting paragraph is generated from multiple copywriting paragraph samples corresponding to the second-level label corresponding thereto. Alternatively, the copywriting paragraph can be generated from one of the multiple copywriting paragraph samples corresponding to the second-level label corresponding thereto.
[0079] Optionally, the value of M can be the same as the value of N, that is, the first copywriting includes the copywriting paragraphs corresponding to each second-level label. The value of M can also be less than the value of N, that is, the first copywriting includes the copywriting paragraphs corresponding to some second-level labels. The value of M can also be greater than the value of N, that is, there may be two different copywriting paragraphs in the first copywriting that are generated based on the copywriting paragraph samples corresponding to the same second-level label.
[0080] Optionally, for two adjacent or close copywriting paragraphs in the first copywriting, that is, the first copywriting paragraph and the second copywriting paragraph, the first copywriting paragraph is generated based on the first copywriting paragraph sample, and the second copywriting paragraph is generated based on the second copywriting paragraph sample. The first copywriting paragraph sample and the second copywriting paragraph sample may be from the same article. To avoid the similarity between the adjacent or close first copywriting paragraph and the second copywriting paragraph and the article being too high, the processor 100 can adjust the first copywriting paragraph or the second copywriting paragraph to be generated based on other copywriting paragraph samples.
[0081] S205: The processor 100 determines whether the generation quality of the first copywriting meets the quality condition.
[0082] S206: If the generation quality does not meet the quality condition, the processor 100 adjusts the first copywriting.
[0083] After the processor 100 generates the first copywriting using the first AI model, the processor 100 can also perform quality detection on the first copywriting. And, in the case where the generation quality of the first copywriting does not meet the quality conditions, the processor 100 can adjust the first copywriting to meet the quality conditions. In this way, problems such as poor quality of the first copywriting obtained by the user and non-compliance of the generated first copywriting are effectively avoided, and the generation quality of the first copywriting is further guaranteed.
[0084] It should be noted that after the processor 100 generates the second copywriting or more copywritings, it can also perform quality detection on the second copywriting and / or more copywritings in a similar manner, which will not be elaborated here.
[0085] In this embodiment, the following implementation examples of the processor 100 performing quality detection on the first copywriting and adjusting the first copywriting are provided.
[0086] As a first implementation example, the processor 100 can use a second AI model to perform dependency syntactic analysis on the first sentence in the first copywriting. Among them, dependency syntactic analysis is specifically used to analyze the dependency relationship between words in a sentence. For example, in a sentence containing a subject, a predicate, and an object, such as the sentence "I love you", the subject "I" and the object "you" depend on the predicate verb "love". In addition, the second AI model can be an AI model with the ability of dependency syntactic analysis and processing, or the second AI model can also have other NLP capabilities, which are not limited here.
[0087] Based on this, the processor 100 can determine whether the dependency relationship between words in the first sentence meets the quality conditions. Among them, the quality conditions can be, for example, that the number of words depending on the predicate verb is lower than the number threshold. Usually, the number of words depending on the predicate verb in a sentence with qualified quality will not be too low. However, the sentences generated by using an AI model may have the above problems. In this way, by analyzing the dependency relationship between words in the first sentence, it helps to make the first sentence more in line with the writing habits of people.
[0088] Subsequently, if the dependency relationship between words in the first sentence does not meet the quality conditions, the processor can use the second AI model to adjust the first sentence in the first copywriting. For example, the processor 100 can add new words to the sentence so that the number of words depending on the predicate verb reaches the number threshold indicated by the quality conditions. Or, the processor 100 can also adjust the dependency relationship in the sentence, which is not limited here.
[0089] It should be noted that the processor 100 can perform quality detection on the first sentence in the first copywriting, or can perform quality detection based on all or part of the sentences in the first copywriting. The specific detection content can refer to the detection method of the first sentence, which will not be elaborated here.
[0090] As a second implementation example, the processor 100 can use the above-mentioned second AI model to compare the similarity between the second sentence in the first copywriting and the third sentence in the first copywriting material. Exemplarily, the processor 100 can perform dependency syntactic analysis on the second sentence and the third sentence to determine the similarity between them. Generally, the second sentence and the third sentence are similar if the dependency relationships in the second sentence and the third sentence are similar, or if the second sentence and the third sentence contain the same words. The value of the similarity can be, for example, the percentage between the number of words that the second sentence and the third sentence have in common and the total number of words.
[0091] Based on this, the processor 100 can determine whether the above similarity meets the quality condition, where the quality condition can be, for example, that the quantitative value of the similarity meets the first value range, or that the value of the similarity is lower than the similarity threshold. In this way, the situation where the similarity between the sentences in the first copywriting and the sentences in the first copywriting material is too high is effectively avoided.
[0092] Subsequently, if the value of the similarity does not meet the first value range, or if the value of the similarity is lower than the similarity threshold, the processor 100 can use the second AI model to adjust the second sentence in the first copywriting. For example, the processor 100 can replace the words in the second sentence that are repeated with the third sentence with new words so that the similarity after replacement is greater than the similarity threshold. Or, the processor 100 can also adjust the dependency relationship in the second sentence so that there are obvious differences between the dependency relationships of the second sentence and the third sentence.
[0093] It should be noted that the processor 100 can compare the similarity between the second sentence in the first copywriting and all the sentences in the first copywriting material, or the processor 100 can also compare the pairwise similarities between all the sentences in the first copywriting and all the sentences in the first copywriting material respectively. The specific detection content can refer to the detection method of the second sentence, which will not be elaborated here.
[0094] As a third implementation example, the processor 100 can determine the occurrence frequency of the first character included in the first copywriting, where the occurrence frequency is the ratio between the number of occurrences of the first character in the first copywriting and the total number of characters.
[0095] Based on this, the processor 100 can determine whether the above-mentioned occurrence frequency meets the quality condition, where the quality condition can be that the occurrence frequency satisfies the second value range, or the occurrence frequency is higher than the frequency threshold. In this way, through quality detection, the processor 100 can detect the phenomenon of repeatedly using the same character in the first copywriting, avoiding the bad reading experience of "repeating" for readers of the first copywriting.
[0096] Subsequently, if the occurrence frequency does not satisfy the second value range, or the occurrence frequency is higher than the frequency threshold, the processor 100 can adjust the first copywriting using the first AI model or the second AI model. Specifically, the processor 100 can use the first AI model to regenerate the adjusted first copywriting, or the processor 100 can use the second AI model to replace some of the characters that appear in the first copywriting.
[0097] In addition, the processor can also refer to the description in S204 to adjust the connection relationship between different copywritings, which will not be elaborated here.
[0098] It should be noted that the processor 100 can perform quality detection and adjustment on the first copywriting using one or more of the above implementation manners. In actual applications, the processor 100 can also use other implementation manners for quality detection and adjustment, which are not limited herein.
[0099] In addition, the embodiment of the present application also provides a copywriting generation device. Refer to Figure 3 , Figure 3 which shows a schematic structural diagram of a copywriting generation device in the embodiment of the present application. Figure 3 The shown copywriting generation device 300 includes:
[0100] An acquisition module 301, where the acquisition module 301 is used to acquire a generation theme; the acquisition module 301 is further used to acquire the first copywriting material corresponding to the first first-level tag according to the database, and the first first-level tag is one of the multiple first-level tags included in the database;
[0101] A determination module 302, where the determination module 302 is used to determine the first first-level tag according to the generation theme; the determination module 302 is further used to determine whether the generation quality of the first copywriting meets the quality condition;
[0102] A generation module 303, where the generation module 303 is used to generate the first copywriting using the first artificial intelligence AI model according to the first copywriting material;
[0103] An adjustment module 304, where the adjustment module 304 is used to adjust the first copywriting if the generation quality does not meet the quality condition.
[0104] In a possible implementation manner, the determining module 302 is specifically configured to: perform dependency syntactic analysis on the first sentence by using a second AI model, where the dependency syntactic analysis is used to analyze the dependency relationships between words in the sentence, and the first sentence is a sentence in the first copywriting; determine whether the dependency relationships between words in the first sentence meet the quality condition; the adjusting module 304 is specifically configured to: if the dependency relationships between words in the first sentence do not meet the quality condition, adjust the first sentence in the first copywriting by using the second AI model.
[0105] In a possible implementation manner, the determining module 302 is specifically configured to: compare the similarity between a second sentence and a third sentence by using the second AI model, where the second sentence is a sentence in the first copywriting, and the third sentence is a sentence in the first copywriting material; determine whether the similarity meets the quality condition, where the quality condition includes that the value of the similarity satisfies a first value range; the adjusting module 304 is specifically configured to: if the value of the similarity does not satisfy the first value range, adjust the second sentence in the first copywriting by using the second AI model.
[0106] In a possible implementation manner, the determining module 302 is specifically configured to: determine the occurrence frequency of the first character included in the first copywriting in the first copywriting; determine whether the occurrence frequency meets the quality condition, where the quality condition includes that the occurrence frequency satisfies a second value range; the adjusting module 304 is specifically configured to: if the occurrence frequency does not satisfy the second value range, adjust the first copywriting by using the first AI model.
[0107] In a possible implementation manner, the first first-level tag corresponds to N second-level tags, the first copywriting material includes multiple copywriting paragraph samples respectively corresponding to each of the N second-level tags, the first copywriting includes M copywriting paragraphs, and the copywriting paragraph corresponds to one of the N second-level tags, where N and M are positive integers; the generating module 303 is specifically configured to: generate the M copywriting paragraphs by using the first AI model, and the copywriting paragraphs are generated based on the multiple copywriting paragraph samples corresponding to the second-level tag corresponding to the copywriting paragraph.
[0108] In a possible implementation, the determining module 302 is further configured to determine a second first-level tag according to the generated theme, where the second first-level tag is one of the multiple first-level tags included in the database; the obtaining module 301 is further configured to obtain, according to the database, second copywriting materials corresponding to the second first-level tag; the generating module 303 is further configured to generate a second copywriting according to the first copywriting and the second copywriting materials by using a first AI model, and the second copywriting is after the first copywriting.
[0109] In a possible implementation, the obtaining module 301 is further configured to obtain a generation style, where the generation style is used to indicate the content style of the first copywriting; specifically, the generating module 303 is configured to generate the first copywriting according to the generation style and the first copywriting materials by using the first AI model.
[0110] In a possible implementation, the obtaining module 301 is further configured to obtain a target article; the copywriting generating device 300 further includes an analysis module 305 (not shown in the figure), where the analysis module 305 is configured to analyze the target article to obtain at least one copywriting material; the copywriting generating device 300 further includes an updating module 306 (not shown in the figure), where the updating module 306 is configured to update the database according to the at least one copywriting material.
[0111] It should be noted that, for the information interaction, execution process, etc. among the above-mentioned device modules and units, since they are based on the same concept as the method embodiments in the present application, the technical effects brought by them are the same as those of the method embodiments in the present application. For specific content, reference can be made to the description in the method embodiments shown above in the embodiments of the present application, and details are not described herein again.
[0112] In addition, an embodiment of the present application further provides a computing device. Refer to Figure 4 , Figure 4 which shows a schematic hardware structure diagram of a computing device in an embodiment of the present application. As Figure 4 shown, the computing device 400 may include a processor 401 and a memory 402.
[0113] Among them, the memory 402 is used to store a computer program;
[0114] The processor 401 is configured to execute the copywriting generating method described in the above method embodiment according to the computer program.
[0115] In addition, an embodiment of the present application further provides a computer-readable storage medium, where the computer-readable storage medium is used to store a computer program, and the computer program is used to execute the copywriting generating method described in the above method embodiment.
[0116] In addition, an embodiment of the present application further provides a computer program product including instructions. When it runs on a computing device, it causes the computing device to execute the copywriting generation method described in the above method embodiment.
[0117] In the embodiment of the present application, the "first" in names such as "first copywriting" is only used as a name identifier and does not represent the first in order. This rule also applies to "second", "third", etc.
[0118] From the description of the above embodiments, those skilled in the art can clearly understand that all or part of the steps in the above embodiment methods can be implemented by means of software plus a general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as read-only memory (ROM) / RAM, magnetic disk, optical disc, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network communication device such as a router) to execute the methods described in various embodiments or some parts of the embodiments of the present application.
[0119] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device embodiments described above are only illustrative. The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.
[0120] The above description is only an exemplary embodiment of the present application and is not used to limit the protection scope of the present application.
Claims
1. A copywriting generation method, characterized in that: The method comprises: Get the generated theme; Determine a first level tag according to the generated topic, where the first level tag is one of the multiple level tags included in the database; According to the database, obtaining a first copy material corresponding to the first level label; Generate a first copy using a first artificial intelligence AI model based on the first copy material; Determining whether the generated quality of the first copy meets the quality condition; If the generated quality does not meet the quality condition, the first copy is adjusted.
2. The method according to claim 1, characterized in that The determining whether the generated quality of the first copy meets the quality condition includes: Using the second AI model, performing dependency syntactic analysis on the first sentence, wherein the dependency syntactic analysis is used to analyze dependency relationships between words in a sentence, and the first sentence is a sentence in the first copy; Determining whether the dependency relationship between the words in the first sentence meets the quality condition; If the generated quality does not meet the quality condition, adjusting the first text includes: If the dependency relationship between the words in the first sentence does not meet the quality condition, the second AI model is used to adjust the first sentence in the first copy.
3. The method according to claim 2, characterized in that The determining whether the generated quality of the first copy meets the quality condition includes: Using the second AI model, comparing the similarity between a second sentence and a third sentence, where the second sentence is a sentence in the first copywriting material, and the third sentence is a sentence in the first copywriting material; Determining whether the similarity satisfies the quality condition, wherein the quality condition includes that a value of the similarity satisfies a first value range; If the generated quality does not meet the quality condition, adjusting the first text includes: If the value of the similarity does not satisfy the first value range, the second sentence in the first copy is adjusted using the second AI model.
4. The method according to any one of claims 1 to 3, characterized in that: The determining whether the generated quality of the first copy meets the quality condition includes: Determining the frequency of occurrence of a first character included in the first text in the first text; Determine whether the occurrence frequency meets the quality condition, the quality condition including that the occurrence frequency satisfies a second value range; If the generated quality does not meet the quality condition, adjusting the first text includes: If the occurrence frequency does not satisfy the second value range, the first copy is adjusted using the first AI model.
5. The method according to claim 1, characterized in that The first first-level tag corresponds to N second-level tags, the first copy material includes a plurality of copy paragraph samples corresponding to each of the N second-level tags, the first copy includes M copy paragraphs, the copy paragraph corresponds to one of the N second-level tags, and N and M are positive integers; The step of generating a first copy using a first AI model according to the first copy material includes: The M copy paragraphs are generated using the first AI model, and the copy paragraphs are generated based on multiple copy paragraph samples corresponding to the secondary tags corresponding to the copy paragraphs.
6. The method according to claim 1, characterized in that The method further comprises: Determine a second first-level tag according to the generated topic, where the second first-level tag is one of the plurality of first-level tags included in the database; According to the database, obtaining a second copy material corresponding to the second first-level label; A second copy is generated using a first AI model according to the first copy and the second copy materials, and the second copy is after the first copy.
7. The method according to claim 1, characterized in that The method further comprises: Acquire a generation style, where the generation style is used to indicate a content style of the first copy; The step of generating a first copy using a first AI model according to the first copy material includes: The first copy is generated using the first AI model according to the generation style and the first copy material.
8. The method according to claim 1, characterized in that The method further comprises: Get the target article; Parsing the target article to obtain at least one copywriting material; The database is updated according to the at least one text material.
9. A copywriting generation device, characterized in that: The device comprises: An acquisition module, the acquisition module is used to acquire a generated theme; the acquisition module is also used to acquire a first copy material corresponding to a first first-level tag according to a database, the first first-level tag being one of the multiple first-level tags included in the database; A determination module, the determination module is used to determine the first level label according to the generated theme; the determination module is also used to determine whether the generated quality of the first copy meets the quality condition; A generation module, the generation module is used to generate a first copy according to the first copy material using a first artificial intelligence AI model; An adjustment module is used to adjust the first text if the generated quality does not meet the quality condition.
10. A computer program product comprising instructions, characterized in that When the method is executed on a computing device, the computing device is caused to execute the method according to any one of claims 1 to 8.