Automatic content generation method, computer equipment, medium and system

Abstract

The invention relates to the technical field of computers and provides an automatic content generation method, computer equipment, a medium and a system. The method comprises the following steps: acquiring a hotspot theme through a hotspot screening module; inputting the hotspot topics into a plurality of generative artificial intelligence models in a hotspot analysis module so as to obtain contents associated with the hotspot topics generated by the plurality of generative artificial intelligence models respectively; transmitting contents associated with the hot topics respectively generated by the plurality of generative artificial intelligence models to a style conversion module; and through the style conversion module, based on a system cue word at least comprising a style cue word, generating stylized content for the hot topic by using the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models. In this way, the overall efficiency is improved, and individual requirements are met.

Description

Technical Field

[0001] This application relates to the field of computer technology, and particularly to a method for automatically generating content, a computer device, a medium, and a system. Background Art

[0002] Generative Artificial Intelligence (GAI) refers to generating new content with logic and coherence based on the input data of users. For example, it generates content such as text, images, sounds, and videos based on algorithms and models. Through the adoption of various large model technologies, multimodal technologies, and generation algorithms, etc., Artificial Intelligence Generated Content (AIGC) has been widely applied in various fields. Generally, by inputting a prompt, it generates works such as matching text, images, audio, and video. In the field of hot news information processing, it is often necessary to quickly generate reports and various analysis results of relevant hot news data based on hot topics. Moreover, it is necessary to consider that hot topics often have strong timeliness, that is, the popularity may disappear in a short time, and it is also necessary to consider the complex and changing needs of the audience. Therefore, it is necessary to be able to quickly generate relevant hot news information processing results for market hotspots in a timely manner. It should not only be able to cover the audience and relevant information within the scope of different subject industries, but also be able to provide results with different levels of content richness and different analysis perspectives, so as to meet the needs of different audiences in a targeted manner. All these pose challenges to the application of GAI and AIGC in the field of hot news information processing. In the prior art, one content generation solution is to rely on manual retrieval for content classification management and rely on manual review to overcome the uncontrollability and instability of GAI in content generation. However, this results in high time costs, low execution efficiency, and it is difficult to capture hotspots in a timely manner and provide targeted services for different customer needs. Another content generation solution is to use stylized prompts to match the corresponding style type of GAI from multiple candidate GAIs, and then generate content based on the matched GAI, that is, to overcome the uncontrollability and instability of GAI in content generation through prompts. Examples in this regard can be seen in the Chinese patent application with the application number CN117115303A, "Method, System, Computing Device, and Storage Medium for Content Generation". However, such a content generation solution using stylized prompts and the matched GAI must first find the matched GAI every time content is generated and then can only rely on the matched GAI for content generation. In this way, the utilization rate of model resources is low; moreover, with the rapid change of hot topics and the rapid change of the needs of the audience, it may be necessary to frequently generate stylized prompts, perform model matching, and regenerate content, making it difficult to achieve rapid response and wide coverage, and it is also not conducive to improving the overall efficiency.

[0003] Therefore, this application provides a content automatic generation method, computer device, medium, and system to address the technical problems in the prior art. Summary of the Invention

[0004] In a first aspect, the present application provides a method for automatically generating content. The method for automatically generating content includes: obtaining a hot topic through a hot topic screening module; inputting the hot topic into a plurality of generative artificial intelligence models in a hot topic analysis module to obtain content associated with the hot topic generated by each of the plurality of generative artificial intelligence models, and transmitting the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models to a style conversion module, wherein the optimization strategies of each of the plurality of generative artificial intelligence models have different optimization objectives for model inputs, and the plurality of generative artificial intelligence models are optimized independently of each other; through the style conversion module, based on a system prompt word including at least a style prompt word, using the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models, generating stylized content for the hot topic, wherein the stylized content for the hot topic has a style corresponding to the style prompt word.

[0005] Through the first aspect of the present application, through the separate design of content generation and stylization processing, and by using a plurality of generative artificial intelligence models to respectively generate content associated with the hot topic, this helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be improved by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative artificial intelligence models. By performing stylization processing on the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models that have been generated, it can support the free combination of core scenario content and content style conversion, and support the generation of content in different styles to meet the personalized needs of different audience groups. In this way, not only can it cover the audience groups and relevant information within the scope of different subject industries, but also it can provide results with different levels of content richness and different analysis perspectives, which is beneficial to meeting the needs of different audience groups in a targeted manner, and has the advantages of low time cost, high execution efficiency, being able to capture hot topics in a timely manner, and providing targeted services for different customer needs. Moreover, through the parallel operation of multiple models, the utilization rate of model resources is improved, and with the rapid changes in hot topics and the rapid changes in the needs of the audience group, through content recombination, rapid response and wide coverage can be achieved, which is beneficial to improving the overall efficiency.

[0006] In a possible implementation of the first aspect of the present application, the hot topic is used to indicate the content characteristics of the stylized content for the hot topic, the style prompt word is used to indicate the style characteristics of the stylized content for the hot topic, the system prompt word further includes a plurality of task prompt words, the plurality of task prompt words correspond one-to-one with the plurality of generative artificial intelligence models, and each task prompt word in the plurality of task prompt words is used for the model service call of the generative artificial intelligence model corresponding to the task prompt word in the plurality of generative artificial intelligence models.

[0007] In a possible implementation of the first aspect of the present application, the system prompt word is provided by the prompt word arrangement module to the hot topic analysis module and the style conversion module, and the function of the prompt word arrangement module for generating the plurality of task prompt words is based on the content prompt word, the content preference corresponding to the content prompt word, and the content associated with the content prompt word generated by each of the plurality of generative artificial intelligence models.

[0008] In a possible implementation of the first aspect of the present application, the plurality of task prompt words include the settings of the plurality of generative artificial intelligence models for artificial intelligence agents, domain common sense, special variables, processing tasks, and response formats respectively.

[0009] In a possible implementation of the first aspect of the present application, through the style conversion module, based on the system prompt word including at least the style prompt word, using the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models, the stylized content for the hot topic is generated, including: integrating the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models according to the standard style, so as to obtain the standardized content for the hot topic with the standard style; based on the style prompt word, selectively performing content deletion operations and selectively performing stylized element addition operations on the standardized content for the hot topic, so as to obtain the stylized content for the hot topic with the style corresponding to the style prompt word.

[0010] In a possible implementation of the first aspect of the present application, the standard style is independent of the style corresponding to the style prompt word, and the standardized content for the hot topic with the standard style is independent of the style prompt word.

[0011] In a possible implementation manner of the first aspect of the present application, the stylized element adding operation includes adding at least one stylized element of the style corresponding to the style prompt word, and the at least one stylized element is selected from a plurality of stylized elements, and the plurality of stylized elements include symbols, expressions, and pictures.

[0012] In a possible implementation manner of the first aspect of the present application, the plurality of generative artificial intelligence models provide the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models through multi-model parallel operation, and the updates of each of the plurality of generative artificial intelligence models are performed independently of each other.

[0013] In a possible implementation manner of the first aspect of the present application, the system prompt word further includes the weight ratio of each of the plurality of generative artificial intelligence models, and the weight ratio of each of the plurality of generative artificial intelligence models is used to indicate the importance degree of the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models relative to the stylized content for the hot topic. And when there is a generative artificial intelligence model being updated among the plurality of generative artificial intelligence models and the weight ratio of the generative artificial intelligence model being updated is not higher than a preset threshold, the old model of the generative artificial intelligence model being updated is used to ensure that the multi-model parallel operation of the plurality of generative artificial intelligence models is not interrupted.

[0014] In a possible implementation manner of the first aspect of the present application, through the style conversion module, based on the system prompt word that at least includes the style prompt word, and using the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models, the stylized content for the hot topic is generated, including: integrating the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models respectively according to a plurality of preset standard styles, so as to obtain a plurality of standardized contents for the hot topic, and each of the plurality of standardized contents for the hot topic has the plurality of preset standard styles; based on the style prompt word, selecting the preset standard style that matches the style prompt word from the plurality of preset standard styles, and determining the standardized content for the hot topic that has the selected preset standard style among the plurality of standardized contents for the hot topic as the reference standardized content; based on the style prompt word, selectively performing a content deletion operation and a stylized element adding operation on the reference standardized content, so as to obtain the stylized content for the hot topic that has the style corresponding to the style prompt word.

[0015] In a possible implementation of the first aspect of the present application, the multiple preset standard styles are key styles determined based on user demand prediction.

[0016] In a possible implementation of the first aspect of the present application, the hot topic is a key hot topic determined based on the user demand prediction.

[0017] In a possible implementation of the first aspect of the present application, the style prompt words are selected from multiple preset style prompt words, and the multiple preset style prompt words include standard style, consulting style, advisory style, relaxed style, and concise style.

[0018] In a possible implementation of the first aspect of the present application, the multiple generative artificial intelligence models include a first generative artificial intelligence model, a second generative artificial intelligence model, a third generative artificial intelligence model, a fourth generative artificial intelligence model, and a fifth generative artificial intelligence model. Among them, the optimization strategy of the first generative artificial intelligence model has an optimization goal for the relevance and timeliness of the model input, the optimization strategy of the second generative artificial intelligence model has an optimization goal for the information extraction and in-depth research of the model input, the third generative artificial intelligence model has an optimization goal for the trend prediction of the model input, the optimization strategy of the fourth generative artificial intelligence model has an optimization goal for the multi-industry analysis results of the model input, and the optimization strategy of the fifth generative artificial intelligence model has an optimization goal for the event context analysis results of the model input.

[0019] In a possible implementation of the first aspect of the present application, the stylized content for the hot topic is a hot topic analysis report for the hot topic. The content associated with the hot topic generated by the first generative artificial intelligence model is the content of the hot topic-related hot news scenario. The content associated with the hot topic generated by the second generative artificial intelligence model is the content of the research report analysis scenario related to the hot topic. The content associated with the hot topic generated by the third generative artificial intelligence model is the content of the external market analysis scenario related to the hot topic. The content associated with the hot topic generated by the fourth generative artificial intelligence model is the content of the multi-industry analysis scenario related to the hot topic. The content associated with the hot topic generated by the fifth generative artificial intelligence model is the content of the event development context scenario related to the hot topic.

[0020] In a possible implementation of the first aspect of the present application, the style prompt words are used to set the combination between the scenario contents generated by each of the multiple generative artificial intelligence models.

[0021] In a possible implementation of the first aspect of the present application, the content of the hot news scenario associated with the hot topic is based on the vectorized hot news information associated with the hot topic. The content of the event development context scenario associated with the hot topic is the output result of the event development context sorted by the large language model according to the time axis associated with the hot topic. The content of the external market analysis scenario associated with the hot topic and the content of the multi-industry analysis scenario associated with the hot topic are obtained through hot topic analysis by an artificial intelligence agent and an analysis framework.

[0022] In a possible implementation of the first aspect of the present application, the content automatic generation method further includes: through the style conversion module, based on another style prompt word different from the style prompt word, using the content associated with the hot topic generated by each of the multiple generative artificial intelligence models, generating another stylized content for the hot topic, where the another stylized content for the hot topic has a style corresponding to the another style prompt word.

[0023] In a possible implementation of the first aspect of the present application, the hot topic is user-defined, or the hot topic is obtained by processing the public domain hot news data through the hot topic screening module based on the hot topic capture algorithm, the hot topic trend analysis algorithm, and the hot word analysis algorithm.

[0024] In a second aspect, an embodiment of the present application further provides a computer device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the method according to any one of the implementations of any of the above aspects.

[0025] In a third aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores computer instructions. When the computer instructions run on a computer device, the computer device is caused to execute the method according to any one of the implementations of any of the above aspects.

[0026] In a fourth aspect, an embodiment of the present application further provides a computer program product, which includes instructions stored on a computer-readable storage medium. When the instructions run on a computer device, the computer device is caused to execute the method according to any one of the implementations of any of the above aspects.

[0027] Fifth aspect, the present application provides a content automatic generation system. The content automatic generation system includes: a hot topic screening module, configured to obtain hot topics; a hot topic analysis module, the hot topic analysis module includes a plurality of generative artificial intelligence models, and the hot topic analysis module is configured to: input the hot topics into the plurality of generative artificial intelligence models, so as to obtain the content associated with the hot topics generated by each of the plurality of generative artificial intelligence models, and transmit the content associated with the hot topics generated by each of the plurality of generative artificial intelligence models to a style conversion module, wherein the optimization strategies of each of the plurality of generative artificial intelligence models have different optimization objectives for model inputs, and the plurality of generative artificial intelligence models are optimized independently of each other; the style conversion module, configured to: based on a system prompt word that at least includes a style prompt word, use the content associated with the hot topics generated by each of the plurality of generative artificial intelligence models to generate stylized content for the hot topics, wherein the stylized content for the hot topics has a style corresponding to the style prompt word.

[0028] Through the fifth aspect of the present application, through the separate design of content generation and stylization processing, and by using a plurality of generative artificial intelligence models to separately generate the content associated with the hot topics, this helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be improved by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative artificial intelligence models. By performing stylization processing on the content associated with the hot topics generated by each of the plurality of generative artificial intelligence models that have been generated, it is possible to support the free combination of core scenario content and content style conversion, and support the generation of content with different styles to meet the personalized needs of different audience groups. In this way, not only can it cover the audience groups and relevant information within the scope of different subject industries, but also it can provide results with different content richness levels and different analysis perspectives, which is beneficial to meeting the needs of different audience groups in a targeted manner. It has the advantages of low time cost, high execution efficiency, being able to capture hot topics in a timely manner, and providing targeted services for different customer needs. Moreover, through the parallel operation of multiple models, the utilization rate of model resources is improved, and as the hot topics and the needs of the audience groups change rapidly, through content recombination, rapid response and wide coverage can be achieved, which is beneficial to improving the overall efficiency.

[0029] In a possible implementation of the fifth aspect of the present application, the hot topic is used to indicate the content features of the stylized content for the hot topic, the style prompt word is used to indicate the style features of the stylized content for the hot topic, the system prompt word further includes a plurality of task prompt words, and the plurality of task prompt words correspond to the plurality of generative artificial intelligence models one by one. Each task prompt word in the plurality of task prompt words is used for model service invocation of the generative artificial intelligence model corresponding to this task prompt word in the plurality of generative artificial intelligence models.

[0030] In a possible implementation of the fifth aspect of the present application, the style conversion module is configured to: based on the system prompt word that at least includes the style prompt word, use the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models to generate the stylized content for the hot topic, including: integrating the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models according to the standard style to obtain the standardized content for the hot topic with the standard style; based on the style prompt word, selectively perform content deletion operations and selectively perform stylized element addition operations on the standardized content for the hot topic, so as to obtain the stylized content for the hot topic with the style corresponding to the style prompt word.

[0031] In a possible implementation of the fifth aspect of the present application, the plurality of generative artificial intelligence models include a first generative artificial intelligence model, a second generative artificial intelligence model, a third generative artificial intelligence model, a fourth generative artificial intelligence model, and a fifth generative artificial intelligence model. Among them, the optimization strategy of the first generative artificial intelligence model has an optimization goal for the relevance and timeliness of the model input, the optimization strategy of the second generative artificial intelligence model has an optimization goal for information extraction and in-depth research of the model input, the third generative artificial intelligence model has an optimization goal for trend prediction of the model input, the optimization strategy of the fourth generative artificial intelligence model has an optimization goal for the parsing results of multiple industries of the model input, and the optimization strategy of the fifth generative artificial intelligence model has an optimization goal for the parsing results of the event context of the model input. Brief Description of the Drawings

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0033] Figure 1 A flowchart of a content automatic generation method provided by an embodiment of this application; Figure 2 The first one provided by an embodiment of this application is based on Figure 1 A schematic diagram of the service processing flow of the content automatic generation method shown; Figure 3 The second one provided by an embodiment of this application is based on Figure 1 A schematic diagram of the service processing flow of the content automatic generation method shown; Figure 4 A schematic diagram of a content automatic generation system provided by an embodiment of this application; Figure 5 A schematic diagram of the structure of a computing device provided by an embodiment of this application. Detailed implementation manners

[0034] Hereinafter, embodiments of this application will be further described in detail with reference to the accompanying drawings.

[0035] It should be understood that in the description of this application, "at least one" means one or more, and "a plurality" means two or more. In addition, words such as "first" and "second" are only used for the purpose of distinguishing descriptions unless otherwise specified, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying an order.

[0036] Figure 1 A flowchart of a content automatic generation method provided by an embodiment of this application. As Figure 1 shown, the content automatic generation method includes the following steps.

[0037] Step S101: Obtain a hot topic through a hot topic screening module.

[0038] Step S103: Input the hot topic into multiple generative artificial intelligence models in a hot topic analysis module, so as to obtain the content associated with the hot topic generated by each of the multiple generative artificial intelligence models, and transmit the content associated with the hot topic generated by each of the multiple generative artificial intelligence models to a style conversion module, wherein the optimization strategies of each of the multiple generative artificial intelligence models have different optimization objectives for model inputs, and the multiple generative artificial intelligence models are optimized independently of each other.

[0039] Step S105: Through the style conversion module, based on a system prompt word that at least includes a style prompt word, use the content associated with the hot topic generated by each of the multiple generative artificial intelligence models to generate a stylized content for the hot topic, wherein the stylized content for the hot topic has a style corresponding to the style prompt word.

[0040] Figure 1 The content generation method shown is applied to the field of hot news information processing. For example, it can quickly generate reports and various analysis results of relevant hot news data based on hot topics. It utilizes Generative Artificial Intelligence (GAI) technology and Artificial Intelligence Generated Content (AIGC). It can not only cover the audience groups and relevant information within different thematic industries, but also provide results with different levels of content richness and different analysis perspectives, which is conducive to meeting the needs of different audience groups in a targeted manner. It has the advantages of low time cost, high execution efficiency, being able to capture hot topics in a timely manner, and providing targeted services for different customer needs. Moreover, different from the existing content generation solutions, through the parallel operation of multiple models, the utilization rate of model resources is improved, and with the rapid changes in hot topics and the rapid changes in the needs of the audience group, through content recombination, it can achieve rapid response and wide coverage, which is conducive to improving the overall efficiency.

[0041] Refer to Figure 1, in step S101, the hot topic screening module is used to obtain hot topics. Here, the hot topic screening module is used to provide hot topics, which can be user-defined hot topics, such as specific international events, etc. The hot topics can also be obtained through analysis and sorting by the hot topic screening module, such as based on market hotspots and independent hot topic topics, such as trade topics, or hot events. Then, in step S103, the hot topics are input into multiple generative artificial intelligence models in the hot topic analysis module, so as to obtain the content associated with the hot topics generated by each of the multiple generative artificial intelligence models, and the content associated with the hot topics generated by each of the multiple generative artificial intelligence models is transmitted to the style conversion module. Among them, the optimization strategies of each of the multiple generative artificial intelligence models have different optimization objectives for the model input, and the multiple generative artificial intelligence models are optimized independently of each other. In this way, it is divided into the first stage of obtaining hot topics and the second stage of content generation. In the second stage of content generation, based on the hot topics, the content associated with the hot topics generated by each of the multiple generative artificial intelligence models is obtained. It should be understood that the optimization strategies of each of the multiple generative artificial intelligence models have different optimization objectives for the model input, and the multiple generative artificial intelligence models are optimized independently of each other. This means that for the same model input, such as the same prompt word representing a hot topic, different generative artificial intelligence models will generate content with different focuses based on the optimization strategies with different optimization objectives. By setting different optimization objectives and optimizing the multiple generative artificial intelligence models independently of each other, it can ensure that the content generated by the multiple generative artificial intelligence models has a better overall content richness and can provide different analysis perspectives. Considering that when a single generative artificial intelligence model generates content based on a model input (such as a prompt word), the generated content has a certain degree of uncontrollability and instability. For example, for two consecutive content generation requests for the same prompt word by the same generative artificial intelligence model, the generated content may be somewhat different. Therefore, the hot topics are input into multiple generative artificial intelligence models in the hot topic analysis module, so as to obtain the content associated with the hot topics generated by each of the multiple generative artificial intelligence models. In this way, using the different optimization objectives of the optimization strategies of each of the multiple generative artificial intelligence models helps to control the uncontrollability and instability of the content generated by the multiple generative artificial intelligence models as a whole.In other words, compared to the content generated by a single generative AI model for a given prompt (such as a certain hot event), the content generated by multiple generative AI models for the same given prompt generally performs more stably and controllably overall. Moreover, by setting different optimization objectives for multiple generative AI models as the optimization directions of their respective optimization strategies, it is also possible to increase the overall coverage of the content, which helps to avoid missing important content features or having insufficient output of key content features.

[0042] Continue to refer to Figure 1, in step S105, through the style conversion module, based on the system prompt words that at least include style prompt words, using the content associated with the hot topic generated by each of the multiple generative AI models, generate stylized content for the hot topic. Among them, the stylized content for the hot topic has a style corresponding to the style prompt words. As described above, it is divided into a first stage of obtaining the hot topic and a second stage of content generation. Among them, in the second stage of content generation, using the different optimization objectives of the respective optimization strategies of the multiple generative AI models, and by inputting the hot topic into the multiple generative AI models in the hot topic analysis module, the content associated with the hot topic generated by each of the multiple generative AI models is obtained. In this way, the content generated by the multiple generative AI models has better controllability and stability as a whole, and also increases the overall coverage of the content, which helps to avoid missing important content features or insufficient output of key content features. Therefore, on the basis of the content generation in step S103, a third stage of stylization processing is performed in step S105, that is, after the first stage of obtaining the hot topic and the second stage of content generation, a third stage of stylization processing is performed. Specifically, in the third stage of stylization processing, that is, in step S105, based on the system prompt words that at least include style prompt words, using the content associated with the hot topic generated by each of the multiple generative AI models, generate stylized content for the hot topic. In this way, a separate design of content generation and stylization processing is achieved; in the second stage of content generation, that is, in step S103, it does not involve style prompt words and the style corresponding to the style prompt words, but each of the multiple generative AI models is used to generate the content associated with the hot topic. This helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be improved by increasing the parallelism of model operation, and the overall coverage can also be expanded by increasing the differences between different generative AI models. For example, the optimization objective of one generative AI model can be set to focus more on timeliness and relevance (that is, more focus on current information and events), while the optimization objective of another generative AI model can be set to focus more on the collation of historical event contexts (that is, more focus on retrospective information and historical events), and another generative AI model can be set to focus more on trend prediction. On the basis of trying to increase the richness of the generated content and expand its coverage, in the third stage of stylization processing, that is, in step S105, it does not involve content generation, that is, it does not call the large model to regenerate content, but uses the content associated with the hot topic generated by each of the multiple generative AI models that have been generated, and through stylization processing, the stylized content for the hot topic is transformed and obtained.Thus, based on making full use of the model resources of multiple generative AI models in the hot topic analysis module, by stylizing the content associated with the hot topics generated by each of the multiple generative AI models that have been generated, it is possible to support the free combination of core scenario content and content style conversion, support the generation of content in different styles to meet the personalized needs of different audience groups. It should be understood that when the hot topics change rapidly, parallel processing of multiple generative AI models can be used to generate content for the new hot topics in a timely manner, achieving rapid response; when the needs of the audience groups change rapidly, new style prompt words can be used to match the new personalized needs, thus ensuring wide coverage. By using the separate design of the above-mentioned content generation and stylization processing, it is possible to selectively optimize the hot topic analysis module and selectively optimize the style conversion module. For example, a new hot topic does not necessarily lead to new personalized needs, so it is possible to use the hot topic analysis module to generate new content according to the new hot topic while keeping the style conversion module. Another example is that in the face of new personalized needs that may arise for the original hot topics, corresponding stylization processing can be carried out by introducing new style prompt words on the basis of the content generated by the hot topic analysis module, so as to obtain stylized content with a new style. In this way, it is beneficial to improve the overall efficiency. Moreover, in terms of the update and upgrade of large models, the hot topic analysis module responsible for content generation and the style conversion module responsible for stylization processing can also be upgraded separately, which helps to improve the upgrade efficiency.

[0043] In summary, Figure 1The content generation method shown can, through the separate design of content generation and stylization processing, generate content associated with the hot topic through multiple generative artificial intelligence models respectively. This helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be increased by increasing the parallelism of model operation. The overall coverage can also be expanded by increasing the differences between different generative artificial intelligence models. By performing stylization processing on the content associated with the hot topic generated by each of the multiple generative artificial intelligence models that have been generated, it is possible to support the free combination of core scenario content and content style conversion, support the generation of content in different styles to meet the personalized needs of different audience groups. In this way, not only can the audience groups and relevant information within the scope of different theme industries be covered, but also results with different levels of content richness and different analysis perspectives can be provided, which is beneficial to meeting the needs of different audience groups in a targeted manner. It has the advantages of low time cost, high execution efficiency, being able to promptly capture hot topics, and providing targeted services for different customer needs. Moreover, through the parallel operation of multiple models, the utilization rate of model resources is improved. And with the rapid changes in hot topics and the rapid changes in the needs of the audience group, through content recombination, rapid response and wide coverage can be achieved, which is beneficial to improving the overall efficiency.

[0044] Figure 2 The first provided by the embodiment of the present application is based on Figure 1 Schematic diagram of the business processing flow of the content automatic generation method shown. Figure 2 The business processing flow shown refers to Figure 1The content generation method shown combines public domain hotspots, internal information, research reports and other data, and uses prompt arrangement and large model technology to generate hotspot analysis results. Specifically, starting from obtaining hotspots 201, the subsequent steps are hotspot vectorization 202, knowledge base retrieval and sorting 203, extracting related information 204, adding citation traceability information 205, obtaining prompts 206, and large model analysis 207; in this way, a one-way business processing flow is formed by multiple nodes. And during the process of this one-way business processing flow, there are also other nodes for providing support functions. For example, hotspot synchronization 210 is used to support obtaining hotspots 201. Another example is that knowledge base retrieval and sorting 203 can call the vector knowledge base 220, and the maintenance and update of the vector knowledge base 220 can be achieved through the information database 230, semantic segmentation service 231 and message queue 232. Another example is that after obtaining prompts 206, the prompt service 241 can be used to better support large model analysis 207, and the prompt arrangement 240 can be used to support the prompt service 241. And a hotspot list module can be provided to further support obtaining hotspots 201. For example, aggregating hotspots in a specific field is provided, and the switching of hourly, daily, and weekly rankings can also be supported. And each hotspot can support viewing details (such as hotspot trends, related information, hot word analysis), etc. A custom hotspot module can also be provided, supporting users to customize the input of hotspot titles and generate analysis reports. For large model analysis 207, four content structures such as title, self-developed large model research report analysis, market analysis, and industry analysis can be provided. Each structure content supports functions such as re-obtaining, copying, and viewing historical versions, and each hotspot also supports sorting out the context of event development. And for the finally output analysis results, a style conversion module can be provided for content processing after generating the analysis report, to support generating content in different styles to meet the personalized needs of different audience groups and facilitate meeting the needs of different users. In this way, the final output of the analysis results is achieved, and based on market hotspots and custom hotspots, hotspot analysis reports are quickly generated for different audience groups, reducing the investment in labor costs and the error rate of content, improving the richness of content, and meeting customer demands in a targeted manner.

[0045] Figure 3 The second one provided by the embodiment of the present application is based on Figure 1 The schematic diagram of the business processing flow of the content generation method shown. Figure 3 The business processing flow shown refers to Figure 1The content generation method shown starts from the login content platform 301, goes through hot spot screening 303 and AI model content creation 305, and finally is for business scenario use 307. Among them, in hot spot screening 303, initial hot spot screening 310 is first carried out, and then hot spot trend viewing 311, hot word analysis viewing 312, relevant information viewing 313 can be carried out synchronously, and finally the determination of interested hot spots 314 is carried out. In AI model content creation 305, the AI model 320 is first called, and then the following can be carried out synchronously: making the AI model write analysis content through hot spots and market data 321, multi-module content combination 322, AI model conversion of content style 323, and finally publishing 324. In this way, based on the market hot spot ability of the content platform, the hot spot data is vectorially retrieved through the existing information in the library, and then the information is sliced and correlated for analysis and the prompt words are trained through the existing database, and finally the analysis results are output. In hot spot screening 303, optimization can be carried out in obtaining hot spots. For example, hot spot insight first accurately captures market hot spots and related data through external algorithm capabilities, including heat trends and hot word analysis, and at the same time supports users to customize hot spot themes for creation to ensure the comprehensiveness and personalization of hot spot information. In addition, in initializing the information knowledge base, content segmentation and vectorization technologies can be used to transform and land the information in certain fields of the content system to the knowledge base system. In arranging system prompt words, for specific tasks, different prompt words can be arranged to complete the setting of information such as AI agent roles, domain common sense, special variables, processing tasks, response formats, etc. In vectorially matching hot spot-related information, after obtaining hot spots, vectorization technology is used to efficiently match the relevant information in the content library and perform intelligent sorting according to the relevance and time freshness of the information to ensure that users can quickly obtain the most relevant and latest information. In generating hot spot information by the AI model, key information in the relevant information is extracted through content slicing technology, and the prompt words set in the previous steps are obtained, and the large model service interface parameters are assembled together to call the large model service to complete the content creation of hot spots. And, in setting the AI model prompt words, in some embodiments, the prompt word creation covers five core scenario capabilities: generating hot spot information titles, self-developed large model research report analysis, external model market analysis, industry analysis, and event context analysis. The prompt word for generating hot spot information titles generates hot spot popular titles by using vectorization technology matching and AI large model technology. The self-developed large model prompt words extract information from hot spot-related research reports, deeply dig into the essence of the research reports, and refine key information and viewpoints. The external market analysis and industry analysis prompt words rely on the computing power of the large model itself and the information slice information of the vector database to accurately predict market trends. The event context analysis prompt words clearly sort out the development context of events through the extraction of relevant information and large model technology to help users quickly understand the whole picture of events.Based on the above five core scenario capabilities, not only can the content of each scenario be regenerated, but also the content of multiple scenarios can be freely combined. Through the setting of prompt words, the hot topic insight function also supports content style conversion. Whether it is a formal, humorous or concise style, it can be easily achieved to ensure that the content meets the needs of various users. In this way, it helps to achieve coverage of large model research report analysis, external model market analysis, independent industry analysis, event context analysis, etc. Moreover, the finally generated results support style conversion, making the results more accessible to different audiences.

[0046] See Figure 1 , Figure 2 and Figure 3 , in a possible implementation manner, the hot topic is used to indicate the content characteristics of the stylized content for the hot topic, the style prompt word is used to indicate the style characteristics of the stylized content for the hot topic, the system prompt word further includes a plurality of task prompt words, the plurality of task prompt words correspond to the plurality of generative artificial intelligence models one by one, and each task prompt word in the plurality of task prompt words is used for the model service call of the generative artificial intelligence model corresponding to this task prompt word in the plurality of generative artificial intelligence models. As described above, through the separate design of content generation and stylization processing, the content associated with the hot topic is generated by the plurality of generative artificial intelligence models respectively. This helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be increased by increasing the model operation parallelism, and the overall coverage can be expanded by increasing the differences between different generative artificial intelligence models. By stylizing the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models that have been generated, it is possible to support the free combination of core scenario content and content style conversion, and support the generation of content in different styles to meet the personalized needs of different audiences. Here, the hot topic is used to indicate the content characteristics of the stylized content for the hot topic, so it is applicable to the second stage of content generation; the style prompt word is used to indicate the style characteristics of the stylized content for the hot topic, so it is applicable to the third stage of stylization processing. Combined with a plurality of task prompt words, the model service call of the generative artificial intelligence model corresponding to this task prompt word in the plurality of generative artificial intelligence models can be executed more effectively. In this way, it helps to improve the model resource utilization efficiency.

[0047] In some embodiments, the system prompt words are provided by the prompt word arrangement module to the hotspot analysis module and the style conversion module. The function of the prompt word arrangement module for generating the multiple task prompt words is based on the content prompt words, the content preferences corresponding to the content prompt words, and the content associated with the content prompt words generated by each of the multiple generative AI models. Here, the prompt word arrangement module generates task prompt words to arrange the tasks of each generative AI model. For example, assume a certain generative AI model, such as a self-developed large model, extracts information, refines viewpoints, and generates research reports according to the corresponding task prompt words. Therefore, the task prompt words are related to content generation. It can be seen that the hotspot analysis module uses multiple generative AI models for content generation, and then the style conversion module performs specific style adjustments and style conversions. That is, a standard press release is first generated, and then content deletion and stylization processing are carried out. In some examples, a method of training the prompt word arrangement module is to use the content prompt words as input training data, compare the content preferences corresponding to the content prompt words and the content generated by the generative AI models, which can improve the quality of the content generated by the generative AI models. For example, the hotspot news model should generate content that better conforms to the content preferences of the hotspot news title. Further, debugging in combination with specific business scenarios can improve the generation quality of the task prompt words. Specifically, according to the hotspots selected by the user or custom hotspots, this is used as the "input hotspot", and combined with functions such as the information library, vector library, system prompt words, and basic large model, so as to rewrite the "input hotspot" to make the expression of the hotspot information more complete, facilitating the generation of subsequent standard version style content. Among them, the system prompt word arrangement and debugging link is as described above. In this way, by optimizing the design of the function of the prompt word arrangement module for generating the multiple task prompt words, designing the training loss function and model convergence strategy based on the content prompt words, the content preferences corresponding to the content prompt words, and the content associated with the content prompt words generated by each of the multiple generative AI models, the performance in content generation can be enhanced, so as to better utilize the separate design of the above content generation and stylization processing, and use the task prompt words to improve the effect of model service calls, which helps to improve the efficiency of model resource utilization.

[0048] In some embodiments, the multiple task prompt words include the settings of each of the multiple generative AI models regarding AI agents, domain common sense, special variables, processing tasks, and response formats. In this way, it helps to further improve the effect of model service calls and the efficiency of model resource utilization through diverse settings.

[0049] In a possible implementation, through the style conversion module, based on the system prompt that at least includes the style prompt word, using the content associated with the hot topic generated by each of the multiple generative artificial intelligence models, generate the stylized content for the hot topic, including: integrating the content associated with the hot topic generated by each of the multiple generative artificial intelligence models according to the standard style, so as to obtain the standardized content for the hot topic with the standard style; based on the style prompt word, selectively perform content deletion operations and selectively perform stylized element addition operations on the standardized content for the hot topic, so as to obtain the stylized content for the hot topic with the style corresponding to the style prompt word. As described above, using the different optimization objectives of the optimization strategies of each of the multiple generative artificial intelligence models helps to control the uncontrollability and instability of the content generated by the multiple generative artificial intelligence models as a whole. In other words, compared with the content generated by a single generative artificial intelligence model for a given prompt word (such as a certain hot event), the content generated by multiple generative artificial intelligence models for the given prompt word generally performs more stably and controllably as a whole, and by setting different optimization objectives for the multiple generative artificial intelligence models as the optimization directions of their respective optimization strategies, it is also possible to increase the overall coverage of the content, which helps to avoid missing important content features or insufficient output of key content features. Further, in the second stage of content generation, using the different optimization objectives of the optimization strategies of each of the multiple generative artificial intelligence models, and by inputting the hot topic into the multiple generative artificial intelligence models in the hot topic analysis module, so as to obtain the content associated with the hot topic generated by each of the multiple generative artificial intelligence models. In this way, the content generated by the multiple generative artificial intelligence models has good controllability and stability as a whole, and also increases the overall coverage of the content, which helps to avoid missing important content features or insufficient output of key content features. Here, integrate the content associated with the hot topic generated by each of the multiple generative artificial intelligence models according to the standard style, so as to obtain the standardized content for the hot topic with the standard style. This means that the content generated by each of the multiple generative artificial intelligence models can be integrated by referring to the form of a standard press release, which helps to provide the standardized content for the hot topic, so as to achieve timely and efficient provision of targeted standardized content with a standard style for different hot topics, which helps with subsequent stylization processing.Moreover, since the content associated with the hot topics generated by the multiple generative AI models is integrated in a standard style to obtain the standardized content for the hot topics with the standard style, the separate design of content generation and stylization processing is realized. In the second stage of content generation, it does not involve style prompting words and the corresponding style, but the content associated with the hot topics is generated by multiple generative AI models respectively. This helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be improved by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative AI models. Further, on the basis of obtaining the standardized content for the hot topics with the standard style, based on the style prompting words, content deletion operations and stylistic element addition operations are selectively performed on the standardized content for the hot topics, so as to obtain the stylized content for the hot topics with the style corresponding to the style prompting words. In this way, in the third stage of stylization processing, it does not involve content generation, that is, the large model is not called to regenerate content, but the content associated with the hot topics generated by the multiple generative AI models that have been generated is used for stylization processing to transform into the stylized content for the hot topics. In this way, on the basis of making full use of the model resources of the multiple generative AI models in the hot topic analysis module, by stylizing the content associated with the hot topics generated by the multiple generative AI models that have been generated, it is possible to support the free combination of core scenario content and content style conversion, and support the generation of content in different styles to meet the personalized needs of different audience groups. Moreover, since content deletion operations and stylistic element addition operations are selectively performed, the overall complexity is simplified as much as possible, that is, only partial content deletion is required to meet personalized needs, and the maximum coverage obtained in the content generation stage is also fully utilized. Therefore, using content deletion operations to transform the standardized content for the hot topics with the standard style into the stylized content for the hot topics with the style corresponding to the style prompting words not only improves the content generation speed by using multi-model parallel operation and the expanded overall coverage, but also improves the stylization processing speed by deleting content from the expanded overall coverage, which is beneficial to improving the overall efficiency. And using stylistic element addition operations to increase stylistic elements is beneficial to responding to the personalized needs of different audience groups.In some examples, depending on the display platform of the finally generated content, such as chat software on mobile phones, online video platforms, and any Internet media platforms, etc., various restrictions may be imposed on the overall data scale of the generated content, such as the total number of words, the total number of lines, the total data length, etc. These can be understood as part of personalized requirements. Thus, in the stage of stylization processing, through content deletion operations, that is, selectively performing content deletion operations, the content specifications on different platforms can be met.

[0050] The following uses a specific example to further illustrate how to integrate the content related to the hot topic generated by each of the multiple generative AI models according to the standard style, so as to obtain the standardized content for the hot topic with the standard style, and how to selectively perform content deletion operations and selectively add stylistic elements to the standardized content for the hot topic based on the style prompt words, so as to obtain the stylized content for the hot topic with the style corresponding to the style prompt words. As mentioned above, through the hot topic screening module, hot topics are obtained. Here, the hot topic screening module can provide a hot topic list, in which multiple hot topics can be listed according to hours, days, weeks, months or any other sorting method, and the user can also input custom hot topics. Here, hot topic insight content can be generated by real-time capturing public domain hot topics, combined with big data and artificial intelligence technologies. For example, the hot topic list can list hot topics such as "the change in the container transportation volume between two countries after a certain event", and also hot topics such as the start of construction of a cross-border railway and the opening of an international road transportation route. Here, taking "the change in the container transportation volume between two countries after a certain event" as an exemplary hot topic, the hot topic is input into multiple generative AI models in the hot topic analysis module, so as to obtain the content related to the hot topic generated by each of the multiple generative AI models. Here, the respective optimization strategies of the multiple generative AI models have different optimization goals for the model input, and the multiple generative AI models are optimized independently of each other. Exemplarily, a generative AI model has optimization goals for the relevance and timeliness of the model input. Therefore, for the hot topic of "the change in the container transportation volume between two countries after a certain event", it can provide content such as export demand and capacity adjustment. For example, the capacity freight rate increased month-on-month; another generative AI model has an optimization goal for trend prediction of the model input. Therefore, for the hot topic of "the change in the container transportation volume between two countries after a certain event", it can provide content such as future prospects. For example, it is expected that the pre-order volume of container transportation will remain high. The content related to the hot topic generated by each of the multiple generative AI models obtained in this way, starting from different optimization goals, helps to increase the richness of the generated content and expand its coverage. Then, integrate the content related to the hot topic generated by each of the multiple generative AI models according to the standard style, and obtain the standardized content for the hot topic, that is, "the change in the container transportation volume between two countries after a certain event". Then, based on the style prompt words, selectively perform content deletion operations and selectively add stylistic elements to the standardized content for the hot topic.For example, the standard style can be adopted, that is, the original standardized content is retained; for another example, the information style can be adopted to highlight timeliness; for another example, the consultant style can be adopted to provide in-depth analysis and thinking; for another example, the relaxed style can be adopted to add more fashionable stylistic elements to increase the reading pleasure of readers; for another example, the concise style can be adopted to streamline the content and compress the data specifications to suit the readers of mobile terminals. Therefore, based on the standardized content, content deletion operations and stylistic element addition operations are selectively performed on the standardized content for the hot topic, which can support the free combination of core scenario content and content style conversion, and support the generation of content in different styles to meet the personalized needs of different audience groups. Moreover, when the hot topic changes rapidly, for example, from the current hot topic, that is, "the change in the container transportation volume between two countries after a certain event", due to a sudden international event, it becomes "the change in the container transportation volume between two countries after another unexpected event after a certain event", such a new hot topic, content can be generated in a timely manner for the new hot topic through the parallel processing of multiple generative artificial intelligence models, achieving rapid response; when the needs of the audience group change rapidly, new personalized needs can be matched through new style prompt words, ensuring wide coverage.

[0051] In some embodiments, the standard style is independent of the style corresponding to the style prompt word, and the standardized content for the hot topic with the standard style is independent of the style prompt word. In this way, a separate design of content generation and stylistic processing is achieved, making full use of the largest possible coverage achieved in the content generation stage. The standardized content for the hot topic with the standard style is transformed through content deletion operations to obtain the stylized content for the hot topic with the style corresponding to the style prompt word. This not only improves the content generation speed by using multi-model parallel operation and the enlarged overall coverage, but also improves the stylistic processing speed by deleting the content in the enlarged overall coverage, which is beneficial to improving the overall efficiency. Moreover, adding stylistic elements through the stylistic element addition operation is beneficial to responding to the personalized needs of different audience groups.

[0052] In some embodiments, the stylistic element addition operation includes adding at least one stylistic element of the style corresponding to the style prompt word, and the at least one stylistic element is selected from a variety of stylistic elements, and the variety of stylistic elements includes symbols, expressions, and pictures. In this way, by adding stylistic elements, it is beneficial to responding to the personalized needs of different audience groups.

[0053] In a possible implementation, the multiple generative AI models provide the content associated with the hot topics generated by each of the multiple generative AI models through parallel operation of multiple models, and the updates of the multiple generative AI models are independent of each other. In this way, each of the multiple generative AI models generates the content associated with the hot topics, which helps to give full play to the design advantages of parallel processing of multiple models. The content generation speed can be increased by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative AI models.

[0054] In some embodiments, the system prompt words further include the weight ratios of the respective generative AI models. The weight ratios of the respective generative AI models are used to indicate the importance degree of the content associated with the hot topic generated by each of the generative AI models relative to the stylized content for the hot topic. And when there is a generative AI model being updated among the multiple generative AI models and the weight ratio of the generative AI model being updated is not higher than a preset threshold, the old model of the generative AI model being updated is used to ensure that the parallel operation of the multiple generative AI models is not interrupted. As described above, by separating the content generation and stylization processes into two consecutive stages, the generative AI model can be updated without interrupting the business. For example, if the system prompt words indicate that the main purpose is to provide hot news about a certain hot topic issue, then the weight ratio of the generative AI model responsible for generating the event context may be relatively low. This means that while updating the generative AI model for the event context, the generation of the analysis report can be maintained, and the impact of using the old model on the report quality is very small. In this way, through the separate design of content generation and stylization, each of the multiple generative AI models is used to generate the content associated with the hot topic, which helps to give full play to the advantages of the parallel processing design of multiple models. The content generation speed can be increased by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative AI models. By stylizing the content associated with the hot topic generated by each of the multiple generative AI models that have been generated, the free combination of core scenario content and content style conversion can be supported, and the generation of content in different styles can be supported to meet the personalized needs of different audience groups. Further, in the face of the situation of model update that may occur from time to time, by utilizing the design feature that the multiple generative AI models are optimized independently of each other, the old model of the model being updated with relatively low importance degree can be used for the parallel operation of multiple models, thus ensuring that the business is not interrupted and the impact on the quality of the finally generated content is also limited, which is beneficial to meeting the continuity and stability of the business.

[0055] In a possible implementation, through the style conversion module, based on the system prompt words that at least include the style prompt words, and using the content associated with the hot topic generated by each of the multiple generative artificial intelligence models, the stylized content for the hot topic is generated, including: integrating the content associated with the hot topic generated by each of the multiple generative artificial intelligence models respectively according to multiple preset standard styles, so as to obtain multiple standardized contents for the hot topic, and each of the multiple standardized contents for the hot topic has the multiple preset standard styles; based on the style prompt words, selecting the preset standard style that matches the style prompt words from the multiple preset standard styles, and determining the standardized content for the hot topic that has the selected preset standard style among the multiple standardized contents for the hot topic as the reference standardized content; based on the style prompt words, selectively performing a content deletion operation and a stylized element addition operation on the reference standardized content, so as to obtain the stylized content for the hot topic with the style corresponding to the style prompt words. Here, by using the separate design of content generation and stylization processing, and by taking advantage of the parallel processing of multiple models, analysis reports in multiple key styles can be generated, and then the key style that matches the user's style preference can be selected as a reference, and further user customization processing can be carried out. In this way, multiple batches of generated content can be generated for key hot topics and key styles, and then screened in combination with specific hot topics, which helps to improve the overall efficiency and meet the personalized needs of different audience groups. In addition, based on the determination of the reference standardized content and then performing stylization processing, the combination of core scenario content can be used to generate content that better and faster meets the user's customization requirements.

[0056] In some embodiments, the multiple preset standard styles are key styles determined based on user demand prediction. In this way, it helps to improve the overall efficiency and meet the personalized needs of different audience groups.

[0057] In some embodiments, the hot topic is a key hot topic determined based on the user demand prediction. In this way, it helps to improve the overall efficiency and meet the personalized needs of different audience groups.

[0058] In a possible implementation, the style prompt words are selected from a variety of preset style prompt words, and the variety of preset style prompt words include a standard style, a consulting style, an advisor style, a relaxed style, and a concise style. In some examples, depending on the display platform of the finally generated content, such as chat software on mobile phones, online video platforms, and any Internet media platform, etc., various restrictions may be imposed on the overall data scale of the generated content, such as the total number of words, the total number of lines, the total data length, etc. These can be understood as part of the personalized requirements. Thus, in the stage of stylization processing, through content deletion operations, that is, selectively performing content deletion operations, the content specifications on different platforms can be met. Here, different preset style prompt words can take into account the content specifications on different platforms. For example, one preset style prompt word corresponds to the content specifications on one platform. For example, the consulting style can correspond to the content specifications on a platform with strong timeliness, such as the publishing platform of short news. For another example, the advisor style can correspond to the content specifications on a platform with strong professionalism, such as requiring in-depth and thorough information analysis and sorting out the complete context of events. For another example, the relaxed style can correspond to the content specifications on a platform with strong entertainment, such as online video platforms, webcast platforms, etc. For another example, the concise style can correspond to the platforms on mobile terminals, such as chat software on mobile phones. For this type of platform, due to the limited visual area of the user display terminal, the generated content is required to be as concise as possible for easy display to the terminal users. In this way, by providing a variety of preset style prompt words, it helps to meet the personalized needs of different audience groups.

[0059] In a possible implementation, the multiple generative artificial intelligence models include a first generative artificial intelligence model, a second generative artificial intelligence model, a third generative artificial intelligence model, a fourth generative artificial intelligence model, and a fifth generative artificial intelligence model. Among them, the optimization strategy of the first generative artificial intelligence model has an optimization goal for the relevance and timeliness of the model input. The optimization strategy of the second generative artificial intelligence model has an optimization goal for the information extraction and in-depth research of the model input. The third generative artificial intelligence model has an optimization goal for the trend prediction of the model input. The optimization strategy of the fourth generative artificial intelligence model has an optimization goal for the analysis results of multiple industries of the model input. The optimization strategy of the fifth generative artificial intelligence model has an optimization goal for the analysis results of the event context of the model input. The respective optimization strategies of the multiple generative artificial intelligence models have different optimization goals for the model input, and the multiple generative artificial intelligence models are optimized independently of each other. This means that for the same model input, such as the same prompt representing a hot topic, different generative artificial intelligence models will generate content with different focuses based on the optimization strategies with different optimization goals. By setting different optimization goals and optimizing the multiple generative artificial intelligence models independently of each other, it can ensure that the content generated by the multiple generative artificial intelligence models has better overall content richness and can provide different analysis perspectives. Considering that when a single generative artificial intelligence model generates content based on a model input (such as a prompt), the content it generates has a certain degree of uncontrollability and instability. For example, for two consecutive content generation requests for the same prompt by the same generative artificial intelligence model, it may generate content with certain differences. Therefore, the hot topic is input into the multiple generative artificial intelligence models in the hot topic analysis module to obtain the content associated with the hot topic generated by each of the multiple generative artificial intelligence models. In this way, using the different optimization goals of the respective optimization strategies of the multiple generative artificial intelligence models helps to control the uncontrollability and instability of the content generated by the multiple generative artificial intelligence models as a whole. In other words, compared with the content generated by a single generative artificial intelligence model for a given prompt (such as a hot event between two countries), the content generated by multiple generative artificial intelligence models for the given prompt is generally more stable and controllable as a whole. Moreover, by setting different optimization goals for the multiple generative artificial intelligence models as the optimization directions of their respective optimization strategies, it can also increase the overall content coverage range and help avoid missing important content features or having insufficient output of key content features.Thus, by providing the above-mentioned first generative artificial intelligence model, second generative artificial intelligence model, third generative artificial intelligence model, fourth generative artificial intelligence model, and fifth generative artificial intelligence model, and utilizing the respective optimization strategies and different optimization objectives of these generative artificial intelligence models, it helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be increased by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative artificial intelligence models.

[0060] In some embodiments, the stylized content for the hot topic is a hot analysis report for the hot topic. The content associated with the hot topic generated by the first generative artificial intelligence model is the hot news scenario content associated with the hot topic. The content associated with the hot topic generated by the second generative artificial intelligence model is the research report analysis scenario content associated with the hot topic. The content associated with the hot topic generated by the third generative artificial intelligence model is the external market analysis scenario content associated with the hot topic. The content associated with the hot topic generated by the fourth generative artificial intelligence model is the multi-industry analysis scenario content associated with the hot topic. The content associated with the hot topic generated by the fifth generative artificial intelligence model is the event development context scenario content associated with the hot topic. Thus, through the separate design of content generation and stylization processing, and by using multiple generative artificial intelligence models to generate the content associated with the hot topic respectively, it helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be increased by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative artificial intelligence models. By stylizing the content associated with the hot topic generated by each of the multiple generative artificial intelligence models that have been generated, it can support the free combination of core scenario content and content style conversion, support the generation of content with different styles to meet the personalized needs of different audience groups. Thus, it can not only cover the audience groups and relevant information within different topic industries, but also provide results with different levels of content richness and different analysis perspectives, which is beneficial to meeting the needs of different audience groups in a targeted manner. It has the advantages of low time cost, high execution efficiency, being able to capture hot topics in a timely manner, and providing targeted services for different customer needs. Moreover, through the parallel operation of multiple models, the utilization rate of model resources is improved, and with the rapid changes in hot topics and the rapid changes in the needs of the audience group, through content recombination, rapid response and wide coverage can be achieved, which is beneficial to improving the overall efficiency.

[0061] In some embodiments, the style prompt words are used to set the combination between the scene contents generated by each of the multiple generative AI models. In this way, through the separate design of content generation and stylization processing, each of the multiple generative AI models is used to generate the content associated with the hot topic. This helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be increased by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative AI models. By stylizing the content associated with the hot topic generated by each of the multiple generative AI models that have been generated, the free combination of core scene content and content style conversion can be supported, and the generation of content with different styles can be supported to meet the personalized needs of different audience groups. In this way, not only can the audience groups and relevant information within the scope of different thematic industries be covered, but also results with different content richness and different analysis angles can be provided, which is beneficial to meeting the needs of different audience groups in a targeted manner. It has the advantages of low time cost, high execution efficiency, being able to capture hotspots in a timely manner, and providing targeted services for different customer needs. Moreover, through the parallel operation of multiple models, the utilization rate of model resources is improved, and with the rapid changes in hot topics and the rapid changes in the needs of the audience group, through content recombination, rapid response and wide coverage can be achieved, which is beneficial to improving the overall efficiency.

[0062] In some embodiments, the hot news scenario content associated with the hot topic is based on the vectorized hot news information associated with the hot topic. The event development context scenario content associated with the hot topic is the output result of the event development context sorted by the large language model according to the time axis associated with the hot topic. The external market analysis scenario content associated with the hot topic and the multi-industry analysis scenario content associated with the hot topic are obtained through hot topic analysis by an artificial intelligence agent and an analysis framework. In this way, through the separate design of content generation and stylization processing, multiple generative artificial intelligence models are used to generate the content associated with the hot topic respectively. This helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be increased by increasing the parallelism of model operation. The overall coverage can be expanded by increasing the differences between different generative artificial intelligence models. By stylizing the content associated with the hot topic generated by each of the multiple generative artificial intelligence models that have been generated, the free combination of core scenario content and content style conversion can be supported, and the generation of content with different styles can be supported to meet the personalized needs of different audience groups. In this way, not only can the audience groups and relevant news within the scope of different topic industries be covered, but also results with different content richness and different analysis angles can be provided, which is conducive to meeting the needs of different audience groups in a targeted manner, with the advantages of low time cost, high execution efficiency, being able to capture hot topics in a timely manner, and providing targeted services for different customer needs. Moreover, through the parallel operation of multiple models, the utilization rate of model resources is improved, and with the rapid changes in hot topics and the rapid changes in the needs of the audience group, through content recombination, rapid response and wide coverage can be achieved, which is conducive to improving the overall efficiency.

[0063] In a possible implementation, the content automatic generation method further includes: through the style conversion module, based on another style prompt word different from the style prompt word, using the content associated with the hot topic generated by each of the multiple generative artificial intelligence models, generating another stylized content for the hot topic, where the another stylized content for the hot topic has a style corresponding to the another style prompt word. In this way, through the separate design of content generation and stylization processing, each of the multiple generative artificial intelligence models is used to generate the content associated with the hot topic, which helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be increased by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative artificial intelligence models. By performing stylization processing on the content associated with the hot topic generated by each of the multiple generative artificial intelligence models that have been generated, the free combination of core scenario content and content style conversion can be supported, and the generation of content in different styles can be supported to meet the personalized needs of different audience groups. In this way, not only can the audience groups and relevant information within the scope of different topic industries be covered, but also the results with different content richness levels and different analysis angles can be provided, which is beneficial to meeting the needs of different audience groups in a targeted manner. It has the advantages of low time cost, high execution efficiency, being able to capture hotspots in a timely manner, and providing targeted services for different customer needs. Moreover, through the parallel operation of multiple models, the utilization rate of model resources is improved, and with the rapid changes in hot topics and the rapid changes in the needs of the audience group, through content recombination, rapid response and wide coverage can be achieved, which is beneficial to improving the overall efficiency.

[0064] In a possible implementation manner, the hot topic is user-defined, or the hot topic is obtained by processing public domain hot news data through the hot topic screening module based on a hot topic capture algorithm, a hot topic trend analysis algorithm, and a hot word analysis algorithm. In this way, through the separate design of content generation and stylization processing, multiple generative artificial intelligence models are used to separately generate content associated with the hot topic. This helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be increased by increasing the parallelism of model operation. The overall coverage can also be expanded by increasing the differences between different generative artificial intelligence models. By performing stylization processing on the content associated with the hot topic generated by each of the multiple generative artificial intelligence models that have been generated, the free combination of core scenario content and content style conversion can be supported, and the generation of content with different styles can be supported to meet the personalized needs of different audience groups. In this way, not only can the audience groups and relevant information within the scope of different topic industries be covered, but also results with different content richness levels and different analysis angles can be provided, which is beneficial to meeting the needs of different audience groups in a targeted manner, and has the advantages of low time cost, high execution efficiency, being able to capture hot topics in a timely manner, and providing targeted services for different customer needs. Moreover, through the parallel operation of multiple models, the utilization rate of model resources is improved, and with the rapid changes in hot topics and the rapid changes in the needs of the audience group, through content recombination, rapid response and wide coverage can be achieved, which is beneficial to improving the overall efficiency.

[0065] Figure 4 Schematic diagram of a content automatic generation system provided by an embodiment of this application. As Figure 4As shown, the content automatic generation system 400 includes a hot topic screening module 401, a hot topic analysis module 403, and a style conversion module 405. Among them, the hot topic screening module 401 is used to obtain hot topics. The hot topic analysis module 403 includes multiple generative artificial intelligence models. The hot topic analysis module 403 is configured to: input the hot topic into the multiple generative artificial intelligence models, so as to obtain the content associated with the hot topic generated by each of the multiple generative artificial intelligence models, and transmit the content associated with the hot topic generated by each of the multiple generative artificial intelligence models to the style conversion module 405. Among them, the optimization strategies of each of the multiple generative artificial intelligence models have different optimization objectives for the model input, and the multiple generative artificial intelligence models are optimized independently of each other. The style conversion module 405 is configured to: based on a system prompt word that at least includes a style prompt word, use the content associated with the hot topic generated by each of the multiple generative artificial intelligence models to generate a stylized content for the hot topic, where the stylized content for the hot topic has a style corresponding to the style prompt word. Figure 4 Exemplarily shown in [the figure] is that the multiple generative artificial intelligence models in the hot topic analysis module 403 include a first generative artificial intelligence model 411, a second generative artificial intelligence model 412, a third generative artificial intelligence model 413, a fourth generative artificial intelligence model 414, and a fifth generative artificial intelligence model 415.

[0066] Figure 4The content generation system shown, through the separate design of content generation and stylization processing, uses multiple generative artificial intelligence models to individually generate content associated with the hot topics. This helps to fully utilize the design advantages of multi-model parallel processing. The content generation speed can be increased by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative artificial intelligence models. By stylizing the content associated with the hot topics generated by each of the multiple generative artificial intelligence models that have been generated, it is possible to support the free combination of core scenario content and content style conversion, and support the generation of content in different styles to meet the personalized needs of different audience groups. In this way, not only can it cover the audience groups and relevant information within the scope of different thematic industries, but also it can provide results with different levels of content richness and different analysis perspectives, which is conducive to meeting the needs of different audience groups in a targeted manner. It has the advantages of low time cost, high execution efficiency, being able to promptly capture hot topics, and providing targeted services for different customer needs. Moreover, through the parallel operation of multiple models, the utilization rate of model resources is improved, and with the rapid changes in hot topics and the rapid changes in the needs of the audience group, through content recombination, rapid response and wide coverage can be achieved, which is conducive to improving the overall efficiency.

[0067] Refer to Figure 4, in a possible implementation, the hot topic is used to indicate the content features of the stylized content for the hot topic, the style prompt word is used to indicate the style features of the stylized content for the hot topic, the system prompt word further includes a plurality of task prompt words, the plurality of task prompt words correspond to the plurality of generative artificial intelligence models one by one, and each task prompt word in the plurality of task prompt words is used for the model service call of the generative artificial intelligence model corresponding to the task prompt word in the plurality of generative artificial intelligence models. As described above, through the separate design of content generation and stylization processing, the content associated with the hot topic is generated respectively by a plurality of generative artificial intelligence models, which helps to give full play to the design advantages of multi-model parallel processing. The content generation speed can be improved by increasing the parallelism of model operation, and the overall coverage can be expanded by increasing the differences between different generative artificial intelligence models. By stylizing the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models that have been generated, it is possible to support the free combination of core scenario content and content style conversion, support the generation of content in different styles so as to meet the personalized needs of different audience groups. Here, the hot topic is used to indicate the content features of the stylized content for the hot topic, so it is applicable to the second stage of content generation; the style prompt word is used to indicate the style features of the stylized content for the hot topic, so it is applicable to the third stage of stylization processing. Combining with a plurality of task prompt words, the model service call of the generative artificial intelligence model corresponding to the task prompt word in the plurality of generative artificial intelligence models can be executed more effectively. Thus, it helps to improve the utilization efficiency of model resources.

[0068] In a possible implementation manner, the style conversion module 405 is configured to: based on the system prompt words that at least include the style prompt words, and utilize the content associated with the hot topic generated by each of the multiple generative artificial intelligence models, generate the stylized content for the hot topic, including: integrating the content associated with the hot topic generated by each of the multiple generative artificial intelligence models according to the standard style, so as to obtain the standardized content for the hot topic with the standard style; based on the style prompt words, selectively perform a content deletion operation and a stylized element addition operation on the standardized content for the hot topic, so as to obtain the stylized content for the hot topic with the style corresponding to the style prompt words. In this way, a separate design of content generation and stylization processing is achieved, and the maximum possible coverage range obtained in the content generation stage is fully utilized. Using the content deletion operation to transform the standardized content for the hot topic with the standard style into the stylized content for the hot topic with the style corresponding to the style prompt words not only improves the content generation speed by using multi-model parallel operation and the enlarged overall coverage range, but also improves the stylization processing speed by deleting the content in the enlarged overall coverage range, which is beneficial to improving the overall efficiency. Moreover, using the stylized element addition operation to add stylized elements is beneficial to meeting the personalized needs of different audience groups.

[0069] In a possible implementation manner, the multiple generative artificial intelligence models include a first generative artificial intelligence model 411, a second generative artificial intelligence model 412, a third generative artificial intelligence model 413, a fourth generative artificial intelligence model 414, and a fifth generative artificial intelligence model 415. Among them, the optimization strategy of the first generative artificial intelligence model 411 has an optimization goal for the relevance and timeliness of the model input, the optimization strategy of the second generative artificial intelligence model 412 has an optimization goal for the information extraction and in-depth research of the model input, the third generative artificial intelligence model 413 has an optimization goal for the trend prediction of the model input, the optimization strategy of the fourth generative artificial intelligence model 414 has an optimization goal for the multi-industry analysis results of the model input, and the optimization strategy of the fifth generative artificial intelligence model 415 has an optimization goal for the event context analysis results of the model input. In this way, by using the respective optimization strategies and different optimization goals of these generative artificial intelligence models, it is helpful to give full play to the design advantages of multi-model parallel processing. The content generation speed can be improved by increasing the parallelism of model operation, and the overall coverage range can also be enlarged by increasing the differences between different generative artificial intelligence models.

[0070] Refer to Figure 1, Figure 2 , Figure 3 and Figure 4, in some embodiments, a content automatic generation method, computer device, medium and system provided by the specific embodiments and implementation manners of the present application include: providing a system function for generating an analysis report based on hotspots; providing a system function for generating a time context of a hot event with the aid of an artificial intelligence large model for a hot event; providing a system function for generating different content styles for different hot events and different user groups; the provided analysis report generally includes multiple modules, including but not limited to: relevant industry research reports, market analysis, industry analysis, event context sorting, and content style evolution. In some embodiments, the content automatic generation method, computer device, medium and system provided by the present application can be applied to the processing of hot news information in the financial field, and provides a content generation solution for integrating public domain hotspots with internal information, research reports and other data into an event context in the financial field, and at the same time generating hot analysis in combination with prompt engineering and large model technology. The content generation solution includes the following steps. (1) Integrating public domain hotspots with internal information, research reports and other data and generating hot analysis in combination with prompt engineering and large model technology, specifically, it can include: writing a method for segmenting and vectorizing financial news, and constructing a financial information knowledge base in the financial field; writing system prompt engineering code to implement the setting of information such as the proxy role of the artificial intelligence model, domain common sense, special variables, processing tasks, response format, etc.; writing a method for obtaining public domain hotspots and a method for text vectorization; writing a method for retrieving relevant information data according to the public domain hotspot vector in the financial information knowledge base in the financial field; writing a method for filtering and sorting relevant information to extract target knowledge; writing code for an adapter for calling different large model services; writing a method for obtaining system prompts for hot analysis tasks; writing a method for docking the large model adapter according to associated knowledge and relevant prompts. (2) Integrating public domain hotspots with internal information, research reports and other data and integrating them into an event context in the financial field in combination with large model technology, specifically, it can include: writing a method for segmenting and vectorizing financial news, and constructing a financial information knowledge base in the financial field; writing system prompt engineering code to implement the setting of information such as the proxy role of the artificial intelligence model, domain common sense, special variables, processing tasks, response format, etc.; writing a method for obtaining public domain hotspots and a method for text vectorization; writing a method for retrieving relevant information data according to the public domain hotspot vector in the financial information knowledge base in the financial field; writing a method for filtering and sorting relevant information data to extract target knowledge; writing code for an adapter for calling different large model services; writing system prompts for obtaining event context tasks; writing a method for docking the large model adapter according to associated knowledge and relevant prompts.Thus, the content generation solution has the following advantages: matching news, research reports, and information related to relevant hotspots through large model technology and vectorization technology; by constructing an analysis agent and inputting exclusive analysis frameworks and relevant data, it can analyze various hotspots and output exclusive results such as market analysis and industry analysis; customizable styles: by training style conversion prompt words, optimizing diversified styles and the results output in the styles set by users independently, to meet the needs of different content style audience groups; automatically generated event context based on the timeline: through the large language model's understanding of relevant data, the event context sorted by the timeline is output as a result. Further, the content generation solution quickly completes the screening of hotspots through hotspots aggregation display, accurately outputs hotspot analysis through large model prompt word training, and quickly completes the retrieval of relevant information through vectorized content library matching, reducing the generation of manual time costs.

[0071] Figure 5 FIG. is a schematic structural diagram of a computing device provided by an embodiment of the present application. The computing device 500 includes: one or more processors 510, a communication interface 520, and a memory 530. The processor 510, the communication interface 520, and the memory 530 are interconnected through a bus 540. Optionally, the computing device 500 may further include an input / output interface 550, and the input / output interface 550 is connected to input / output devices for receiving parameters set by users, etc. The computing device 500 can be used to implement part or all of the functions of the device embodiment or system embodiment in the above embodiment of the present application; the processor 510 can also be used to implement part or all of the operation steps of the method embodiment in the above embodiment of the present application. For example, the specific implementation of the computing device 500 performing various operations can refer to the specific details in the above embodiments, such as the processor 510 being used to execute part or all of the steps or part or all of the operations in the above method embodiment. For another example, in the embodiment of the present application, the computing device 500 can be used to implement part or all of the functions of one or more components in the above device embodiment. In addition, the communication interface 520 can specifically be used for the communication functions necessary to implement the functions of these devices and components, etc., and the processor 510 can specifically be used for the processing functions necessary to implement the functions of these devices and components, etc.

[0072] It should be understood that Figure 5The computing device 500 may include one or more processors 510, and the multiple processors 510 may cooperate to provide processing capabilities in a parallel connection mode, a serial connection mode, a serial-parallel connection mode, or any connection mode, or the multiple processors 510 may form a processor sequence or a processor array, or the multiple processors 510 may be divided into a main processor and an auxiliary processor, or the multiple processors 510 may have different architectures such as a heterogeneous computing architecture. Additionally, Figure 5 For the computing device 500 shown, the related structural and functional descriptions are exemplary and non-limiting. In some exemplary embodiments, the computing device 500 may include more or fewer components than Figure 5 shown, or combine certain components, or split certain components, or have a different component arrangement.

[0073] The processor 510 can have multiple specific implementation forms. For example, the processor 510 can include one or a combination of a central processing unit (CPU), a graphic processing unit (GPU), a neural-network processing unit (NPU), a tensor processing unit (TPU), or a data processing unit (DPU), etc. The embodiments of the present application do not make specific limitations. The processor 510 can also be a single-core processor or a multi-core processor. The processor 510 can be a combination of a CPU and a hardware chip. The above-mentioned hardware chip can be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The above-mentioned PLD can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof. The processor 510 can also be implemented by a logic device with built-in processing logic alone, such as an FPGA or a digital signal processor (DSP), etc. The communication interface 520 can be a wired interface or a wireless interface for communicating with other modules or devices. The wired interface can be an Ethernet interface, a local interconnect network (LIN), etc., and the wireless interface can be a cellular network interface or use a wireless local area network interface, etc.

[0074] The memory 530 can be a non-volatile memory, for example, a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The memory 530 can also be a volatile memory, and the volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM). The memory 530 can also be used to store program code and data, so that the processor 510 can call the program code stored in the memory 530 to execute some or all of the operation steps in the above method embodiments, or execute the corresponding functions in the above device embodiments. In addition, the computing device 500 may include more or fewer components than Figure 5 shown, or have a different component configuration.

[0075] The bus 540 can be a Peripheral Component Interconnect Express (PCIe) bus, or an Extended Industry Standard Architecture (EISA) bus, a Unified Bus (Ubus or UB), a Compute Express Link (CXL), a Cache Coherent Interconnect for Accelerators (CCIX), etc. The bus 540 can be divided into an address bus, a data bus, a control bus, etc. In addition to including a data bus, the bus 540 can also include a power bus, a control bus, and a status signal bus, etc. However, for the sake of clarity,Figure 5 It is represented by only one thick line, but it does not mean that there is only one bus or one type of bus.

[0076] The method and device provided by the embodiments of the present application are based on the same inventive concept. Since the principles of the method and the device for solving problems are similar, the embodiments, implementation manners, examples or implementation modes of the method and the device can be referred to each other, and the repeated parts will not be described again. The embodiments of the present application also provide a system, which includes a plurality of computing devices, and the structure of each computing device can refer to the structure of the computing device described above. The functions or operations that the system can implement can refer to the specific implementation steps in the above method embodiments and / or the specific functions described in the above device embodiments, and will not be described again here.

[0077] The embodiments of the present application also provide a computer-readable storage medium, in which computer instructions are stored. When the computer instructions run on a computer device (such as one or more processors), the method steps in the above method embodiments can be implemented. The specific implementation of the processor of the computer-readable storage medium when executing the above method steps can refer to the specific operations described in the above method embodiments and / or the specific functions described in the above device embodiments, and will not be described again here.

[0078] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. The present application can take the form of an all-hardware embodiment, an all-software embodiment, or an embodiment combining software and hardware aspects. The embodiments of the present application can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The present application can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The computer program product includes one or more computer instructions. When the computer program instructions are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (such as coaxial cable, optical fiber, digital subscriber line) or wireless (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more sets of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media, or semiconductor media. The semiconductor media can be a solid-state drive, or random access memory, flash memory, read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, or any other suitable form of storage medium.

[0079] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. Each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks. These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions in the process Figure 1One process or multiple processes and / or blocks Figure 1 The functions specified in one block or multiple blocks. These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide for implementing the process Figure 1 One process or multiple processes and / or blocks Figure 1 The steps of the functions specified in one block or multiple blocks.

[0080] In the above embodiments, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. The steps in the method embodiments of the present application can be adjusted, combined or deleted according to actual needs; the modules in the system embodiments of the present application can be divided, combined or deleted according to actual needs. If these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these changes and variations.

Claims

1. A method for automatically generating content, characterized in that, The method for automatically generating the content includes: Obtaining a hot topic through a hot topic screening module; Inputting the hot topic into multiple generative artificial intelligence models in a hot topic analysis module, thereby obtaining the content associated with the hot topic generated by each of the multiple generative artificial intelligence models, and transmitting the content associated with the hot topic generated by each of the multiple generative artificial intelligence models to a style conversion module, wherein the optimization strategies of each of the multiple generative artificial intelligence models have different optimization objectives for the model input, and the multiple generative artificial intelligence models are optimized independently of each other; Generating, through the style conversion module, stylized content for the hot topic based on a system prompt word that at least includes a style prompt word, using the content associated with the hot topic generated by each of the multiple generative artificial intelligence models, wherein the stylized content for the hot topic has a style corresponding to the style prompt word.

2. The content automatic generation method according to claim 1, characterized in that The hot topic is used to indicate the content characteristics of the stylized content for the hot topic, the style prompt word is used to indicate the style characteristics of the stylized content for the hot topic, the system prompt word further includes a plurality of task prompt words, the plurality of task prompt words correspond one-to-one with the multiple generative artificial intelligence models, and each task prompt word in the plurality of task prompt words is used for the model service call of the generative artificial intelligence model corresponding to the task prompt word in the multiple generative artificial intelligence models.

3. The method for automatically generating content according to claim 2, wherein The system prompt word is provided by a prompt word arrangement module to the hot topic analysis module and the style conversion module, and the function of the prompt word arrangement module for generating the plurality of task prompt words is based on a content prompt word, the content preference corresponding to the content prompt word, and the content associated with the content prompt word generated by each of the multiple generative artificial intelligence models.

4. The method for automatically generating content according to claim 2, wherein The plurality of task prompt words include the settings of each of the multiple generative artificial intelligence models regarding artificial intelligence agents, domain common sense, special variables, processing tasks, and response formats.

5. The content automatic generation method according to claim 1, characterized in that Generating, through the style conversion module, the stylized content for the hot topic based on the system prompt word that at least includes the style prompt word, using the content associated with the hot topic generated by each of the multiple generative artificial intelligence models, includes: Integrating the content associated with the hot topic generated by each of the multiple generative artificial intelligence models according to a standard style, thereby obtaining standardized content for the hot topic with the standard style; Based on the style prompt word, selectively performing a content deletion operation and selectively performing a stylized element addition operation on the standardized content for the hot topic, thereby obtaining the stylized content for the hot topic with a style corresponding to the style prompt word.

6. The method for automatically generating according to the content described in claim 5, characterized in that, The standard style is independent of the style corresponding to the style prompt word, and the standardized content for the hot topic with the standard style is independent of the style prompt word.

7. The content automatic generation method according to claim 5, wherein The stylized element addition operation includes adding at least one stylized element of the style corresponding to the style prompt word, and the at least one stylized element is selected from a plurality of stylized elements, and the plurality of stylized elements include symbols, expressions, and pictures.

8. The content automatic generation method according to claim 1, characterized in that The plurality of generative AI models provide the content associated with the hot topic generated by each of the plurality of generative AI models through parallel operation of multiple models, and the updates of the plurality of generative AI models are performed independently of each other.

9. The content automatic generation method according to claim 8, characterized in that, The system prompt word further includes the weight ratio of each of the plurality of generative AI models, and the weight ratio of each of the plurality of generative AI models is used to indicate the importance degree of the content associated with the hot topic generated by each of the plurality of generative AI models relative to the stylized content for the hot topic. And when there is a generative AI model being updated among the plurality of generative AI models and the weight ratio of the generative AI model being updated is not higher than a preset threshold, the old model of the generative AI model being updated is used to ensure that the parallel operation of the multiple generative AI models is not interrupted.

10. The content automatic generation method according to claim 1, characterized in that, Through the style conversion module, based on the system prompt word that at least includes the style prompt word, and using the content associated with the hot topic generated by each of the plurality of generative AI models, generate the stylized content for the hot topic, including: Integrate the content associated with the hot topic generated by each of the plurality of generative AI models respectively according to a plurality of preset standard styles, so as to obtain a plurality of standardized contents for the hot topic, and each of the plurality of standardized contents for the hot topic has the plurality of preset standard styles; Based on the style prompt word, select the preset standard style that matches the style prompt word from the plurality of preset standard styles, and determine the standardized content for the hot topic that has the selected preset standard style among the plurality of standardized contents for the hot topic as the reference standardized content; Based on the style prompt word, selectively perform content deletion operations and selectively perform stylized element addition operations on the reference standardized content, so as to obtain the stylized content for the hot topic with the style corresponding to the style prompt word.

11. The method for automatically generating content according to claim 10, characterized in that The plurality of preset standard styles are key styles determined based on user demand prediction.

12. The method for automatically generating content according to claim 11, wherein The hot topic is a key hot topic determined based on user demand prediction.

13. The method for automatically generating according to the content described in claim 1, characterized in that, The style prompt word is selected from a plurality of preset style prompt words, and the plurality of preset style prompt words include standard style, consulting style, advisory style, relaxed style, and concise style.

14. The method for automatically generating according to the content described in claim 1, characterized in that, The multiple generative AI models include a first generative AI model, a second generative AI model, a third generative AI model, a fourth generative AI model, and a fifth generative AI model. Among them, the optimization strategy of the first generative AI model has an optimization goal for the relevance and timeliness of model inputs, the optimization strategy of the second generative AI model has an optimization goal for information refinement and in-depth research of model inputs, the third generative AI model has an optimization goal for trend prediction of model inputs, the optimization strategy of the fourth generative AI model has an optimization goal for multi-industry analysis results of model inputs, and the optimization strategy of the fifth generative AI model has an optimization goal for event context analysis results of model inputs.

15. The content automatic generation method according to claim 14, wherein The stylized content for the hot topic is a hot topic analysis report for the hot topic. The content associated with the hot topic generated by the first generative AI model is hot news scenario content associated with the hot topic. The content associated with the hot topic generated by the second generative AI model is research report analysis scenario content associated with the hot topic. The content associated with the hot topic generated by the third generative AI model is external market analysis scenario content associated with the hot topic. The content associated with the hot topic generated by the fourth generative AI model is multi-industry analysis scenario content associated with the hot topic. The content associated with the hot topic generated by the fifth generative AI model is event development context scenario content associated with the hot topic.

16. The content automatic generation method according to claim 15, characterized in that, The style prompt words are used to set the combination between the scenario contents generated by each of the multiple generative AI models.

17. The method for automatically generating content according to claim 15, characterized in that, The hot news scenario content associated with the hot topic is based on the hot news information associated with the hot topic after vectorization processing. The event development context scenario content associated with the hot topic is the output result of the event development context sorted by the large language model according to the time axis associated with the hot topic. The external market analysis scenario content associated with the hot topic and the multi-industry analysis scenario content associated with the hot topic are obtained through hot topic analysis by an AI agent and an analysis framework.

18. The content automatic generation method according to claim 1, characterized in that, The content automatic generation method further includes: Through the style conversion module, based on another style prompt word different from the style prompt word, using the content associated with the hot topic generated by each of the multiple generative AI models, generate another stylized content for the hot topic, where the another stylized content for the hot topic has a style corresponding to the another style prompt word.

19. The method for automatically generating content according to claim 1, wherein The hot topic is user-defined, or the hot topic is obtained by processing the public domain hot news data by the hot topic screening module based on hot topic capture algorithms, hot topic trend analysis algorithms, and hot word analysis algorithms.

20. A computer device, characterized in that, The computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method according to any one of claims 1 to 19 is implemented.

21. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions. When the computer instructions run on a computer device, the computer device is caused to execute the method according to any one of claims 1 to 19.

22. A content automatic generation system, characterized in that The content automatic generation system includes: A hot topic screening module for obtaining hot topics; A hot topic analysis module, the hot topic analysis module including a plurality of generative artificial intelligence models. The hot topic analysis module is configured to: input the hot topic into the plurality of generative artificial intelligence models to obtain content associated with the hot topic generated by each of the plurality of generative artificial intelligence models, and transmit the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models to a style conversion module. Among them, the respective optimization strategies of the plurality of generative artificial intelligence models have different optimization objectives for model inputs, and the plurality of generative artificial intelligence models are optimized independently of each other; The style conversion module is configured to: based on a system prompt word that at least includes a style prompt word, use the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models to generate a stylized content for the hot topic, where the stylized content for the hot topic has a style corresponding to the style prompt word.

23. The content automatic generation system according to claim 22, characterized in that, The hot topic is used to indicate the content characteristics of the stylized content for the hot topic, the style prompt word is used to indicate the style characteristics of the stylized content for the hot topic, the system prompt word further includes a plurality of task prompt words, the plurality of task prompt words correspond one-to-one with the plurality of generative artificial intelligence models, and each task prompt word in the plurality of task prompt words is used for model service invocation of the generative artificial intelligence model corresponding to the task prompt word in the plurality of generative artificial intelligence models.

24. The content automatic generation system according to claim 22, wherein The style conversion module is configured to: based on the system prompt word that at least includes the style prompt word, use the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models to generate the stylized content for the hot topic, including: Integrating the content associated with the hot topic generated by each of the plurality of generative artificial intelligence models according to a standard style to obtain a standardized content for the hot topic with the standard style; Based on the style prompt word, selectively perform a content deletion operation and a stylized element addition operation on the standardized content for the hot topic, so as to obtain the stylized content for the hot topic with a style corresponding to the style prompt word.

25. The content automatic generation system according to claim 22, wherein The multiple generative artificial intelligence models include a first generative artificial intelligence model, a second generative artificial intelligence model, a third generative artificial intelligence model, a fourth generative artificial intelligence model, and a fifth generative artificial intelligence model. Among them, the optimization strategy of the first generative artificial intelligence model has an optimization goal for the relevance and timeliness of the model input, the optimization strategy of the second generative artificial intelligence model has an optimization goal for the information refinement and in-depth research of the model input, the third generative artificial intelligence model has an optimization goal for the trend prediction of the model input, the optimization strategy of the fourth generative artificial intelligence model has an optimization goal for the multi-industry analysis results of the model input, and the optimization strategy of the fifth generative artificial intelligence model has an optimization goal for the event context analysis results of the model input.

Images