Method and device for generating marketing information, equipment and medium

Abstract

The embodiment of the invention discloses a method and device for generating marketing information, equipment and a medium. According to the scheme, the method comprises the steps that a marketing information generation request for a target product is input into an intelligent agent used for generating the marketing information; the agent generates marketing information based on the marketing information generation request; wherein the marketing information is obtained by inputting a marketing prompt word containing product information of the target product into a large language model; the product information is information about the target product acquired from a knowledge base; the product information at least comprises at least one of function information of the target product or right and interest information which can be obtained by accessing the target product.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to a method and device for generating marketing information, equipment and medium. BACKGROUND

[0002] Marketing content refers to various forms of information materials used by enterprises or brand parties when promoting products or services. The marketing content can be used to convey brand value, attract target audiences, promote sales, etc.

[0003] In actual application, the marketing content can be manually generated based on product operation personnel. However, the manual generation of the marketing content requires the product operation personnel to think for a long time, and the experience of the product operation personnel is difficult to replicate, so that the generation efficiency of the marketing content generated based on the product operation personnel is not high. SUMMARY

[0004] Therefore, the embodiments of the present application provide a method and device for generating marketing information, equipment and medium to improve the generation efficiency of the marketing information.

[0005] To solve the above technical problems, the embodiments of the present application are implemented as follows:

[0006] The method for generating marketing information provided by the embodiments of the present application comprises the following steps.

[0007] The request for generating marketing information for a target product is input into an intelligent agent for generating marketing information.

[0008] The intelligent agent generates marketing information based on the request for generating marketing information. The marketing information is obtained by inputting marketing prompt words containing product information of the target product into a large language model. The product information is information about the target product obtained from a knowledge base. The product information at least includes at least one of function information of the target product or benefit information available by accessing the target product.

[0009] The device for generating marketing information provided by the embodiments of the present application comprises the following steps.

[0010] The request input module is configured to input the request for generating marketing information for a target product into an intelligent agent for generating marketing information.

[0011] The marketing information generation module is configured to generate marketing information based on the marketing information generation request, wherein the marketing information is obtained by inputting a marketing prompt word containing product information of the target product into a large language model, the product information is information about the target product obtained from a knowledge base, and the product information at least includes at least one of function information of the target product or benefit information available by accessing the target product.

[0012] The device for generating marketing information provided by the embodiments of the present specification includes:

[0013] at least one processor; and

[0014] a memory in communication connection with the at least one processor; wherein

[0015] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to implement the method for generating marketing information described above.

[0016] The computer readable medium provided by the embodiments of the present specification stores computer readable instructions executable by a processor to implement the method for generating marketing information described above.

[0017] At least one embodiment provided in the present specification can achieve the following beneficial effects: by inputting a marketing information generation request for a target product into an agent for generating marketing information; then, the agent generates marketing information based on the marketing information generation request, wherein the marketing information is obtained by inputting a marketing prompt word containing product information of the target product into a large language model; the product information is information about the target product obtained from a knowledge base; the product information at least includes at least one of function information of the target product or benefit information available by accessing the target product, thereby the embodiments of the present specification can utilize the agent for generating marketing information to generate marketing information through a large language model, which can improve the marketing information generation efficiency compared with manually generating marketing information.

[0018] On the other hand, the embodiments of the present specification can obtain product information of a target product from a knowledge base, generate a prompt word for inputting into a large language model based on the product information, so that the general large language model can be used to generate marketing information, without fine-tuning a pre-trained model using product information to generate marketing information using the fine-tuned large language model, thereby saving model fine-tuning time and further improving marketing information generation efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present specification or the related art, the following will briefly introduce the drawings needed to be used in the embodiment or related art description. Obviously, the drawings in the following description only some embodiments described in the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0020] Figure 1 is a schematic diagram of an application scenario of a method for generating marketing information provided by an embodiment of the present specification;

[0021] Figure 2 is a flowchart of a method for generating marketing information provided by an embodiment of the present specification;

[0022] Figure 3 is a flowchart of a method for generating marketing information provided by an embodiment of the present specification;

[0023] Figure 4 is a structural schematic diagram of a device for generating marketing information corresponding to Figure 2 provided by an embodiment of the present specification;

[0024] Figure 5 is a structural schematic diagram of a device for generating marketing information corresponding to Figure 2 provided by an embodiment of the present specification. DETAILED DESCRIPTION

[0025] In the following description, a lot of specific details are set forth in order to facilitate a thorough understanding of the present application. However, the present application can be implemented in many different ways than described herein, and those skilled in the art can make similar extensions without departing from the spirit of the present application, so the present application is not limited to the specific implementation disclosed below.

[0026] The terms used in one or more embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of the present application. The singular forms "a", "said" and "the" used in one or more embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in one or more embodiments of the present application means and includes any or all possible combinations of one or more associated listed items.

[0027] It should be understood that, although the terms first, second, etc. can be employed in describing various information in one or more embodiments, the information should not be limited to such terms. These terms are only used to distinguish one particular information from another particular information. For example, a first can be termed a second, and, similarly, a second can be termed a first, without departing from the scope of one or more embodiments. As used herein, the term "if' can be construed to mean "when" or "upon" or "in response to determining" terms in contexts of the present disclosure.

[0028] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards in the relevant region, and provide corresponding operation portal for user to choose authorization or refusal.

[0029] The following is an explanation of the terms involved in the embodiments of the present specification.

[0030] Agent: refers to an autonomous system that can perceive the environment and take actions to achieve goals. Agents can complete tasks through autonomous planning. Specifically, the processing of tasks can be simply described as perception, planning and action, where perception refers to the agent obtaining information from the environment. Planning refers to the decision-making process of the agent to complete the task. Action refers to the action based on the environment and planning. Agents can autonomously take actions to achieve goals and can improve their performance through learning or acquiring knowledge.

[0031] Large Language Model (LLM): refers to a deep learning model in the field of natural language processing that is pre-trained on a large corpus of text using an autoregressive approach. It can understand and generate natural language text. Large language models can learn statistical patterns and semantic information of natural language text to predict the next word or sentence. As the input data set and parameter space continue to expand, the capabilities of large language models will also expand accordingly. It can be applied to machine learning, machine translation, speech recognition, image processing and other fields.

[0032] Prompt: refers to the command or instruction provided by the user when interacting with the large language model. Specifically, it can include questions, keywords, context information, etc., which are used to indicate the actions or outputs generated by the large language model. Users can provide clear prompts to guide the large language model to generate replies that meet expectations, thereby improving the effectiveness and quality of the interaction.

[0033] Prompt Template: A structured text used to generate prompts. Generating prompts based on prompt templates can help users interact with large language models, allowing the models to better understand user intent and generate more accurate and user-specific responses.

[0034] Retrieval-Augmented Generation (RAG): A method that combines information retrieval and language generation techniques. This method aims to enhance the performance and accuracy of processing natural language tasks by retrieving relevant information from external knowledge sources. The core idea of retrieval-augmented generation is to use a retrieval system to find relevant fragments from large-scale text data related to the current task, and then input these fragments as additional context into a large language model, generating more accurate and informative text.

[0035] To solve the defects in the prior art, the present scheme provides the following embodiments:

[0036] Figure 1 is an application scenario diagram of a method for generating marketing information provided by an embodiment of the present specification. As shown in Figure 1 the present specification, the embodiment can input a marketing information generation request for a target product into an agent 100 for generating marketing information, and then generate marketing information in response to the marketing information generation request by the agent 100.

[0037] The agent 100 can be deployed on a server. The server where the agent 100 is deployed can be a standalone physical server, a server cluster or a distributed file system composed of multiple physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and big data and artificial intelligence platforms. Basic cloud computing services such as platform.

[0038] Specifically, the agent 100 can obtain product information of the target product from the knowledge base based on the marketing information generation request, generate marketing prompts based on the product information, input the marketing prompts into the large language model, and obtain the marketing information generated by the large language model.

[0039] The method for generating marketing information provided by the embodiment of the present specification is described below in conjunction with the drawings.

[0040] Figure 2 is a flowchart of a method for generating marketing information provided by an embodiment of the present specification. From the perspective of software, Figure 2The execution subject of the method can be a program module that invokes an agent for generating marketing information. From the hardware perspective, Figure 2 The execution subject of the method can be a server or terminal that deploys an agent for generating marketing information.

[0041] As Figure 2 The method for generating marketing information can include the following steps.

[0042] Step 202: Input a marketing information generation request for a target product into an agent for generating marketing information.

[0043] In actual application, the execution subject can receive a marketing information generation request for a target product, and invoke an agent to generate marketing information based on the marketing information generation request.

[0044] In the embodiments of the present specification, the marketing information can refer to various forms of information materials used by an enterprise or brand party when promoting products or services. The dissemination of marketing information can convey brand value, attract target audiences, promote sales, and the like.

[0045] Optionally, the marketing information for the target product can refer to information materials designed and disseminated around the target product. The marketing information for the target product can promote the target product by highlighting the uniqueness, functional value, use scenario, and additional benefits of the target product, and ultimately promote the acceptance of the target product by the audience users.

[0046] Optionally, the target product can refer to a program product for realizing a certain business function, such as a functional module or applet, and the like. For example, it can be a member service, life service, financial product, insurance product, and the like in an APP.

[0047] As a specific implementation, the target product can also be a physical product, such as home appliances, furniture, cars, clothing, snacks, and the like.

[0048] In the embodiments of the present specification, the marketing information generation request can carry operation demand information provided by a product operator. The operation demand information can be an operation demand for the marketing information, such as a display position of the marketing information, a display time of the marketing information, and the like. For example, a specific marketing information generation request can be "please help me generate a marketing information of product A displayed on the APP opening page", and the like. For example, it can be "please help me generate a marketing information of product B displayed from 8:00 to 9:00", and the like.

[0049] Optionally, the operation demand information can be information in natural language form.

[0050] In an embodiment of the present specification, the marketing information generation request can also carry a product identifier of the target product. The product identifier can be an ID of the target product, a name of the target product, etc. In order to obtain product information of the target product according to the product identifier of the target product.

[0051] In actual application, a user with a marketing information generation requirement, such as a product operator, can provide the agent with a marketing information generation request, such as inputting a marketing information generation request for a target product through a server or terminal deployed with the agent, so that the agent can receive the marketing information generation request for the target product.

[0052] In related technologies, an agent refers to an autonomous system that can perceive the environment and take actions to achieve goals. The agent can complete tasks through autonomous planning. Specifically, the processing of the task can be simply described as perception, planning and action, wherein perception refers to the agent obtaining information from the environment. Planning refers to the decision-making process of the agent to complete the task. Action refers to the action based on the environment and planning. The agent can autonomously take actions to achieve goals, and can improve its performance through learning or acquiring knowledge.

[0053] Optionally, the agent can generate marketing information according to the marketing information generation request based on at least one of the two thinking modes of CoT (Chain of Thought) and REAct (Reasoning and Acting).

[0054] Step 204: The agent generates marketing information based on the marketing information generation request; the marketing information is obtained by inputting a marketing prompt word containing product information of the target product into a large language model; the product information is information about the target product obtained from a knowledge base; and the product information at least includes at least one of function information of the target product or benefit information available by accessing the target product.

[0055] Further, the process of the agent generating marketing information based on the marketing information generation request in step 204 can include: first, the product information of the target product can be obtained from the knowledge base in response to the marketing information generation request; the product information at least includes at least one of function information of the target product or benefit information available by accessing the target product. Then, the marketing prompt word can be generated based on the product information. After that, the marketing prompt word can be input into the large language model to obtain the marketing information generated by the large language model.

[0056] In one or more embodiments of the present specification, a specific scheme for obtaining product information of the target product from the knowledge base is further provided.

[0057] In an embodiment of the present specification, the knowledge base can store product information of one or more products. The product information of a product can be information describing the product, such as information describing the marketing points of the product. The marketing points of the product can refer to points attracting users to access the product, which can be attracting users to purchase the product, browse the product, download the product, and the like.

[0058] Optionally, the product information can include at least one of function information of the product or benefit information obtainable by accessing the product. Specifically, the function information of the product can be information describing the functional value of the product, such as information describing the role of the product, information describing the parameters of the product, and the like. The benefit information of the product can be benefit information obtainable by users accessing the product, such as a coupon, a discount coupon, and the like.

[0059] For example, for the product of a pension account, the product function information of the product can be information describing the role of the pension account, such as "the pension account is a special account for managing and accumulating personal pension funds, aiming to provide economic security for individuals after retirement", and the benefit information of the product can be benefit information obtainable by users opening the pension account, such as "opening a pension account to send a red packet", "opening a pension account to send a coupon", and the like.

[0060] In actual applications, the knowledge base can be constructed in one or more ways. For example, it can be constructed based on expert experience, such as by inputting product information of products in the knowledge base by operating personnel. Alternatively, it can also be constructed based on a large language model, such as providing various information of a product, such as basic information of the product, product introduction information, and related report information, to the large language model, instructing the large language model to output product information of the product from the foregoing various information of the product and store it in the knowledge base, so as to be used in the execution process of the embodiments of the present application.

[0061] Optionally, the agent can obtain the product information of the target product from the knowledge base according to the marketing information generation request.

[0062] In actual applications, the agent can retrieve a product identifier identical to or similar to the product identifier carried by the marketing information generation request from the knowledge base, and then can take the product information of the product corresponding to the identical or similar product identifier in the database as the product information of the target product.

[0063] Specifically, the intelligent agent can perform intent recognition on the marketing information generation request, identify keywords in the marketing information generation request, such as the product identifier of the target product. Then, a search vector can be constructed based on the product identifier of the target product, and the constructed search vector can be matched with the vectorized data of the product information in the knowledge base to obtain a product identifier that is the same as or similar to the product identifier carried by the marketing information generation request, and then the product information of the target product can be obtained. Optionally, in order to quickly obtain the product information of the target product from the knowledge base, the product information in the knowledge base can be pre-processed to obtain the vectorized data of the product information in the knowledge base. Alternatively, the knowledge base of the embodiments of the present specification can store the vectorized data of the product information.

[0064] As a specific implementation, the intelligent agent can obtain the product information of the target product from the knowledge base based on a search enhancement generation technology.

[0065] In the embodiments of the present specification, the product information in the knowledge base can be dynamically updated or in real time. Specifically, the update frequency of the product information in the knowledge base can be greater than or equal to a preset threshold.

[0066] In the embodiments of the present specification, the product information of the target product can be obtained from the knowledge base and then provided to the large language model. The problem of insufficient data provided when interacting with the large language model is solved. It can help the large language model better understand the context of generating marketing information, ensure that the marketing information generated by the large language model is closer to the real application scenario, and effectively improve the accuracy of the generated marketing information.

[0067] In related technologies, a pre-trained model can be fine-tuned using product information to obtain a large language model dedicated to generating marketing information, so as to generate marketing information using the obtained large language model. However, in actual applications, the product information of the product updates very quickly, and it is necessary to fine-tune the model using the product information of the new product, which is a large amount of work and limits the generation efficiency of marketing information. In the embodiments of the present specification, the product information of the target product can be obtained from the knowledge base to provide the large language model to generate marketing information, thereby saving the cost and time of fine-tuning the model, and improving the generation efficiency of marketing information, while reducing the generation cost of marketing information.

[0068] In one or more embodiments of the present specification, a specific scheme for generating a marketing prompt based on the product information is further provided.

[0069] The prompt refers to a command or instruction provided to the large language model when interacting with the large language model, which is used to indicate the action to be performed or the output to be generated by the large language model.

[0070] In the embodiments of the present specification, the marketing prompt word can be a prompt word used to instruct the large language model to generate marketing information.

[0071] Optionally, the marketing prompt word can include product information.

[0072] Further, the marketing prompt word can also include task information instructing the large language model to generate marketing information for the target product based on the product information.

[0073] For example, a specific marketing prompt word can be “The following is the product information of the pension account: ‘The pension account is a special account for managing and accumulating personal pension funds, aiming to provide economic security for individuals after retirement; open a pension account to get a discount coupon.’ Please generate marketing information for the pension account based on the product information of the pension account.”. Among them, “The pension account is a special account for managing and accumulating personal pension funds, aiming to provide economic security for individuals after retirement; open a pension account to get a discount coupon” can be product information, and “Please generate marketing information for the pension account based on the product information of the pension account” can be task information instructing the large language model to generate marketing information for the target product based on the product information.

[0074] As a specific implementation, the marketing prompt word used to input to the large language model can be generated based on a prompt word template.

[0075] The prompt word template is a structured text used to generate prompt words. Based on the pre-set prompt word template, high-quality prompt words can be efficiently generated, so that the large language model can better understand the user's intention and generate more accurate and more user-demand-conforming replies.

[0076] For example, the prompt word template can include a system prompt word. The system prompt word can be description information for the task. The description information can specifically include one or more of role information of the large language model, task information instructing the large language model to complete, task requirement information, and output data format information.

[0077] The role information of the large language model can be role information that the large language model needs to play for this task. For example, the large language model needs to play a product operator, and the role information of the large language model can be "you are a product operator". The task information indicating the large language model to complete can be task information that the large language model needs to complete for this task, such as task information indicating the large language model to generate marketing information for the target product based on product information. The task requirement information can be requirement information for this task. For example, the task requirement information can be requirement information for the marketing information to be colloquial, requirement information for the data size of the marketing information, and the like. The output data format information can be data format information of the data output by the large language model for this task. For example, the output data format information can be JSON format information, text format information, picture format information, and the like.

[0078] In one or more embodiments of the present specification, a specific scheme is further provided for inputting the marketing prompt word into the large language model to obtain the marketing information generated by the large language model.

[0079] In the embodiments of the present specification, the large language model can be a model of the GPT series, such as GPT-3.5, GPT-4, and GPT-4.5 large models. Of course, it can also be a model of other series, such as the Tongyi Qianwen model, the Ant Hundredling large model, the DeepSeek series large model, and the like, which are not limited herein.

[0080] In actual applications, the generated marketing information can be information in various formats, such as at least one of text format, image format, combination of text and image format, dynamic picture format, and video format.

[0081] Optionally, the generated marketing information can be one piece of marketing information, or multiple pieces of marketing information.

[0082] The embodiments of the present specification can generate marketing information by using an intelligent agent for generating marketing information. The user can only need to input a marketing information generation request for a target product, and the intelligent agent can generate marketing information in response to the marketing information generation request. Compared with manually generating marketing information by a product operator, the generation efficiency of the marketing information and the convenience of the marketing information generation can be improved.

[0083] In actual applications, an operator can have limited understanding of product information of a target product, and can only provide some basic information such as a product name and a quantity of marketing information to be generated when submitting a request for generating marketing information of the target product. It is difficult to generate high-quality marketing information only based on the information. In the embodiments of the present specification, product information of the target product can be obtained from the knowledge base to generate prompt words for inputting a large language model, which can help the large language model better understand the context of generating marketing information, so as to obtain high-quality marketing information output by the large language model.

[0084] In the embodiments of the present specification, product information of the target product is obtained from the knowledge base to generate prompt words for inputting a large language model, so that marketing information can be generated by using a general large language model without fine-tuning a pre-trained model based on product information to generate marketing information by using the fine-tuned large language model, thereby saving the cost and time of fine-tuning the model, improving the generation efficiency of marketing information, and reducing the generation cost of marketing information.

[0085] Based on the method, Figure 2 The embodiments of the present specification also provide some implementation manners of the method, which are described below.

[0086] In actual applications, the operation demand information of the operator can include demand information for a specified information position. For example, the operator can specify to promote the target product on an APP opening page. Optionally, the marketing information generation request of the user can be a request for generating marketing information for a specified information position. The embodiments of the present specification can generate marketing information for a specified information position, such as marketing information of an APP opening page.

[0087] Optionally, the marketing information generation request is specifically used to request to generate marketing information for display on a specified information position. The agent can generate marketing information based on the marketing information generation request, specifically including: in response to the marketing information generation request, obtaining position information for describing the specified information position from the knowledge base. The position information specifically includes at least one of position requirement information and position preference information. The position requirement information is used to indicate requirements for information displayed on the information position. The position preference information is used to reflect the user's preference for historical marketing information displayed on the information position. Marketing prompt words are generated based on the product information and the position information.

[0088] In actual application, the marketing information can be displayed at various positions of the user terminal page. For example, the marketing information can be displayed on the opening page or the home page of the APP or the home page of the applet, displayed on the pop-up window or the floating window of the page, or fixedly displayed at the top, the bottom, the middle, or the two sides of the page. The specified information display position of the embodiments of the present specification can include at least one of the opening page, the home page, the pop-up window, the floating window, the top of the page, the bottom of the page, the middle of the page, and the two sides of the page.

[0089] Optionally, the knowledge base can further include display position information for describing the information display position. The display position information can be obtained through product description documents, display position introduction documents, expert experience, or large language models.

[0090] Specifically, the information display position can refer to a position in a web page or an APP page for displaying information. For example, the opening page, the pop-up window, the bottom of the page, the two sides of the page, and the like mentioned above.

[0091] The display position information can refer to information describing the information display position. For example, it can include at least one of display position requirement information and display position preference information.

[0092] Optionally, the display position requirement information can be used to represent requirements for the information displayed in the information display position. For example, the number of words of the displayed information, the format of the displayed information, the content of the displayed information, and the like. Specifically, the number of words of the displayed information can be required to be within 30 words, required to be between 15-20 words, and the like. The format of the displayed information can be required to be in text format, video format, or a combination of text and image format. The content of the displayed information can be required to contain no sensitive content, contain no certain characters or symbols, use a certain font, font size, font color, and the like. Optionally, the display requirement information can be determined according to expert experience or set according to actual needs. Optionally, the data form of the display position requirement information can be a document form or a rule form.

[0093] Optionally, the position preference information can be used to reflect the user's preference for the historical marketing information displayed in the information position. For example, the position preference information can be a style preference for the historical marketing information, such as a preference for a humorous style, a popular science style, a poetic style, etc. It can also be a format preference for the historical marketing information, such as a preference for a text format, an animation format, a video format, etc. It can also be a preference for the product type corresponding to the historical marketing information, such as a preference for service type products, a preference for insurance type products, a preference for home appliance type products, etc. In the embodiments of the present specification, the position preference information can be determined based on historical marketing record information. The historical marketing record information can refer to the click rate, browsing time, etc. of the historical marketing information. Optionally, the historical marketing information can be marketing information for one or more products. Optionally, the data form of the position preference information can be a label form.

[0094] In practical applications, different information positions can correspond to different position information. Of course, according to actual conditions, different information positions can correspond to part of the same position information.

[0095] In the embodiments of the present specification, marketing prompt words can be generated based on product information and position information. The marketing prompt words can include product information and position information. Thus, the large language model can generate marketing information for display in the specified information position based on the product information and the position information. Optionally, the inputting of the marketing prompt words into the large language model to obtain the marketing information generated by the large language model can specifically include: inputting the marketing prompt words into the large language model to obtain the marketing information generated by the large language model with reference to the position information and the product information, which is suitable for display in the specified information position.

[0096] In practical applications, marketing information can also be displayed to terminal users. Optionally, after inputting the marketing prompt words into the large language model to obtain the marketing information generated by the large language model with reference to the position information and the product information, which is suitable for display in the specified information position, the method can further include: displaying the marketing information in the specified information position in the user terminal.

[0097] In the embodiments of the present specification, the position information describing the specified information position can be obtained from the knowledge base, so that the marketing prompt words are generated based on the product information and the position information, so that the large language model can consider the product information and the position information when generating the marketing information, and the click rate, conversion rate, etc. of the marketing information is improved.

[0098] In the embodiments of the present specification, in order to improve the accuracy of the marketing information generation, the reference case information can also be used to generate the marketing prompt word. Optionally, the intelligent agent generates the marketing information based on the marketing information generation request, and specifically can include: obtaining first reference case information corresponding to at least one of the target product and the specified information position from the knowledge base. The first reference case information is the marketing information meeting the user preference selected from the historical marketing information corresponding to at least one of the target product and the specified information position. The marketing prompt word is generated based on the product information, the position information and the first reference case information.

[0099] In the embodiments of the present specification, the first reference case information can be information used by the large language model for reference to generate the marketing information. Specifically, the first reference case information can be historical marketing information.

[0100] Optionally, the first reference case information can be the marketing information meeting the user preference selected from the historical marketing information corresponding to at least one of the target product and the specified information position. Specifically, the historical marketing information corresponding to the target product can be marketing information for the target product. The historical marketing information corresponding to the specified information position can be marketing information for the specified information position. The marketing information meeting the user preference can include marketing information meeting a first preset condition. The first preset condition can include that the user click rate of the first reference case information is higher than the user click rate of other marketing information in the historical marketing information, the user browsing time of the first reference case information is greater than the user browsing time of other marketing information in the historical marketing information, and the like.

[0101] Optionally, the first reference case information can include marketing information meeting the first preset condition selected from the first historical marketing information displayed at the information position. Or, it can include marketing information meeting the first preset condition selected from the second historical marketing information for the target product. Or, it can include marketing information meeting the first preset condition selected from the third historical marketing information for the target product displayed at the information position.

[0102] In actual application, in the process of constructing the knowledge base, the first reference case information corresponding to at least one of the target product and the specified information position can also be stored into the knowledge base. Optionally, the knowledge base can include the first reference case information corresponding to at least one of the target product and the specified information position. The knowledge base can include one or more first reference case information.

[0103] Further, the first reference case information corresponding to at least one of the target product and the specified information position in the knowledge base can be obtained to generate the marketing prompt word. Optionally, the first reference case information obtained from the knowledge base can be one first reference case information or multiple first reference case information, which is not limited herein.

[0104] In the embodiments of the present specification, the marketing prompt word can be generated based on the product information, the position information, and the first reference case information. The marketing prompt word can include the product information, the position information, and the first reference case information. Thus, the large language model can refer to the first reference case information to generate the marketing information for display in the specified information position, improve the accuracy of the generated marketing information, and improve the click rate, conversion rate, and the like of the marketing information.

[0105] In actual application, the operation demand information of the operator can be demand information for a specified user group. For example, the operator can promote products for an elderly user group. Optionally, the user's marketing information generation request can be a request for generating marketing information for a specified user group. The embodiments of the present specification can generate marketing information for a specified user group, such as marketing information for an elderly user group. Optionally, the marketing information generation request is specifically used to request to generate marketing information displayed to a specified user group. The agent can generate marketing information based on the marketing information generation request, specifically including: in response to the marketing information generation request, obtaining group information for describing the specified user group from the knowledge base. The group information specifically includes at least one of group feature information and group preference information. The group feature information is used to reflect the basic attribute characteristics of the specified user group, and the group preference information is used to reflect the preference of the specified user group for historical marketing information. The product information and the group information are used to generate a marketing prompt word.

[0106] In the embodiments of the present specification, the specified user group can be a group divided according to each dimension. For example, it can include a child user group, a teenager user group, an adult user group, and an elderly user group divided by age. It can include a student user group, a job seeker user group, a freelancer user group, and a retiree user group divided by occupation. It can include a city user group, a rural user group, a user group in A province, and a user group in B city divided by geographical area. It can include a high-consumption user group, a medium-consumption user group, and a low-consumption user group divided by consumption ability. It can include a technology enthusiast user group, a sports enthusiast user group, a food enthusiast user group, and a movie and television enthusiast user group divided by interest and hobby. It can also include a junior college user group, a college user group, a graduate student user group, and a doctoral user group divided by education level.

[0107] Optionally, the knowledge base can further include group information for describing a specified user group. The group information of the specified user group can be obtained by statistics on information of the specified user group.

[0108] Optionally, the group information of the specified user group can refer to information for describing the specified user group. For example, the group information can specifically include at least one of group feature information and group preference information.

[0109] The group feature information can be used to reflect the basic attribute features of the specified user group. The basic attribute features can include grade features, education features, occupation features, interest and hobby features, consumption ability features, geographical area features, asset features, etc. Optionally, the data form of the group feature information can be a label form.

[0110] The group preference information can be used to reflect the preferences of the specified user group for historical marketing information. For example, style preferences, format preferences, product function information preferences, product benefit information preferences, etc. for historical marketing information. Specifically, the information about the preferences of the specified user group for historical marketing information can be determined based on historical marketing record information. The historical marketing record information can refer to click rates, browsing times, etc. for historical marketing information. Optionally, the data form of the group preference information can be a label form.

[0111] In the embodiments of the present specification, marketing prompt words can be generated based on product information and group information. The marketing prompt words can include product information and group information. Thus, a large language model can generate marketing information for display to a specified user group based on product information and group information. Optionally, the input of the marketing prompt words into the large language model to obtain the marketing information generated by the large language model can specifically include: inputting the marketing prompt words into the large language model to obtain marketing information generated by the large language model based on the group information and the product information and suitable for display to the specified user group.

[0112] Further, the marketing information can also be displayed to the specified user group. Optionally, the method can further include: displaying the marketing information in a user terminal used by the specified user group.

[0113] In actual application, the operation demand information of the operation personnel can be demand information for a specified information display position and a specified user group. For example, the operation personnel can specify to promote a financial product to a high-consumption user group in a page pop-up window. Optionally, the user's marketing information generation request can be a generation request for marketing information for a specified information display position and a specified user group. The embodiments of the present specification can generate marketing information for a specified information display position and a specified user group, such as marketing information for promoting a financial product to a high-consumption group in a page pop-up window. Optionally, the generating of the marketing prompt based on the product information can specifically include: generating a marketing prompt based on the product information, the display position information, and the group information.

[0114] The inputting of the marketing prompt into the large language model to obtain marketing information generated by the large language model can specifically include: inputting the marketing prompt into the large language model to obtain marketing information generated by the large language model with reference to the group information and the product information, which is suitable for display to the specified user group.

[0115] Further, the marketing information can also be used to display at a specified information display position in a user terminal used by the specified user group. Optionally, the method can further include: displaying the marketing information at the specified information display position in the user terminal used by the specified user group.

[0116] In the embodiments of the present specification, the group information describing the specified user group can be obtained from the knowledge base, so that different marketing information, such as marketing information of different styles, different formats, and different products, can be adaptively generated for different user groups. The user experience is improved, thereby improving the click rate and conversion rate of the marketing information.

[0117] As a specific implementation, in order to improve the click rate and conversion rate of the specified user group for the marketing information, the marketing prompt can also be generated according to the product preference information of the specified user group for the target product, so as to obtain the specified user group preferred marketing information. Optionally, the method of generating marketing information can further include: determining product preference information of the specified user group for the target product based on the product information and the group preference information. The marketing prompt is generated based on the product information, the group information, and the product preference information.

[0118] In actual application, the product information of the target product can include a plurality of product selling point information. The product selling point information can be a certain function information of the product, a certain benefit information of the product, etc. The product selling point information matched with the group preference information can be determined from the plurality of product selling point information according to the group preference information of the specified user group, such as preference for a certain function information, preference for a certain benefit information, etc.

[0119] Further, the marketing prompt words for inputting the large language model can be generated according to the product information, the group information, and the product preference information. Thus, the large language model can generate marketing information suitable for display to the specified user group with reference to the group information, the product information, and the product preference information.

[0120] In the embodiments of the present specification, after obtaining the marketing information generated by the large language model, the marketing information can also be sent to the requestor of the marketing information generation request, such as the product operator, etc., so that the requestor of the marketing information generation request processes the marketing information, such as modifying the marketing information, displaying the marketing information on the user terminal of the user group, storing the marketing information into the knowledge base, etc. Optionally, the method can further include: sending the marketing information to the requestor of the marketing information generation request.

[0121] In the related art, the marketing information can be generated by using the fine-tuned large language model. However, this marketing information generation method is difficult to accurately control the style, content, and user preference of the marketing information, etc., so that personalized marketing information cannot be provided for different user groups. In the embodiments of the present specification, the marketing prompt words can be generated according to at least one of the display information of the information display position, the group information of the user group, the product preference information, etc., and then the personalized marketing information is generated based on the large language model, which is beneficial to improve the click rate and conversion rate of the marketing information.

[0122] In actual application, in order to ensure the quality of the marketing information, the intelligent agent can also perform quality evaluation on the marketing information before sending the marketing information to the requestor of the marketing information generation request. Optionally, the method for generating marketing information can further include: the intelligent agent calling a preset quality evaluation tool to perform quality evaluation on the marketing information to obtain a quality evaluation result; and if the quality evaluation result indicates that the marketing information does not meet a preset quality condition, rewriting the marketing information to obtain rewritten marketing information.

[0123] In the embodiments of the present specification, the quality evaluation tool can be a quality evaluation large model. The quality evaluation large model can be obtained based on the expert experience for training the quality evaluation on the marketing information. Alternatively, the intelligent agent can directly perform quality evaluation on the marketing information through expert experience, such as sending the marketing information to a marketing information evaluation expert for quality evaluation.

[0124] Optionally, the quality evaluation on the marketing information can be evaluation on the syntax, semantics, language specification, word count, accuracy, readability, legal compliance, etc. of the marketing information. Which aspect of the marketing information is evaluated can be determined according to actual needs, which is not limited herein.

[0125] As a specific implementation, if the quality evaluation result indicates that the marketing information does not meet the preset quality condition, the marketing information can be filtered out and the marketing information is regenerated.

[0126] As a specific implementation, if the quality evaluation result indicates that the marketing information does not meet the preset quality condition, the marketing information can also be revised to obtain revised marketing information that meets the preset quality condition. For example, the revision can be performed manually or by a large language model. Optionally, the marketing information generation request requester is sent the marketing information, specifically, the revised marketing information can be sent to the marketing information generation request requester.

[0127] Optionally, if the quality evaluation result indicates that the marketing information meets the preset quality condition, the marketing information can be directly sent to the marketing information generation request requester.

[0128] In practical applications, high-quality marketing information can be stored in a knowledge base to serve as reference case information for a large language model, thereby improving the quality of marketing information generated by the subsequent large language model. Optionally, after generating the marketing information, the method can further include: obtaining a marking operation of an audit personnel on specified marketing information in the marketing information. The marking operation is used to indicate that the specified marketing information meets the condition of serving as a reference case for marketing information; and updating the specified marketing information to the knowledge base in response to the marking operation. The specified marketing information is used as reference knowledge when generating subsequent marketing information.

[0129] In the embodiments of the present specification, the audit personnel can refer to personnel who audit marketing information, and the audit personnel can have experience in auditing marketing information. The audit personnel can perform a marking operation on marketing information that meets the condition of serving as a reference case for marketing information, so that the agent can update the specified marketing information to the knowledge base according to the marking operation of the audit personnel.

[0130] Optionally, the specified marketing information can be used as reference case information for a subsequent large language model, for example, the specified marketing information can be used to generate a prompt word for a large language model, so that the large language model can refer to the specified marketing information to generate subsequent marketing information.

[0131] In practical applications, the marketing information can be sent to a marketing information audit platform, so that the audit personnel can perform a marking operation based on the marketing information audit platform, and the agent can update the specified marketing information to the knowledge base based on the marking operation of the audit personnel in the marketing information audit platform.

[0132] As a specific implementation, the marketing information can be audited by an auditor before being online. If the audit is passed, the marketing information can be online to improve the quality of the online marketing information. Optionally, the method can further include sending the marketing information to a marketing information audit platform.

[0133] The marketing information audit platform obtains audit result information input by an auditor. The audit result information includes first result information or second result information. The first result information is used to indicate that the marketing information passes the audit. The second result information is used to indicate that the marketing information fails the audit.

[0134] If the audit result information is specifically the first result information, the marketing information is sent to a user terminal.

[0135] Optionally, the auditor can perform the audit based on the marketing information audit platform. If the marketing information passes the audit, the marketing information can be sent to a user terminal for displaying the marketing information. If the marketing information fails the audit, the marketing information can be revised or directly abandoned.

[0136] In the embodiments of the present specification, after the marketing information is online, the marketing information can be updated to the knowledge base according to the feedback information of the user for the marketing information, so as to serve as reference knowledge for subsequent generation of marketing information, thereby improving the quality of the subsequent generated marketing information.

[0137] Optionally, the marketing information specifically includes a plurality of marketing information. After the marketing information is generated, the method can further include sending the plurality of marketing information to a plurality of user terminals, sending at least one of the plurality of marketing information to a user terminal, obtaining first feedback information of a terminal user for the plurality of marketing information. The first feedback information includes at least one of interaction information and evaluation information. According to the first feedback information, second reference case information preferred by the terminal user is selected from the plurality of marketing information. The second reference case information is updated to the knowledge base. The second reference case information is used as reference knowledge for subsequent generation of marketing information.

[0138] In the embodiments of the present specification, one marketing prompt word can be input into a large language model to make the large language model generate a marketing information. Thus, a plurality of marketing prompt words can be input into the large language model to obtain a plurality of marketing information corresponding to the plurality of marketing prompt words generated by the large language model. Alternatively, one marketing prompt word can be input into the large language model to make the large language model generate a plurality of marketing information at a time.

[0139] Optionally, the plurality of marketing information can be sent to a plurality of user terminals. For example, the plurality of marketing information can be directly sent to the user terminals. Alternatively, the plurality of marketing information can be sent to a marketing information management platform, and an auditing personnel can perform a marketing information sending operation on the marketing information management platform, so as to send the plurality of marketing information to the user terminals based on the marketing information management platform.

[0140] Specifically, one marketing information can be sent to one user terminal, or a plurality of marketing information can be sent to one user terminal.

[0141] Further, the user terminal can display the marketing information. Different marketing information can be displayed on different user terminals. Alternatively, different marketing information can be displayed on the same user terminal, for example, a first marketing information is displayed on a user terminal at a first time, and a second marketing information is displayed on the user terminal at a second time, which is not limited herein.

[0142] In actual application, the user can feed back the marketing information displayed on the terminal, so as to obtain first feedback information of the terminal user for the marketing information.

[0143] Specifically, the first feedback information can include at least one of interaction information and evaluation information. The interaction information can refer to interaction operation information of the user and the marketing information. For example, information of the user clicking the marketing information, information of the user browsing the marketing information, information of the user accessing a target product corresponding to the marketing information, and the like. The evaluation information can refer to evaluation information of the user for the marketing information. For example, comment information of the user for the marketing information, such as information of the user expressing interest in the marketing information, information of the user expressing disinterest in the marketing information, and the like. For example, questionnaire survey information of the user for the marketing information, which can include information of a display duration of the marketing information that the user can accept, information of a display position of the marketing information that the user can accept, and the like.

[0144] In the embodiments of the present specification, the second reference case information preferred by the terminal user can be screened out according to the first feedback information. The second reference case information can refer to the screened marketing information. For example, marketing information with a high click rate of the user can be screened out from the plurality of marketing information, and marketing information of interest to the user can be screened out. Optionally, the second reference case information preferred by the terminal user can include second reference case information satisfying a second preset condition. The second preset condition can include at least one of the second reference case information having a higher click rate of the user than other marketing information in the plurality of marketing information, and the second reference case information being positive evaluation information of the user. The positive evaluation information can include information of the user expressing interest in the marketing information.

[0145] Optionally, the second reference case information can be updated to the knowledge base, and then used as reference knowledge for generating marketing knowledge subsequently. For example, the second reference case information can be written into a marketing prompt word, so that the large language model can generate subsequent marketing information based on the second reference case information.

[0146] Further, an association relationship between the second reference case information and the product, an association relationship between the second reference case information and the information position, and an association relationship between the second reference case information and the product and the information position can be established. Thus, the corresponding second reference case information can be quickly found based on the above association relationships, and the efficiency of generating marketing information is improved.

[0147] In the embodiments of the present specification, low-quality marketing information that does not meet the preferences of the terminal user can also be filtered out according to the first feedback information. Thus, the low-quality marketing information is not sent to the terminal user. Optionally, after obtaining the first feedback information of the terminal user on the plurality of marketing information, the method can further include: determining, according to the first feedback information, low-quality marketing information in the plurality of marketing information that does not meet the preferences of the terminal user; and offline the low-quality marketing information.

[0148] Optionally, the low-quality marketing information that does not meet the preferences of the terminal user can include marketing information that meets a third preset condition. The third preset condition can include at least one of a user click rate of the marketing information being lower than a preset click rate threshold and the marketing information being negative evaluation information of the user. The negative evaluation information can include information indicating that the user is not interested in the marketing information.

[0149] In the embodiments of the present specification, offline low-quality marketing information can mean that the low-quality marketing information is no longer pushed to the terminal user, but other marketing information is pushed to the terminal user. Specifically, marketing information other than the low-quality marketing information in the plurality of marketing information can be used to replace the low-quality marketing information that is pushed to the terminal user.

[0150] In actual application, the diversity of marketing information can be ensured by controlling the number of marketing information displayed to the terminal user, so as to realize personalized display of marketing information, and then improve the click rate and conversion rate of marketing information. Optionally, after offline the low-quality marketing information, the method can further include: determining whether the number of types of marketing information for the target product pushed to the terminal user is greater than or equal to a preset number threshold, to obtain a number determination result; if the number determination result is no, generating a supplementary marketing prompt word based on the product information; calling the large language model using the supplementary marketing prompt word to obtain supplementary marketing information generated by the large language model; and pushing the supplementary marketing information to at least part of the terminal users.

[0151] In the embodiments of the present specification, the preset quantity threshold can be determined according to actual needs, for example, can be set to 20, 50, 100, etc., which is not limited herein.

[0152] Optionally, if the number of types of marketing information for the target product pushed to the end user is less than the preset quantity threshold, the supplementary marketing prompt word can be generated based on the product information. The supplementary marketing prompt word can be a prompt word for instructing the large language model to regenerate the marketing information.

[0153] Specifically, the supplementary marketing prompt word can be the same prompt word as the marketing prompt word.

[0154] Alternatively, the supplementary marketing prompt word can be a prompt word different from the marketing prompt word. For example, the supplementary marketing prompt word includes reference case information different from the marketing prompt word, and for example, the supplementary marketing prompt word includes instruction information instructing the large language model to generate supplementary marketing information, wherein the supplementary marketing information can be different from the marketing information. Optionally, the supplementary marketing prompt word can include at least one of the display information, the first reference case information, the group information, and the product preference information.

[0155] Optionally, one or more supplementary marketing prompt words can be generated based on the product information.

[0156] In the embodiments of the present specification, the supplementary marketing prompt word can be input into the large language model to obtain the supplementary marketing information generated by the large language model. As a specific implementation, the supplementary marketing information generated by the large language model can be one supplementary marketing information or multiple supplementary marketing information.

[0157] Further, the supplementary marketing information can be pushed to the end user. Specifically, the supplementary marketing information can be pushed to one user, part of the users, or all of the users in the end user. Optionally, one supplementary marketing information can be pushed to one user, or multiple supplementary marketing information can be pushed to one user.

[0158] In actual applications, the decision preference information of the terminal user for the target product can also be determined according to the first feedback information. Therefore, the decision preference information can be updated to the knowledge base, so as to facilitate subsequent generation of marketing information for the target product. Optionally, after the terminal user's first feedback information for the plurality of marketing information is obtained, the method can further include: determining the first decision preference information of the terminal user for the target product based on the first feedback information. The first decision preference information is used to reflect the preference reason of the terminal user for the target product; establishing a first mapping relationship information between the product identifier of the target product and the first decision preference information; and updating the first mapping relationship information to the knowledge base. The first mapping relationship information is used as reference knowledge when generating subsequent marketing information for the target product.

[0159] Specifically, the first decision preference information of the terminal user for the target product can be determined based on the first feedback information and the marketing information. The first decision preference information can reflect the preference reason of the terminal user for the target product. For example, if the marketing information includes product function information and the first feedback information indicates that the user is interested in the marketing information, it can be determined that the terminal user prefers the function of the product. For example, if the marketing information includes product benefit information and the first feedback information indicates that the user accesses the target product corresponding to the marketing information, it can be determined that the terminal user prefers the benefit of the product. For example, if the style preference of the marketing information is a humorous style, and the first feedback information indicates that the user clicks the marketing information, it can be determined that the terminal user prefers the marketing information in a humorous style. For example, if the format of the marketing information is a moving picture format, and the first feedback information indicates that the user browses the marketing information, it can be determined that the terminal user prefers the marketing information in a moving picture format, and so on. The first decision preference information in the embodiments of the present specification can include at least one of product function information, product benefit information, style information of marketing information, and format information of marketing information.

[0160] In the embodiments of the present specification, a first mapping relationship information between the product identifier of the target product and the first decision preference information can be established, so as to facilitate obtaining the first decision preference information corresponding to the target product based on the first mapping relationship information.

[0161] Optionally, the first mapping relationship information can be updated to the knowledge base. Specifically, when obtaining the product information of the target product from the knowledge base, the first mapping relationship can also be obtained. Alternatively, the product information of the target product obtained from the knowledge base can include the first mapping relationship. Therefore, the large language model can obtain the first decision preference information based on the first mapping relationship, and then generate marketing information that meets the user's preference based on the decision preference information, thereby improving the click rate and conversion rate of the marketing information.

[0162] In actual application, decision preference information of different user groups for the target product can be determined, so that marketing information meeting the preferences of different user groups can be generated based on the decision preference information of different user groups, the quality of the marketing information is improved, and then the click rate and conversion rate of the marketing information are improved.

[0163] Optionally, after the first decision preference information of the terminal user for the target product is determined based on the first feedback information, a second mapping relationship information between a first user group and at least part of the first decision preference information can be established according to the first decision preference information of each terminal user for the target product. The second mapping relationship information is used to reflect the group preference of part of the terminal users included in the first user group; the second mapping relationship information is updated to the knowledge base; and the second mapping relationship information is used as reference knowledge when subsequent marketing information displayed to the first user group is generated.

[0164] Optionally, the first user group can be determined from each terminal user. For example, the terminal users can be clustered according to the user characteristics of each terminal user to obtain each user group. The user characteristics of the terminal users belonging to the same user group are the same, for example, the same user label can be carried. Specifically, the first user group determined from each terminal user can refer to any user group, such as a child user group, a job user group, a high-consumption user group, a rural user group, a technology enthusiast user group, a college user group, and the like. Then, the preference information of the first user group can be determined from the first decision preference information. The preference information of the first user group can be at least part of the first decision preference information. For example, the first decision preference information includes function information of a product, benefit information of the product, and style information of marketing information, wherein the decision preference information of the child user group can be the style information of the marketing information, the decision preference information of the job user group can be the benefit information of the product, the preference information of the high-consumption user group can be the function information of the product and the style information of the marketing information, and the like.

[0165] In the embodiments of the present specification, the second mapping relationship information between the first user group and at least part of the first decision preference information can be established, so that the decision preference information corresponding to the first user group can be obtained based on the second mapping relationship information.

[0166] Optionally, the second mapping relationship information can be updated to the knowledge base. When marketing information about the target product needs to be generated for the first user group, the second mapping relationship can be obtained from the knowledge base, and the decision preference information corresponding to the first user group can be obtained based on the second mapping relationship, so as to facilitate the large language model to generate marketing information related to the target product for the first user group based on the decision preference information corresponding to the first user group.

[0167] In actual application, in order to push products to each user group, the decision preference information of each user group can also be determined according to the feedback information of each user group for other products, so as to facilitate the generation of marketing information that meets the preferences of each user group for each product. Optionally, the method can also include:

[0168] Obtaining second feedback information corresponding to another product of the end user. The second feedback information includes at least one of interaction information and evaluation information of the end user for marketing information for the other product.

[0169] Determining second decision preference information of the end user for the other product according to the second feedback information. The second decision preference information is used to reflect the preference reason of the end user for the other product.

[0170] According to the first decision preference information of each end user for the target product and the second decision preference information of each end user for the other product, a third mapping relationship information between a second user group and at least part of the decision preference information in the first decision preference information and the second decision information is established. The third mapping relationship information is used to reflect the group preference of part of the end users included in the second user group.

[0171] Updating the third mapping relationship information to the knowledge base. The third mapping relationship information is used as reference knowledge when subsequent marketing information displayed to the second user group is generated.

[0172] In the embodiments of the present specification, the other product can refer to a product other than the target product. Specifically, the other product can be one product or multiple products.

[0173] Specifically, the second decision preference information can be determined based on the second feedback information and the user's feedback on the marketing information of the other product. The second decision preference information can reflect the reason why the end user prefers the other product. For example, if the marketing information of the other product includes the function information of the product, and the second feedback information indicates that the user is interested in the marketing information, it can be determined that the end user prefers the function of the product. For example, if the marketing information of the other product includes the benefit information of the product, and the second feedback information indicates that the user visits the target product corresponding to the marketing information, it can be determined that the end user prefers the benefit of the product. For example, if the style preference of the marketing information of the other product is a humorous style, and the second feedback information indicates that the user clicks the marketing information, it can be determined that the end user prefers the marketing information in a humorous style. For example, if the format of the marketing information of the other product is an animation format, and the second feedback information indicates that the user browses the marketing information, it can be determined that the end user prefers the marketing information in an animation format, and so on. The second decision preference information in the embodiments of the present specification can include at least one of the function information of the product, the benefit information of the product, the style information of the marketing information, and the format information of the marketing information.

[0174] In the embodiments of the present specification, a third mapping relationship information between the second user group and at least part of the decision preference information in the first decision preference information and the second decision preference information can be established. Thus, it is convenient to obtain the first decision preference information corresponding to the target product based on the first mapping relationship information.

[0175] Optionally, the second user group can refer to any user group, and the second user group can be the same group as the first user group or a different group from the second user group, which is not limited herein.

[0176] Optionally, the third mapping relationship information can be updated to the knowledge base. Specifically, when it is necessary to generate marketing information related to the other product for the second user group, the third mapping relationship can be obtained from the knowledge base, and the decision preference information corresponding to the second user group can be obtained based on the third mapping relationship, so that the large language model generates marketing information related to the other product for the first user group based on the decision preference information corresponding to the second user group.

[0177] In actual applications, in order to improve the diversity of the generated marketing information, different marketing prompt words can also be generated, so that the different marketing prompt words are input into the large language model to obtain different marketing information generated by the large language model. Optionally, the agent can generate marketing information based on the marketing information generation request, specifically, the agent can obtain at least one type of reference knowledge for generating marketing information for the target product from the knowledge base in response to the marketing information generation request. Based on the product information and at least part of the at least one type of reference knowledge, a plurality of marketing prompt words are generated. The first prompt word in the plurality of marketing prompt words is generated based on the product information and the first part of the at least one type of reference knowledge. The second prompt word in the plurality of marketing prompt words is generated based on the product information and the second part of the at least one type of reference knowledge. The first part of the knowledge and the second part of the knowledge contain different types of reference knowledge. The plurality of marketing prompt words are input into a large language model respectively to obtain a plurality of groups of marketing information generated by the large language model; one marketing prompt word corresponds to one group of marketing information.

[0178] Optionally, the agent can obtain various types of reference knowledge from the knowledge base according to the marketing information generation request. For example, at least one of the booth information, the first reference case information, the second reference case information, the group information, the product preference information, the first decision preference information of the terminal user for the target product, and the second decision preference information of the terminal user for another product.

[0179] Optionally, a plurality of marketing prompt words can be generated based on the product information and various types of reference knowledge. Specifically, any marketing prompt word can include product information and any type of reference knowledge. The types of reference knowledge included in different marketing prompt words can be different. For example, the first prompt word can be generated based on the product information and the booth information, the second prompt word can be generated based on the product information and the group information, and so on.

[0180] In the embodiments of the present specification, a plurality of marketing prompt words can be input into a large language model respectively, so as to obtain a plurality of groups of marketing information corresponding to the plurality of marketing prompt words generated by the large language model. One marketing prompt word can correspond to one group of marketing information. One group of marketing information can include one marketing information, or can include a plurality of marketing information.

[0181] In the embodiments of the present specification, different reference knowledge can be used to generate different marketing prompt words, so that the large language model can generate marketing prompt words based on different reference knowledge. The diversity of the marketing information is improved, and the user experience is improved.

[0182] In actual application, the reference knowledge preferred by the terminal user in the marketing prompt word can be determined according to the marketing information preferred by the terminal user, so that the reference knowledge preferred by the terminal user is expanded to obtain an optimized prompt word, and the optimized marketing information can be generated based on the optimized prompt word subsequently. Optionally, the generating of the marketing information can further include:

[0183] The plurality of groups of marketing information are sent to a plurality of user terminals. One group of marketing information is sent to one user terminal.

[0184] Feedback information of the terminal user for the plurality of groups of marketing information is obtained.

[0185] According to the feedback information, a group of marketing information preferred by the terminal user is selected from the plurality of groups of marketing information.

[0186] A marketing prompt word used for generating the group of marketing information is determined as an optimal prompt word.

[0187] The target type knowledge contained in the optimal prompt word is expanded to obtain an optimized prompt word. The optimized prompt word is used for generating marketing information subsequently.

[0188] Optionally, one kind of marketing information can be sent to each user terminal. Different kinds of marketing information correspond to different types of reference knowledge in the marketing prompt word.

[0189] In actual application, the terminal user can feed back the marketing information, such as interacting with the marketing information, evaluating, and the like. Specifically, the terminal user can click the marketing information, browse the marketing information, access the target product corresponding to the marketing information, evaluate that the terminal user is interested in the marketing information, evaluate that the terminal user is not interested in the marketing information, and the like. Optionally, the feedback information of the terminal user for the plurality of groups of marketing information can include at least one of interaction information and evaluation information.

[0190] In the embodiments of the present specification, the group of marketing information preferred by the terminal user can be selected according to the feedback information of the terminal user. For example, a group of marketing information with a click rate greater than or equal to a preset threshold, a group of marketing information evaluated as interesting by the terminal user, and the like.

[0191] In the embodiments of the present specification, the marketing prompt word of the group of marketing information preferred by the terminal user can be determined as an optimal prompt word. Therefore, the target type knowledge contained in the optimal prompt word can be expanded.

[0192] Since different marketing information groups are generated based on marketing prompt words containing different reference knowledge, the feedback information of the end user can represent the feedback information of the reference knowledge in the marketing prompt word. For example, the feedback information of the end user for a certain group of marketing information is information indicating a preference, which can indicate that the end user prefers the reference knowledge in the marketing prompt word corresponding to the group of marketing information. Therefore, the reference knowledge in the marketing prompt word corresponding to the group of marketing information preferred by the end user can be processed for knowledge expansion to obtain an optimized prompt word. The optimized prompt word can be used to generate optimized marketing information, thereby improving the click rate and conversion rate of marketing.

[0193] Optionally, the target type knowledge contained in the preferred prompt word can be processed for knowledge expansion, which can be to add reference knowledge of the target type. For example, the reference knowledge of the target type is reference knowledge of group information, and reference knowledge of group information can be added. For example, the group information included in the preferred prompt word is information reflecting the age characteristics of the user group, and information reflecting the occupation characteristics of the user group, information reflecting the geographical area characteristics of the user group, etc. can be added.

[0194] Optionally, the preferred prompt word can be processed for knowledge expansion to obtain one optimized prompt word, or multiple optimized prompt words.

[0195] As a specific embodiment, marketing prompt words can also be generated according to market changes, so that marketing information generated by a large language model for market changes can be obtained, thereby improving the click rate and conversion rate of marketing information.

[0196] For example, the market change can be holiday information, and task information for generating marketing information for holiday information can be added to the marketing prompt word. For example, the task information can be "the day after tomorrow is the May Day holiday, please generate marketing information with the atmosphere of the May Day holiday." For example, the market change can be a market trend change of the target product, and task information for generating marketing information for the market trend change of the target product can be added to the marketing prompt word. For example, the task information can be "the market trend of the target product is currently relatively low, please generate marketing information that can boost the market trend of the target product."

[0197] Figure 3 is a flowchart of a method for generating marketing information provided by an embodiment of the present specification. As shown in Figure 3 the method for generating marketing information can include the following steps.

[0198] Step 302: An operator proposes operation demand information.

[0199] In practical applications, the operation personnel can propose operation demand information to the intelligent agent. For example, a specific operation demand information can be "please create marketing information for product A on the designated APP's opening page, adapt to the job user group, and please create marketing information with high click rate."

[0200] Step 304: The intelligent agent generates marketing information based on the operation demand information of the operation personnel.

[0201] In the embodiments of the present specification, the marketing information can be generated based on the intelligent agent. Specifically, the intelligent agent can include at least one of a working mode module, an intent recognition module, an RAG module, a knowledge base module, and a tool library module. The following is a specific introduction to the above modules.

[0202] Optionally, the working mode module can refer to a module that defines the working mode of the intelligent agent. The intelligent agent can generate marketing information based on the defined working mode. For example, the working mode of the intelligent agent can be defined as at least one of CoT and REAct to generate marketing information according to the marketing information generation request, so as to ensure the generation of high-quality, diversified, and personalized marketing information.

[0203] For the CoT working mode, the working mode can decompose complex tasks into multiple logical reasoning steps, and then gradually deduce the final result. Specifically, the intelligent agent can first identify the core demand of the user input and convert it into a series of sub-problems; then, for each sub-problem, the intelligent agent generates intermediate reasoning results in turn, and finally integrates them into complete marketing information output. This working mode can significantly improve the logicality and coherence of marketing information generation. In addition, various sub-problems and reasoning paths can be designed to increase the diversity of marketing information.

[0204] For the REAct working mode, the working mode can perceive environmental changes, quickly respond and take corresponding actions. Specifically, the intelligent agent receives user input, analyzes the current context environment, and determines whether to call external resources. Secondly, according to the judgment result, a suitable action path is selected, such as directly generating marketing information, querying related data, or referring to historical cases. Finally, based on the selected path, marketing information that meets the user's demand is generated.

[0205] In practical applications, CoT and REAct can be flexibly combined to ensure the quality of marketing information, while ensuring the diversity of marketing information, thereby improving the user experience.

[0206] Optionally, the intent recognition module can refer to a module for recognizing user intent by the agent. Specifically, the agent can use natural language processing tools such as large language models to analyze the user's operation demand information, extract key and valuable information from it, capture the user's implied intent, and lay the foundation for subsequent generation of high-quality and personalized marketing information.

[0207] Optionally, the RAG module can refer to a module for retrieving knowledge by the agent based on the RAG technology. The agent can use a vector model to extract relevant knowledge from the knowledge base based on the RAG module, and integrate the extracted knowledge into the prompt words for generating marketing information. This can help the large language model that generates marketing information better understand the context of marketing information generation, improve the factual accuracy of generated marketing information, and avoid the problem of model hallucination.

[0208] For example, for the user's operation demand information "Please create marketing information for product A on the opening page of the specified APP, adapt to the job user group, and please create marketing information with high click rate.", the target product is identified as "A" by intent recognition, the information display is "the opening page of the APP", and the user group is "job user group". Then, the internal knowledge base and RAG technology can be used to obtain the function information of product A, the benefit information of product A, the display requirement information of the opening page of the APP, the display preference information of the opening page of the APP, and the group information of the job user group, etc. Thus, more effective context background is provided for marketing information generation.

[0209] For the knowledge base module, it can refer to a module used by the agent to retrieve knowledge for providing to the large language model. The knowledge base can store product information, display information, reference case information, group information, etc.

[0210] In practical applications, with the acceleration of market environment and technological progress, innovation and development in various business fields are also accelerating. New products and services are emerging in an endless stream, which makes even the latest pre-trained models may lag behind the rapid development of products and services in practical applications. The traditional pre-training method is difficult to capture and reflect the rapidly changing information in practical applications in a timely manner. Therefore, product information and other information can be updated to the knowledge base, and knowledge can be obtained from the knowledge base to provide to the large language model to generate marketing information, which can quickly adapt to market changes and technological progress, and maintain the freshness and accuracy of marketing information.

[0211] Optionally, the marketing information generated by the agent can be one piece of marketing information or multiple pieces of marketing information, which is not limited here.

[0212] Optionally, the knowledge base can also store product preference information of the user and high-quality historical marketing information fed back by the user. Thus, the product preference information and the historical marketing information can be provided to the large language model to stimulate the creativity of the large language model, so that the large language model generates rich and diverse marketing information.

[0213] Optionally, the knowledge base can also store review information of the marketing information and feedback information of the user on the marketing information. The above information can be stored through a structured data table, such as storing the review information through a marketing information review table. The review information can include review results, reasons for not passing the review, modification suggestions, and the like. The feedback information can be stored through a marketing information feedback table, such as click information, browsing information, evaluation information, and the like of the user. Specifically, the product preference information of the user can be obtained based on the marketing information feedback table.

[0214] For the tool library module, an agent for quality control of the marketing information can be included. The tool library module can include a word count control tool, a content rewriting tool, and the like, so that the marketing information can be adjusted through the tool library module to ensure the quality of the marketing information.

[0215] Step 306: The reviewer reviews the marketing information.

[0216] In the embodiments of the present specification, the reviewer can review the marketing information, such as rewriting, offline, and the like of some marketing information.

[0217] For example, the reviewer can also update the high-quality marketing information to the knowledge base as reference knowledge for subsequent generation of marketing information.

[0218] Step 308: The marketing information is online.

[0219] In the embodiments of the present specification, the marketing information can be pushed to the user terminal of the job user group, and then the marketing information is displayed on the user terminal of the job user group.

[0220] Step 310: Feedback information of the terminal user is obtained.

[0221] In the embodiments of the present specification, the feedback information of the job user group can be obtained, such as click information, browsing information, access information, and the like on the marketing information.

[0222] Further, the high-quality marketing information and the low-quality marketing information can be determined through the feedback information.

[0223] As a specific implementation, the high-quality marketing information can be updated to the knowledge base as reference knowledge for subsequent generation of marketing information.

[0224] In the embodiments of the present specification, the product preference information of the professional user group can also be determined according to the feedback information of the professional user group, so that the product preference information of the professional user group can be updated to the knowledge base as reference knowledge for subsequent generation of marketing information for the professional user group.

[0225] Step 312: eliminating low-quality marketing information.

[0226] Specifically, low-quality marketing information can be eliminated. Or other marketing information can be used to replace low-quality marketing information.

[0227] The embodiments of the present specification realize the automation and intelligentization of the entire process from the operation demand information to the marketing information output through the synergistic effect of each module of the intelligent agent, greatly improving the generation efficiency and generation quality of the marketing information. In addition, by introducing high-quality and multi-style historical marketing information and product preference information of users, the generated marketing information is controllable and diversified, which can meet the needs of different user groups.

[0228] Based on the same idea, the embodiments of the present specification also provide a device corresponding to the above method.

[0229] Figure 4 is a structural schematic diagram of a device for generating marketing information provided by the embodiments of the present specification. As shown in Figure 2 , the device can include: Figure 4

[0230] The request input module 402 is configured to input a marketing information generation request for a target product into an intelligent agent for generating marketing information.

[0231] The marketing information generation module 404 is configured to generate marketing information based on the marketing information generation request by the intelligent agent; the marketing information is obtained by inputting a marketing prompt word containing product information of the target product into a large language model; the product information is information about the target product obtained from a knowledge base; and the product information at least includes at least one of function information of the target product or benefit information available by accessing the target product.

[0232] Optionally, the marketing information generation module 404 specifically includes:

[0233] The knowledge acquisition unit is configured to acquire product information of the target product from the knowledge base in response to the marketing information generation request; and the product information at least includes at least one of function information of the target product or benefit information available by accessing the target product.

[0234] The prompt word generation unit is configured to generate a marketing prompt word based on the product information.​

[0235] The marketing information generation unit is configured to input the marketing prompt word into a large language model to obtain marketing information generated by the large language model.

[0236] Optionally, the marketing information generation module 404 can further include:

[0237] The booth information acquisition unit is configured to acquire booth information for describing the specified information booth from the knowledge base in response to the marketing information generation request; the booth information specifically includes at least one of booth requirement information and booth preference information; the booth requirement information is used to indicate requirements for information displayed on the information booth; and the booth preference information is used to reflect user preferences for historical marketing information displayed on the information booth.

[0238] The prompt word generation unit can be specifically configured to:

[0239] Generate a marketing prompt word based on the product information and the booth information.

[0240] Optionally, the marketing information generation module 404 can further include:

[0241] The first reference case information acquisition unit is configured to acquire first reference case information corresponding to at least one of the target product and the specified information booth from the knowledge base; the first reference case information is user-preferred marketing information selected from historical marketing information corresponding to at least one of the target product and the specified information booth;

[0242] The prompt word generation unit can be specifically configured to:

[0243] Generate a marketing prompt word based on the product information, the booth information, and the first reference case information.

[0244] Optionally, the marketing information generation module 404 can further include:

[0245] The group information acquisition unit is configured to acquire group information for describing the specified user group from the knowledge base in response to the marketing information generation request; the group information specifically includes at least one of group feature information and group preference information; the group feature information is used to reflect basic attribute features of the specified user group; and the group preference information is used to reflect preferences of the specified user group for historical marketing information.

[0246] The prompt word generation unit can be specifically configured to:

[0247] Generate a marketing prompt word based on the product information and the group information.

[0248] Optionally, the apparatus can further include:

[0249] a product preference information determination module configured to determine product preference information of the target product for the specified user group according to the product information and the group preference information.

[0250] The prompt word generation unit can be specifically configured to:

[0251] generate a marketing prompt word based on the product information, the group information, and the product preference information.

[0252] Optionally, the apparatus can further include:

[0253] a quality evaluation module configured to call a preset quality evaluation tool to perform quality evaluation on the marketing information to obtain a quality evaluation result.

[0254] a rewriting module configured to rewrite the marketing information to obtain rewritten marketing information if the quality evaluation result indicates that the marketing information does not meet a preset quality condition.

[0255] Optionally, the apparatus can further include:

[0256] a marking operation acquisition module configured to acquire a marking operation of an audit personnel for specified marketing information in the marketing information. The marking operation is used to indicate that the specified marketing information meets a condition of serving as a reference case of marketing information.

[0257] a marketing information updating module configured to update the specified marketing information to the knowledge base in response to the marking operation, and the specified marketing information is used as reference knowledge when generating subsequent marketing information.

[0258] Optionally, the apparatus can further include:

[0259] a marketing information sending module configured to send the plurality of marketing information to a plurality of user terminals, and send at least one of the plurality of marketing information to one user terminal.

[0260] a first feedback information acquisition module configured to acquire first feedback information of a terminal user for the plurality of marketing information. The first feedback information includes at least one of interaction information and evaluation information.

[0261] a second reference case information screening module configured to screen second reference case information preferred by the terminal user from the plurality of marketing information according to the first feedback information.

[0262] A second reference case information updating module is configured to update the second reference case information to the knowledge base; the second reference case information is used as reference knowledge when generating marketing information subsequently.

[0263] Optionally, the apparatus can further include:

[0264] A low-quality marketing information determining module is configured to determine, according to the first feedback information, low-quality marketing information in the plurality of marketing information that does not meet the preferences of the terminal users.

[0265] A low-quality marketing information offline module is configured to offline the low-quality marketing information.

[0266] Optionally, the apparatus can further include:

[0267] A category number determining module is configured to determine whether the number of categories of marketing information for the target product pushed to the terminal users is greater than or equal to a preset number threshold, to obtain a number determination result.

[0268] A supplementary marketing prompt word generating module is configured to generate a supplementary marketing prompt word based on the product information if the number determination result is no.

[0269] A large language model calling module is configured to call the large language model by using the supplementary marketing prompt word, to obtain supplementary marketing information generated by the large language model.

[0270] A supplementary marketing information pushing module is configured to push the supplementary marketing information to at least part of the terminal users.

[0271] Optionally, the apparatus can further include:

[0272] A first decision preference information determining module is configured to determine, based on the first feedback information, first decision preference information of the terminal users for the target product; the first decision preference information is used to reflect the reasons for the preferences of the terminal users for the target product.

[0273] A first mapping relationship information establishing module is configured to establish first mapping relationship information between a product identifier of the target product and the first decision preference information.

[0274] A first mapping relationship information updating module is configured to update the first mapping relationship information to the knowledge base; the first mapping relationship information is used as reference knowledge when generating marketing information for the target product subsequently.

[0275] Optionally, the apparatus can further include:

[0276] The second mapping relationship information establishing module is configured to establish second mapping relationship information between a second user group and at least part of the first decision preference information according to the first decision preference information of each of the terminal users for the target product. The second mapping relationship information is used to reflect the group preference of part of the terminal users included in the second user group.

[0277] The second mapping relationship information is updated to the knowledge base. The second mapping relationship information is used as reference knowledge when subsequent marketing information is generated and displayed to the first user group.

[0278] Optionally, the apparatus can further include:

[0279] The second feedback information obtaining module is configured to obtain second feedback information of the terminal user corresponding to another product. The second feedback information includes at least one of interaction information and evaluation information of the terminal user for marketing information corresponding to the another product.

[0280] The second decision preference information determining module is configured to determine second decision preference information of the terminal user for the another product according to the second feedback information. The second decision preference information is used to reflect the preference reason of the terminal user for the another product.

[0281] The third mapping relationship information establishing module is configured to establish third mapping relationship information between a second user group and at least part of the first decision preference information and the second decision preference information according to the first decision preference information of each of the terminal users for the target product and the second decision preference information for the another product. The third mapping relationship information is used to reflect the group preference of part of the terminal users included in the second user group.

[0282] The third mapping relationship information updating module is configured to update the third mapping relationship information to the knowledge base. The third mapping relationship information is used as reference knowledge when subsequent marketing information is generated and displayed to the second user group.

[0283] Optionally, the marketing information generating module 404 can further include:

[0284] The reference knowledge obtaining unit is configured to obtain at least one type of reference knowledge used to generate the marketing information for the target product from the knowledge base in response to the marketing information generation request.

[0285] The prompt word generating unit can be specifically configured to:

[0286] generate a plurality of marketing prompt words based on the product information and at least part of the at least one type of reference knowledge; a first prompt word in the plurality of marketing prompt words is generated based on the product information and a first part of the at least one type of reference knowledge; a second prompt word in the plurality of marketing prompt words is generated based on the product information and a second part of the at least one type of reference knowledge; the first part of the reference knowledge and the second part of the reference knowledge do not completely contain the same type of reference knowledge.

[0287] The marketing information generation unit can be specifically configured to:

[0288] input the plurality of marketing prompt words into a large language model respectively to obtain a plurality of groups of marketing information generated by the large language model; one marketing prompt word corresponds to one group of marketing information.

[0289] Optionally, the apparatus can further include:

[0290] a marketing information sending module configured to send the plurality of groups of marketing information to a plurality of user terminals; and send one type of marketing information in one group of marketing information to one user terminal.

[0291] a feedback information obtaining module configured to obtain feedback information of a terminal user with respect to the plurality of groups of marketing information.

[0292] a marketing information group screening module configured to screen a group of marketing information preferred by the terminal user from the plurality of groups of marketing information according to the feedback information.

[0293] a preferred prompt word determining module configured to determine a marketing prompt word used to generate the group of marketing information as a preferred prompt word.

[0294] a knowledge expansion processing module configured to perform knowledge expansion processing on a target type of knowledge contained in the preferred prompt word to obtain an optimized prompt word; and the optimized prompt word is used to generate marketing information in subsequent generation.

[0295] Based on the same idea, the present specification also provides a device corresponding to the above method.

[0296] Figure 5 is a structural schematic diagram of a device for generating marketing information provided by the present specification. As shown in Figure 2 , the device 500 can include: Figure 5

[0297] at least one processor 510; and

[0298] a memory 530 in communication connection with the at least one processor; wherein

[0299] ​The memory 530 stores instructions 520 executable by the at least one processor 510, and the instructions are executed by the at least one processor 510 to enable the at least one processor 510 to implement the method for generating marketing information described above.

[0300] Based on the same idea, the embodiments of the present specification also provide a computer readable medium corresponding to the above method. The computer readable medium stores computer readable instructions, and the computer readable instructions can be executed by a processor to implement the method for generating marketing information described above.

[0301] Each of the embodiments in the present specification is described in a progressive manner, and the same or similar parts between each of the embodiments can be referred to each other. Each of the embodiments focuses on the difference from other embodiments. In particular, for the device and equipment embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiments. The device and equipment provided by the embodiments of the present specification are corresponding to the method, and therefore the device and equipment also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding device and equipment will not be described here.

[0302] In the 1990s, it was relatively easy to distinguish whether an improvement in a technology was a hardware improvement (e.g., an improvement in the circuit structure of a diode, transistor, switch, etc.) or a software improvement (an improvement in a method flow). However, as technology has evolved, many improvements in method flows today can be considered as direct improvements in hardware circuit structures. Designers almost always obtain the corresponding hardware circuit structures by programming the improved method flows into hardware circuits. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using hardware entity modules. For example, a programmable logic device (PLD) (e.g., a field programmable gate array (FPGA)) is an integrated circuit whose logic function is determined by user programming of the device. A designer programs a digital system "integrated" on a PLD by himself, without having to ask a chip manufacturer to design and manufacture a special integrated circuit chip. Moreover, instead of manually manufacturing integrated circuit chips, this programming is now mostly implemented using "logic compiler" software, which is similar to the software compiler used when developing programs, and the original code before compilation must also be written in a specific programming language, which is called a hardware description language (HDL), and there are many types of HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc., and the most commonly used are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also be aware that it is easy to obtain a hardware circuit that implements a logical method flow by simply logically programming the method flow in the above-mentioned hardware description languages and programming it into an integrated circuit.

[0303] The controller can be implemented in any suitable way, for example, the controller can take the form of a microprocessor or processor and a computer readable medium storing computer readable program code, such as software or firmware, executable by the (micro)processor, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller and an embedded microcontroller, examples of which include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320, the memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that, in addition to being implemented in pure computer readable program code, the controller can also be implemented to perform the same functions in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers and embedded microcontrollers, etc. by logically programming the method steps. Therefore, such a controller can be considered as a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can even be considered as both a software module implementing a method and a structure within a hardware component.

[0304] The systems, apparatuses, modules or units illustrated by the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0305] For the sake of description, the above apparatuses are described in various units by functions respectively. Of course, the functions of each unit can be implemented in the same or multiple software and / or hardware in the implementation of the present application.

[0306] Those skilled in the art will understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0307] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof.

[0308] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof.

[0309] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof.

[0310] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0311] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) and / or cache memory, non-volatile memory, such as read-only memory (ROM), EPROM, and / or flash memory. The memory is an example of computer-readable media.

[0312] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0313] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or apparatus that includes a list of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0314] The present application can be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform particular tasks or implement particular abstract data types. The present application can also be practiced in distributed computing environments where tasks are performed by remote processing devices that are connected through a communication network. In a distributed computing environment, program modules can be located in both local and remote computer storage media including storage devices.

[0315] The above only describes the embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A method for generating marketing messages, comprising: Input the marketing information generation request for the target product into the intelligent agent used to generate the marketing information; The intelligent agent generates marketing information based on the marketing information generation request; The marketing information is obtained by inputting marketing prompts containing product information of the target product into a large language model; the product information is information about the target product obtained from a knowledge base; The product information includes at least one of the following: functional information of the target product or rights information obtainable by accessing the target product.

2. The method as described in claim 1, wherein the marketing information generation request is specifically used to request the generation of marketing information for display on a designated information display space; the intelligent agent generates marketing information based on the marketing information generation request, specifically including: In response to the marketing information generation request, booth information describing the specified information booth is obtained from the knowledge base; the booth information specifically includes at least one of booth requirement information and booth preference information; wherein, the booth requirement information is used to indicate the requirements for the information displayed in the information booth; the booth preference information is used to reflect the user's preference for the historical marketing information displayed in the information booth; Based on the product information and the booth information, marketing prompts are generated.

3. The method as described in claim 2, wherein the intelligent agent generates marketing information based on the marketing information generation request, specifically including: Obtain first reference case information corresponding to at least one of the target product and the designated information booth from the knowledge base; The first reference case information is marketing information that meets user preferences, selected from historical marketing information corresponding to at least one of the target product and the designated information booth. Based on the product information, the booth information, and the first reference case information, marketing prompts are generated.

4. The method as described in claim 1, wherein the marketing information generation request is specifically used to request the generation of marketing information to be displayed to a specified user group; the intelligent agent generates marketing information based on the marketing information generation request, specifically including: In response to the marketing information generation request, group information describing the specified user group is obtained from the knowledge base; The group information specifically includes at least one of group characteristic information and group preference information; wherein, the group characteristic information is used to reflect the basic attribute characteristics of the specified user group; and the group preference information is used to reflect the preferences of the specified user group for historical marketing information. Based on the product information and the group information, marketing prompts are generated.

5. The method of claim 4, further comprising: Based on the product information and the group preference information, determine the product preference information of the specified user group for the target product; Based on the product information, the group information, and the product preference information, marketing prompts are generated.

6. The method of claim 1, further comprising: The intelligent agent invokes a preset quality assessment tool to perform a quality assessment on the marketing information and obtains the quality assessment result. If the quality assessment result indicates that the marketing information does not meet the preset quality conditions, the marketing information is rewritten to obtain the rewritten marketing information.

7. The method of claim 1, further comprising, after generating marketing information: Obtain the marking operations performed by the reviewers on specific marketing information within the aforementioned marketing information; The marking operation is used to indicate that the specified marketing information meets the conditions of being used as a reference case for marketing information; In response to the tagging operation, the specified marketing information is updated to the knowledge base; the specified marketing information is used as reference knowledge when generating subsequent marketing information.

8. The method as described in claim 1, wherein the marketing information specifically includes multiple types of marketing information; after generating the marketing information, it further includes: Send the various marketing messages to multiple user terminals; Send at least one of the aforementioned marketing messages to a user terminal; Obtain first feedback information from end users regarding the various marketing messages; the first feedback information includes at least one of interaction information and evaluation information. Based on the first feedback information, a second reference case information of the end user's preference is selected from the various marketing information; The second reference case information is updated to the knowledge base; the second reference case information is used as reference knowledge when generating marketing information in the future.

9. The method of claim 8, further comprising, after obtaining the first feedback information from the end user regarding the various marketing information: Based on the first feedback information, identify the low-quality marketing information among the various marketing information that does not meet the preferences of the end user; The aforementioned low-quality marketing information has been removed from the website.

10. The method of claim 9, further comprising, after removing the low-quality marketing information: Determine whether the number of different types of marketing information for the target product pushed to the end user is greater than or equal to a preset quantity threshold, and obtain the quantity determination result; If the quantity determination result is negative, then supplementary marketing prompts are generated based on the product information; The supplementary marketing prompts are used to invoke the large language model to obtain the supplementary marketing information generated by the large language model; The supplementary marketing information is pushed to at least some of the end users.

11. The method of claim 8, further comprising, after obtaining the first feedback information from the end user regarding the multiple marketing messages: Based on the first feedback information, the first decision preference information of the end user for the target product is determined; The first decision preference information is used to reflect the reasons why the end user prefers the target product; Establish a first mapping relationship between the product identifier of the target product and the first decision preference information; The first mapping relationship information is updated to the knowledge base; the first mapping relationship information is used as reference knowledge when generating marketing information for the target product in the future.

12. The method of claim 11, further comprising, after determining the terminal user's first decision preference information for the target product based on the first feedback information: Based on the first decision preference information of each terminal user for the target product, a second mapping relationship is established between the first user group and at least a portion of the first decision preference information; The second mapping relationship information is used to reflect the group preferences of some terminal users included in the first user group; The second mapping relationship information is updated to the knowledge base; the second mapping relationship information is used as reference knowledge when generating marketing information to be displayed to the first user group.

13. The method of claim 11, further comprising: Obtain the second feedback information from the end user that corresponds to another product; The second feedback information includes at least one of the end user's interaction information and evaluation information regarding marketing information for the other product; Based on the second feedback information, the second decision preference information of the end user for the other product is determined; the second decision preference information is used to reflect the reasons for the end user's preference for the other product. Based on the first decision preference information of each end user for the target product and the second decision preference information for the other product, a third mapping relationship is established between the second user group and at least a portion of the decision preference information in the first and second decision information. The third mapping relationship information is used to reflect the group preferences of some terminal users included in the second user group; The third mapping relationship information is updated to the knowledge base; the third mapping relationship information is used as reference knowledge when generating marketing information to be displayed to the second user group.

14. The method as described in claim 1, wherein the intelligent agent generates marketing information based on the marketing information generation request, specifically including: In response to the marketing information generation request, the intelligent agent obtains at least one type of reference knowledge from the knowledge base for generating marketing information for the target product; Based on the product information and at least some of the class knowledge in the at least one class of reference knowledge, generate multiple marketing prompt words; The first prompt word among the plurality of marketing prompt words is generated based on the product information and the first part of the knowledge in the at least one type of reference knowledge; The second prompt word among the plurality of marketing prompt words is generated based on the product information and the second part of the knowledge in the at least one type of reference knowledge; The types of reference knowledge contained in the first part of the knowledge are not exactly the same as those in the second part of the knowledge; The multiple marketing prompts are input into the large language model to obtain multiple sets of marketing information generated by the large language model. One marketing prompt corresponds to a set of marketing messages.

15. The method of claim 14, further comprising, after generating the marketing information: Send the multiple sets of marketing information to multiple user terminals; Send one type of marketing message from a set of marketing messages to a user terminal; Obtain feedback information from end users regarding the multiple sets of marketing information; Based on the feedback information, the marketing information group preferred by the end user is selected from the multiple groups of marketing information; The marketing prompt words used to generate the marketing message group are determined as preferred prompt words; The target type knowledge contained in the preferred prompt words is expanded to obtain optimized prompt words; the optimized prompt words are used in subsequent marketing information generation.

16. An apparatus for generating marketing information, comprising: The request input module is used to generate a smart agent that generates marketing information for the target product by taking the request input. A marketing information generation module is used by the intelligent agent to generate marketing information based on the marketing information generation request; The marketing information is obtained by inputting marketing prompts containing product information of the target product into a large language model; the product information is information about the target product obtained from a knowledge base; The product information includes at least one of the following: functional information of the target product or rights information obtainable by accessing the target product.

17. A device for generating marketing information, comprising: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to implement the method for generating marketing information as described in any one of claims 1 to 15.

18. A computer-readable medium having stored thereon computer-readable instructions that can be executed by a processor to implement the method of generating marketing information according to any one of claims 1 to 15.

Images