Intelligent marketing cloud service method and platform based on big data and ai large model

Abstract

The application provides an intelligent marketing cloud service method and platform based on big data and an AI large model, which comprises the following steps: collecting generative text content returned by at least two AI large model platforms for the same target brand keyword, extracting a first text segment and converting it into a semantic representation vector; constructing a cross-platform brand center semantic representation, calculating a cross-platform semantic consistency result; constructing an interaction angle based on direction difference, generating an interaction input text and obtaining a response, extracting a second text segment to calculate an expression stability result; fusing the consistency and stability results, generating a brand comprehensive performance index and a structured brand performance result; and further generating content optimization suggestions and outputting them through a cloud platform. The application quantitatively evaluates the consistency and stability of brand semantic expression of multiple models, realizes objective and interpretable brand marketing performance analysis and optimization, and improves the controllability and reliability of AI generated content in a business scenario.

Description

Technical Field

[0001] This invention belongs to the field of commercial marketing, and in particular relates to a method and platform for intelligent marketing cloud services based on big data and AI big data models. Background Technology

[0002] Whether a brand can be correctly identified on large-scale platforms, whether it can form a consistent semantic image across different platforms, and whether it can be consistently mentioned in various question scenarios are gradually becoming important factors affecting brand competitiveness.

[0003] However, existing technological systems primarily revolve around search engine index data, public opinion texts, or user comments, lacking the ability to systematically analyze the content generated by large-scale models. Even when some solutions begin to collect content output from AI platforms, their analysis methods still focus on keyword frequency or rule matching, failing to depict the overall semantic differences between different platforms or distinguish the relationship between differences in expression style and brand perception. Furthermore, existing technologies generally lack methods for assessing the stability of brand performance under multi-round interaction conditions, making it impossible to determine whether a brand consistently revolves around the same semantic center under different questioning angles. These issues result in a lack of quantifiable, comparable, and locatable analytical foundations for understanding a brand's overall performance within the semantic environment of large-scale AI models, thus hindering the formation of effective content optimization and marketing decision support mechanisms. Summary of the Invention

[0004] The purpose of this invention is to design an intelligent marketing cloud service method and platform based on big data and AI big models, which can generate brand performance indicators that can be directly used for ranking, comparison and problem identification, and further map the indicator results into clear content optimization suggestions.

[0005] To achieve the above objectives, a first aspect of the present invention provides an intelligent marketing cloud service method based on big data and AI big data models, the method comprising: Collect generative text content returned by at least two large AI model platforms for the same target brand keywords; Extract a first text fragment directly related to the target brand keywords from the generative text content; The first text segment is converted into a semantic representation vector using a unified text semantic encoding model; Based on the semantic representation vector, a cross-platform brand-centric semantic representation is constructed, and the directional similarity of the semantic representations of each platform relative to the brand-centric semantic representation is calculated to obtain the cross-platform semantic consistency result. At least one interaction angle is constructed based on the directional differences between the semantic representations of each platform and the semantic representation of the brand center, and corresponding interactive input text is generated for each interaction angle. The interactive input text is input into the at least two AI large model platforms respectively to obtain the response text; Extract a second text fragment directly related to the target brand keywords from the response text; The second text fragment is converted into a semantic representation vector using the text semantic encoding model; Calculate the directional similarity between the semantic representation vector of the second text segment and the semantic representation of the brand center under each interaction angle to obtain the expression stability result; By integrating the cross-platform semantic consistency results and the expression stability results, a comprehensive brand performance index and a structured brand performance result are generated; Based on the structured brand performance results, a pre-set content adjustment template is matched to generate content optimization suggestions for specific interaction angles; The aforementioned content optimization suggestions will be deployed and provided as a service through a cloud platform.

[0006] Furthermore, the extraction of the first or second text fragment includes identifying and removing preset prompt text and standardized response content automatically generated by the AI ​​large model platform, retaining only text fragments containing the target brand keywords or those directly related to their semantics.

[0007] Furthermore, the construction of the brand center semantic representation adopts a weighted aggregation method, and the weight is determined by the number of times the target brand keyword appears in the first text segment and the length of the text segment.

[0008] Furthermore, constructing at least one interaction angle includes: calculating the directional difference vector between the semantic representation of each platform and the semantic representation of the brand center, performing clustering processing on the directional difference vector, and determining the cluster center direction corresponding to each clustering result as an interaction angle.

[0009] Furthermore, the interactive input text is generated through a preset input expression template, which specifies the fixed position of the target brand keywords in the text and configures different semantic emphasis expressions according to the interaction perspective.

[0010] Furthermore, the expression stability result is obtained by performing a double average operation on the directional similarity of all platforms under all interaction angles.

[0011] Furthermore, the overall brand performance index is obtained by weighted fusion of the cross-platform semantic consistency result and the expression stability result, with the weights preset by the system configuration module.

[0012] Furthermore, the generation of the content optimization suggestions includes: identifying interaction angles whose expression stability is lower than a preset threshold based on the structured brand performance results, calling a preset content adjustment template associated with the interaction angle, and substituting the target brand keywords into the template to generate suggested text.

[0013] Furthermore, the content optimization suggestions are output after being sorted according to suggestion weights, and the suggestion weights are jointly determined by the brand's overall performance index and the stability results of the corresponding interaction angle.

[0014] In a second aspect, the present invention provides an intelligent marketing cloud service platform based on big data and AI big data models, the platform comprising: The multi-platform semantic acquisition module is used to collect generative text content returned by at least two AI large model platforms for the same target brand keyword, extract the first text fragment directly related to the target brand keyword, and convert it into a semantic representation vector through a unified text semantic encoding model. The brand consistency modeling module is used to construct a cross-platform brand-centric semantic representation based on the semantic representation vector, and calculate the directional similarity of the semantic representations of each platform relative to the brand-centric semantic representation to obtain cross-platform semantic consistency results. The expression stability verification module is used to construct at least one interaction angle based on the directional difference between the semantic representation of each platform and the semantic representation of the brand center, generate corresponding interactive input text, input the interactive input text to the at least two AI big model platforms to obtain response text, extract the second text fragment directly related to the target brand keyword, convert it into a semantic representation vector through the text semantic encoding model, and calculate its directional similarity with the semantic representation of the brand center to obtain the expression stability result. The brand performance output module is used to integrate the cross-platform semantic consistency results and the expression stability results to generate comprehensive brand performance indicators and structured brand performance results. The suggestion generation and service output module is used to match the pre-set content adjustment template according to the structured brand performance results, generate content optimization suggestions for specific interaction angles, and deploy the content optimization suggestions through the cloud platform to provide service output.

[0015] The beneficial technical effects of the present invention are at least as follows: To address the aforementioned issues, this invention provides an intelligent marketing cloud service method and platform based on big data and AI big model. By transforming content generated from multiple AI platforms into a unified semantic representation structure, a cross-platform brand semantic consistency modeling mechanism is constructed. Furthermore, an expression stability verification process under interactive conditions is introduced, forming a systematic evaluation method for the overall performance of a brand within the big model semantic environment. This method not only structurally characterizes the degree of semantic aggregation of a brand across different platforms but also further verifies the brand's semantic maintenance capabilities by incorporating multi-angle interactive scenarios, thereby achieving a joint evaluation of brand recognition consistency and expression stability. Based on this, the invention structurally integrates the above analysis results to generate brand performance indicators that can be directly used for ranking, comparison, and problem identification. The indicator results are further mapped into explicit content optimization suggestions, which are then provided externally via a cloud platform. Through the above technical solution, this invention realizes a complete technical chain from multi-platform content collection, semantic modeling analysis, interactive verification to optimization suggestion output, providing a systematic method with engineering feasibility for brands to adjust content and make marketing decisions within the AI ​​big model ecosystem. Attached Figure Description

[0016] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0017] Figure 1 This is a flowchart of the intelligent marketing cloud service method based on big data and AI big data models of the present invention.

[0018] Figure 2 This is a framework diagram of the intelligent marketing cloud service platform based on big data and AI big data models of the present invention. Detailed Implementation

[0019] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0020] In one or more embodiments, such as Figure 1 As shown, a method for providing intelligent marketing cloud services based on big data and AI big data models is disclosed, the method including the following: S1: Collect generative text content returned by at least two AI large model platforms for the same target brand keyword; extract a first text fragment directly related to the target brand keyword from the generative text content; convert the first text fragment into a semantic representation vector through a unified text semantic encoding model; Specifically, this step involves acquiring generative text content directly related to the target brand from multiple AI large-scale model platforms and converting this content into a unified semantic representation structure. This eliminates differences in language style, expression length, and response organization across different platforms, providing a directly usable input foundation for subsequent cross-platform modeling. As the initial stage of the entire method, this step's input consists of only two parts: one is a pre-defined set of target brand keywords by the system configuration module, typically composed of the brand name, core product name, or fixed terms highly associated with the brand; the other part is the text-based response content returned by the multiple AI large-scale model platforms for the aforementioned brand keywords.

[0021] In real-world engineering environments, different AI large-scale model platforms typically respond to the same brand keyword through their respective standard interactive interfaces or unified calling channels. Although the interfaces and returned text structures differ across platforms, they share a common thread: the input is a semantic query containing the brand keyword, and the output is generated text in natural language. This step does not attempt to analyze the platform's internal generation logic but rather treats the returned text as an externally observable result. For example, when the target brand is a consumer electronics brand, the content returned by different platforms may focus on product function descriptions, usage scenario descriptions, or brand background information, respectively. These contents differ in wording and information organization, but all revolve around the same brand. This step does not attempt to distinguish which description method is superior but focuses on the brand awareness information that these descriptions collectively point to at the semantic level.

[0022] After collecting text from the platform, the system first restructures the text content, removing generic guiding statements or platform-fixed explanatory content irrelevant to the brand, retaining only sentence fragments directly related to brand keywords. This process does not rely on a complex rule system, but rather revolves around a core objective: ensuring that the text content participating in subsequent semantic modeling accurately reflects the platform's understanding of the brand.

[0023] Subsequently, the processed text content is input into the same text semantic encoding model for representation transformation. This semantic encoding model adopts a fixed structure, consisting of a text embedding layer and a multi-layer self-attention computation structure, used to perform unified semantic modeling on texts of different lengths and expressions, and output fixed-dimensional vector results. Text content from all platforms is processed through the same encoding model, without differentiated configurations for any single platform, to ensure that the results generated by different platforms are in the same semantic space.

[0024] During this process, the system does not update model parameters online, nor does it adjust the encoding strategy based on platform differences. The encoding model is used only as a semantic mapping tool, its function being to transform natural language text into a continuous semantic representation that can be used for subsequent computation. After semantic encoding is completed, each "platform-brand keyword" combination will generate a corresponding semantic representation vector. The system organizes these vectors according to the platform and keyword dimensions to form a unified set of semantic representations. The generated form of this semantic representation is as follows: ; in, Indicates on the platform Above targeting brand keywords The obtained semantic representation results; This indicates that after structural reorganization, it is related to brand keywords. Directly relevant platform text content; The text semantic encoding model refers to a model that performs overall semantic modeling of the input text through an embedding layer and a multi-layer self-attention structure, and outputs a vector representation with fixed dimensions.

[0025] Using the above method, even if the text returned by different platforms differs significantly in expression style and content organization, the resulting semantic representations can still be compared and analyzed within the same semantic space. The output of this step is a set of structured brand semantic representations, where each element exists as a semantic vector and uniquely corresponds to a specific platform and brand keyword combination. This output will be directly used as input data for semantic consistency modeling and expression stability verification in subsequent steps; the original text content will not be reused in later steps.

[0026] S2: Construct a cross-platform brand-centric semantic representation based on the semantic representation vector, and calculate the directional similarity of the semantic representations of each platform relative to the brand-centric semantic representation to obtain cross-platform semantic consistency results; Specifically, based on the unified semantic representation results obtained in Step 1, this step performs consistency modeling of the semantic expression of the same brand keyword across multiple AI large-scale model platforms. The core of this modeling is to distinguish between the "stylistic differences in the text returned by different platforms" and the "differences in the understanding of brand semantics," using a stable and comparable structured result to characterize the degree of semantic aggregation between platforms, while simultaneously providing platform-level deviation information for direct use in the next step of interaction scenario construction and stability verification. The input to this step comes from the output of Step 1, namely the semantic vector corresponding to each "platform-brand keyword" in the semantic representation set. And generate the text fragment corresponding to the vector. .in It refers to the text content directly related to the brand keywords that was retained after the structural organization in step one. Yes The vector result obtained by semantic encoding; platform set with It indicates that brand keywords are based on express.

[0027] In generative content across multiple AI platforms, the same brand is often described from different perspectives on different platforms. For example, the same brand might be described more as a "set of product features" on some platforms, while on others it might be described more as "usage scenarios and target audience profiles." These differences are not necessarily bad in marketing scenarios, but when the differences are so large that the brand appears in unrelated semantic positions across platforms, it directly affects the subsequent judgment of "whether the brand can be stably recognized on different platforms." Therefore, this step treats the semantic vector of each platform as an observation point of the brand's keywords for that platform, constructs a cross-platform central semantic representation, and then uses a robust method to measure the degree of aggregation of each observation point around the center.

[0028] To ensure the central semantic representation more accurately reflects the level of "the platform is indeed talking about the brand," this step incorporates the frequency of brand keywords and the length of the text fragment when constructing the central theme. Intuitively, a higher frequency of occurrence indicates a greater likelihood of the fragment revolving around the brand; a longer fragment suggests the inclusion of general interpretations and extended content from the platform, thus receiving a more subtle weighting. The system selects from the text fragments... Directly count the number of times brand keywords appear in the data. The length is measured by the number of characters or tokens after word segmentation. Both can perform statistics locally, without relying on external data sources. Then, the weights of each platform are normalized and used for the semantic representation of the aggregation center.

[0029] The central semantic representation is calculated in the following form: ; in, Brand keywords Central semantic representation vector in a multi-platform semantic space; Indicates platform Normalized weights when constructing the central representation. The semantic encoding output from step one serves as the basic unit of aggregation; Indicates in text fragment Chinese brand keywords The number of occurrences is counted by... The count obtained after string matching or word segmentation; Represents a text fragment The length of the character can be obtained by counting characters or counting tokens after word segmentation. This represents the set of platforms participating in the analysis. In the formula... This is used to avoid numerical problems when the length is zero and to make the weight change smoothly as the text length increases.

[0030] After obtaining the central semantic representation, this step uses "directional similarity" to characterize the closeness between the platform semantics and the central semantics, and introduces a robust term for generative content to suppress interference with the overall consistency judgment when a few platforms exhibit extreme expressions (such as over-expanded backgrounds or over-generalized descriptions). Specifically, the directional similarity between each platform and the center is first calculated, and then a saturation penalty is applied to the "deviation amount" to ensure that large deviations do not linearly amplify their impact on the overall result. This results in semantic consistency of brand keywords across platforms.

[0031] The semantic consistency result is calculated in the following form: ; in, Brand keywords Cross-platform semantic consistency results; This represents the robustness coefficient, used to adjust the degree of suppression against extreme deviations; its value is given by the system configuration. , , This is consistent with the meaning of the above formula. The first term within the parentheses represents the directional similarity between the platform semantics and the central semantics; the second term is a saturation-type penalty term, which produces a more significant inhibitory effect when the similarity is low, while its influence tends to level off when the similarity is high. This makes the overall consistency result more reflective of the "common semantic position of most platforms," ​​while retaining sensitivity to a few platforms that deviate from the platform without excessively amplifying their impact. At the implementation level, the system calculates... Simultaneously, the deviation of each platform relative to the center is retained, which is used in subsequent steps to directly select the "platform with the more obvious deviation" for stability verification. This deviation can be directly obtained from the directional similarity in the above formula, without the need to introduce additional data types or additional model structures.

[0032] This step outputs two results. The first is the brand keywords. Cross-platform semantic consistency results The second is platform-level deviation set. The index results of the reference relationships in the system enable subsequent steps to locate the corresponding semantic deviation degree according to the platform and organize the verification input accordingly; the index results come from the calculation process of this step. With each The corresponding relationship record.

[0033] S3: Construct at least one interaction angle based on the directional differences between the semantic representations of each platform and the semantic representation of the brand center, and generate corresponding interaction input text for each interaction angle; input the interaction input text into the at least two AI large model platforms respectively to obtain response text; extract a second text fragment directly related to the target brand keyword from the response text; convert the second text fragment into a semantic representation vector through the text semantic encoding model; calculate the directional similarity between the semantic representation vector of the second text fragment and the semantic representation of the brand center under each interaction angle to obtain the expression stability result; Specifically, this step verifies the stability of the semantic structure output from step two under interactive conditions. The verification process uses the brand semantic center formed in step two as a representation. As a semantic anchor, and utilizing a platform-level semantic representation set The system constructs several representative input angles based on the semantic deviation structure reflected in the data. Platform responses are collected from these angles, and semantic alignment calculations are performed to obtain the stability of brand keyword expression under interactive conditions. The system first constructs interactive angles based on the difference direction between the platform-level semantic representation and the central semantics. Specifically, for each platform... The system calculates its semantic representation. Relative to the central semantics The system identifies directional differences and treats each direction as a "potential expression angle." Since different platforms may have similar semantic emphases, the system merges these directions, grouping similar difference vectors into the same category and selecting a representative direction for each category. Each representative direction corresponds to an interpretable decision-making expression angle, such as bias towards functional description, usage scenario, or target audience.

[0034] After merging directions, the system assigns a fixed number to each representative direction and records the correspondence between this number and its corresponding semantic direction features and template type. This number serves as a unique identifier for the interaction angle within the system, used for subsequent interaction input construction and result aggregation. The resulting set of numbers constitutes the interaction angle index set. Each element corresponds to a specific interactive expression perspective.

[0035] Based on the above numbering system, the system calls a preset input expression template to generate interactive input text. The template contains brand keywords. The template is set to appear at a fixed location, and different expression fragments are selected based on the interaction angle number to reflect the semantic emphasis of that angle. The template is a pre-built resource of the system, and its selection result is recorded together with the interaction angle number to ensure that the interaction angle corresponding to each platform response can be accurately traced in the future.

[0036] After each AI platform returns a natural language response to the interactive input text, the system will return the text using the same semantic encoding model as in step one. Convert to semantic vector superscript Corresponding Interaction Angle Index Set A specific number within it. This vector is related to the central semantic representation. Directional similarity is calculated to assess the degree to which the platform retains the brand's semantics from this interaction perspective.

[0037] The stability results are obtained by averaging from both platform and interaction perspectives, and the calculation form is as follows: ; in, Brand keywords The results of expression stability; Represents a collection of platforms; Represents the set of interactive perspective indices The number of elements in; Indicates platform Numbering from the interaction perspective The semantic representation of the returned text is given under the corresponding input conditions. This represents the central semantic representation output from step two. The generation process remains consistent with step one to ensure semantic space consistency. This step outputs two result variables. The first is the brand keyword. Expression stability results The second is the set of interaction angle indexes. In calculation During the process, the system simultaneously retains the ID of each interaction angle. The corresponding intermediate result of angular-level stability is denoted as . .in, Brand keywords Numbering from the interaction perspective The degree of semantic preservation. The aforementioned With the set of interactive angle indexes The numbers in the table correspond one-to-one and serve as direct inputs for generating structured decomposition and optimization suggestions in subsequent steps.

[0038] S4: Integrate the cross-platform semantic consistency results with the expression stability results to generate comprehensive brand performance indicators and structured brand performance results; Specifically, in engineering implementation, and Although both results describe a brand's semantic performance on the AI ​​platform, they originate from different sources and have different emphases. Using them directly side-by-side would increase the complexity of subsequent system interpretation and invocation. Therefore, this step first aligns the two results to allow them to be combined under the same evaluation scale. This alignment is accomplished using fixed mapping rules, which are pre-configured by the system configuration module and do not rely on historical sample statistics or online updates. After mapping, and Both can be regarded as indicators that "the larger the value, the more stable the performance," thus laying the foundation for subsequent integration.

[0039] After scaling, the system performs metric fusion. The fusion process is not a simple concatenation; rather, it uses a configurable weighting coefficient to weight and combine consistency and stability results. This weighting coefficient, set during system initialization, reflects the relative importance of "inter-platform consistency" and "interaction condition stability" in specific business scenarios. For example, in scenarios emphasizing the overall brand image consistency, the weight of consistency results can be increased; in scenarios emphasizing actual user experience, the weight of stability results can be increased. The fused result serves as a comprehensive performance indicator for brand keywords, used for subsequent ranking, filtering, or threshold determination.

[0040] The calculation method for the overall brand performance index is as follows: ; in, Brand keywords Comprehensive performance indicators; This refers to the cross-platform semantic consistency result output in step two. The output of step three is the expression stability result; The weighting coefficients, configured by the system, are used to adjust the proportion of the two types of results in the comprehensive index. This is used to obtain the comprehensive performance index. Subsequently, the system further utilizes the interactive perspective index set. The stability results are decomposed into a structured form. Specifically, the system calculates in step three... At that time, the number of each interaction angle had already been recorded internally. The corresponding intermediate alignment result, which is consistent with The angle numbers in the records correspond one-to-one. This step reorganizes the stability results by angle by reading these records, forming a mapping structure of "interaction angle - stability value". This structure can clearly indicate the differences in the degree of brand semantic retention under different decision-making angles.

[0041] Subsequently, the system will integrate performance indicators. Original consistency results Overall stability results and according to The decomposed angle-level stability maps are collectively encapsulated into a unified data structure. This data structure exists in the system with fixed fields, and the relationships between these fields are explicitly defined in this step, ensuring that subsequent modules can directly call it without needing to re-understand the preceding calculation logic. This step outputs two result variables. The first is the brand keyword. Comprehensive performance indicators The first metric reflects the brand's overall semantic performance level on the AI ​​big data model platform; the second is the structured result of brand performance. The result was obtained by , , and based on The resulting interactive perspective-level stability mapping is used to generate specific optimization suggestions and platform display content in subsequent steps.

[0042] S5: Match the pre-set content adjustment template based on the structured brand performance results to generate content optimization suggestions for specific interaction angles; deploy the content optimization suggestions through the cloud platform and provide service output; Specifically, the input includes two items: one is brand keywords. Comprehensive performance indicators This metric reflects the brand's overall semantic performance across multiple AI large-scale model platforms; secondly, it represents the structured results of brand performance. This includes cross-platform semantic consistency results. Overall expression stability results and an index set based on interaction perspective The resulting angle-level stability mapping relationship. All inputs above are structured numerical or indexed data, no longer involving raw text or semantic vectors. In engineering implementation, the generation of content optimization suggestions depends on... A layer-by-layer analysis of the internal structure. The system first starts from... Read the interactive perspective index set In steps three and four, the set has already determined the correspondence between each interaction angle number and its semantic emphasis type, such as bias towards functional description, bias towards usage scenario, or bias towards applicable user group. This step uses this index to break down the stability results to the specific angle level, forming a basic mapping of "interaction angle number - stability value".

[0043] Subsequently, the system combines the aforementioned angle-level stability values ​​with comprehensive performance indicators. Based on this, it determines which interaction angles require priority content adjustments given the current overall brand status. This determination process does not employ a dynamic learning mechanism but is completed through a predefined sequence of rules. Specifically, when... At a lower level, the system tends to output optimization suggestions from multiple perspectives simultaneously; when When the system is at a medium or high level, it prioritizes generating suggestions from a few angles with relatively low stability values ​​to avoid indiscriminately outputting too many suggestions. After determining the interaction angles for which suggestions need to be generated, the system enters the template mapping stage. Each interaction angle number is bound to a set of content adjustment templates in the system. These templates are pre-set during system initialization and stored in configuration files or rule tables. The template content is not an abstract description but explicitly defines the direction of content adjustment. For example, the proportion of product feature descriptions should be increased from one angle, or specific usage scenario descriptions should be added from another angle. The templates themselves do not contain variable parameters; they are only filled with brand keywords. Instantiation with the platform identifier is completed, thereby ensuring the determinism of the generation process.

[0044] To rank and filter multiple candidate suggestions, this step calculates a weight value for each suggestion, reflecting its priority in the current analysis task. The weight calculation considers both the overall brand performance and the degree of stability deviation from the corresponding interaction perspective, and its calculation form is as follows: ; in, Indicates targeting brand keywords Interaction Angle Number Optimization suggestion weights; The overall performance index output from step four; Indicates from The corresponding interactive angle number read from the middle The stability results are calculated in this way. This calculation method ensures that when a brand's overall performance is weak and its stability is low from a certain perspective, the corresponding recommendation will naturally receive higher priority in the ranking.

[0045] After obtaining the weights, the system sorts all candidate suggestions from highest to lowest weight and selects the top few suggestions as the final result based on the configured output quantity threshold. The selected results are then encapsulated into a unified data object, containing fields such as brand keywords, interaction angle number, platform identifier, suggestion type, and specific suggestion text. The field structure is predefined in the system, eliminating the need for different business modules to re-parse the meaning when reading. Cloud platform deployment is completed synchronously in this step. The system generates a unique task identifier for each analysis task and writes the aforementioned set of suggestion objects to the result service module of the cloud platform after associating the task identifier with it. The cloud platform manages and schedules the suggestion results through this task identifier, supporting queries by task, by brand, or by time period. The cloud platform itself does not participate in suggestion calculation; it is only responsible for the storage, access control, and display of results, thus ensuring a clear separation between calculation logic and service logic.

[0046] In one or more embodiments, such as Figure 2 As shown, a smart marketing cloud service platform based on big data and AI big data models is disclosed, the platform including: The multi-platform semantic acquisition module is used to collect generative text content returned by at least two AI large model platforms for the same target brand keyword, extract the first text fragment directly related to the target brand keyword, and convert it into a semantic representation vector through a unified text semantic encoding model. The brand consistency modeling module is used to construct a cross-platform brand-centric semantic representation based on the semantic representation vector, and calculate the directional similarity of the semantic representations of each platform relative to the brand-centric semantic representation to obtain cross-platform semantic consistency results. The expression stability verification module is used to construct at least one interaction angle based on the directional difference between the semantic representation of each platform and the semantic representation of the brand center, generate corresponding interactive input text, input the interactive input text to the at least two AI big model platforms to obtain response text, extract the second text fragment directly related to the target brand keyword, convert it into a semantic representation vector through the text semantic encoding model, and calculate its directional similarity with the semantic representation of the brand center to obtain the expression stability result. The brand performance output module is used to integrate the cross-platform semantic consistency results and the expression stability results to generate comprehensive brand performance indicators and structured brand performance results. The suggestion generation and service output module is used to match the pre-set content adjustment template according to the structured brand performance results, generate content optimization suggestions for specific interaction angles, and deploy the content optimization suggestions through the cloud platform to provide service output.

[0047] It is worth noting that the specific workflow of the intelligent marketing cloud service platform based on big data and AI big model provided in this embodiment of the invention is the same as that of the intelligent marketing cloud service method based on big data and AI big model described in the above embodiments, and will not be repeated here.

[0048] This invention also provides an intelligent marketing cloud service device based on big data and AI big data models, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps described in the above embodiments of the intelligent marketing cloud service method based on big data and AI big data models. Figure 1 The steps S1 to S5 described above; or, when the processor executes the computer program, it implements the functions of each module in the above platform embodiments.

[0049] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the intelligent marketing cloud service device based on big data and AI models.

[0050] The intelligent marketing cloud service device based on big data and AI models can be a desktop computer, laptop, handheld computer, or cloud server, etc. This device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the intelligent marketing cloud service device based on big data and AI models may also include input / output devices, network access devices, buses, etc.

[0051] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the intelligent marketing cloud service device based on big data and AI models, connecting all parts of the device via various interfaces and lines.

[0052] The memory can be used to store the computer programs and / or modules. The processor, by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory, realizes various functions of the intelligent marketing cloud service device based on big data and AI big data models. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating platform, applications required for at least one function, etc.; the data storage area may store data created based on the operation of the air conditioner controller, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0053] The modules integrated into the intelligent marketing cloud service device based on big data and AI models, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0054] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0055] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A smart marketing cloud service method based on big data and AI big models, characterized in that: The method includes: Collect generative text content returned by at least two large AI model platforms for the same target brand keywords; Extract a first text fragment directly related to the target brand keywords from the generative text content; The first text segment is converted into a semantic representation vector using a unified text semantic encoding model; Based on the semantic representation vector, a cross-platform brand-centric semantic representation is constructed, and the directional similarity of the semantic representations of each platform relative to the brand-centric semantic representation is calculated to obtain the cross-platform semantic consistency result. At least one interaction angle is constructed based on the directional differences between the semantic representations of each platform and the semantic representation of the brand center, and corresponding interactive input text is generated for each interaction angle. The interactive input text is input into the at least two AI large model platforms respectively to obtain the response text; Extract a second text fragment directly related to the target brand keywords from the response text; The second text fragment is converted into a semantic representation vector using the text semantic encoding model; Calculate the directional similarity between the semantic representation vector of the second text segment and the semantic representation of the brand center under each interaction angle to obtain the expression stability result; By integrating the cross-platform semantic consistency results and the expression stability results, a comprehensive brand performance index and a structured brand performance result are generated; Based on the structured brand performance results, a pre-set content adjustment template is matched to generate content optimization suggestions for specific interaction angles; The aforementioned content optimization suggestions will be deployed and provided as a service through a cloud platform.

2. The intelligent marketing cloud service method based on big data and AI big data models according to claim 1, characterized in that, The extraction of the first or second text fragment includes identifying and removing preset prompt text and standardized response content automatically generated by the AI ​​large model platform, retaining only text fragments containing the target brand keywords or those directly related to their semantics.

3. The intelligent marketing cloud service method based on big data and AI big data models according to claim 1, characterized in that, The construction of the brand center semantic representation adopts a weighted aggregation method, and the weight is determined by the number of times the target brand keyword appears in the first text segment and the length of the text segment.

4. The intelligent marketing cloud service method based on big data and AI big data models according to claim 1, characterized in that, The construction of at least one interaction angle includes: calculating the directional difference vector of the semantic representation of each platform relative to the semantic representation of the brand center, performing clustering processing on the directional difference vector, and determining the cluster center direction corresponding to each clustering result as an interaction angle.

5. The intelligent marketing cloud service method based on big data and AI big data models according to claim 1, characterized in that, The interactive input text is generated through a preset input expression template, which specifies the fixed position of the target brand keywords in the text and configures different semantic emphasis expressions according to the interaction perspective.

6. The intelligent marketing cloud service method based on big data and AI big data models according to claim 1, characterized in that, The expression stability results were obtained by double averaging the directional similarity of all platforms under all interaction angles.

7. The intelligent marketing cloud service method based on big data and AI big data models according to claim 1, characterized in that, The overall brand performance index is obtained by weighted fusion of the cross-platform semantic consistency result and the expression stability result, with the weights preset by the system configuration module.

8. The intelligent marketing cloud service method based on big data and AI big data models according to claim 1, characterized in that, The generation of the content optimization suggestions includes: identifying interaction angles whose expression stability is lower than a preset threshold based on the structured brand performance results, calling the preset content adjustment template associated with the interaction angle, and substituting the target brand keywords into the template to generate suggested text.

9. The intelligent marketing cloud service method based on big data and AI big data models according to claim 1, characterized in that, The content optimization suggestions are output after being sorted by suggestion weight, and the suggestion weight is determined by the overall brand performance index and the stability results of the corresponding interaction angle.

10. An intelligent marketing cloud service platform based on big data and AI big data models, characterized in that: The platform includes: The multi-platform semantic acquisition module is used to collect generative text content returned by at least two AI large model platforms for the same target brand keyword, extract the first text fragment directly related to the target brand keyword, and convert it into a semantic representation vector through a unified text semantic encoding model. The brand consistency modeling module is used to construct a cross-platform brand-centric semantic representation based on the semantic representation vector, and calculate the directional similarity of the semantic representations of each platform relative to the brand-centric semantic representation to obtain cross-platform semantic consistency results. The expression stability verification module is used to construct at least one interaction angle based on the directional difference between the semantic representation of each platform and the semantic representation of the brand center, generate corresponding interactive input text, input the interactive input text to the at least two AI big model platforms to obtain response text, extract the second text fragment directly related to the target brand keyword, convert it into a semantic representation vector through the text semantic encoding model, and calculate its directional similarity with the semantic representation of the brand center to obtain the expression stability result. The brand performance output module is used to integrate the cross-platform semantic consistency results and the expression stability results to generate comprehensive brand performance indicators and structured brand performance results. The suggestion generation and service output module is used to match the pre-set content adjustment template according to the structured brand performance results, generate content optimization suggestions for specific interaction angles, and deploy the content optimization suggestions through the cloud platform to provide service output.

Images