An intelligent marketing document generation device and method based on multi-source data fusion

Abstract

The application discloses an intelligent marketing document generation device and method based on multi-source data fusion, and the method comprises four processes: user instruction analysis, regular cleaning, model extraction parameter, Neo4j knowledge graph verification, and output of structured parameters; external data acquisition, parameter calling of a Baidu large model to collect data, de-duplication, semantic enhancement, and weighted screening of high-quality data; internal case processing, regular slicing of case storage, filtering, propagation calculation, multi-dimensional scoring, and generation of a CoT reasoning chain report; multi-modal output, format adaptation, data integration by a BART model, and output of a metadata document. The device and the method are coordinated, multi-source fusion, knowledge graph association and multi-modal conversion are realized, high-quality marketing documents suitable for multiple industries are efficiently generated, decision-making is supported, and the market response and output efficiency of enterprises are improved.

Description

Technical Field

[0001] This invention belongs to the field of computer application technology, specifically relating to an intelligent marketing document generation device and method based on multi-source data fusion. Background Technology

[0002] Current mainstream marketing document generation tools are mainly based on discrete data retrieval patterns and fixed template engines, which have significant technical bottlenecks in terms of data integration efficiency, content intelligence, and output format, as specifically manifested as follows:

[0003] I. Overview of Existing Technical Solutions

[0004] Data integration approach: A distributed data storage architecture is adopted, with brand data (such as CRM system), industry information (third-party API platform), and marketing cases (local database) stored in different formats such as JSON / XML / CSV. The communication protocols (RESTful API / SOAP) are not unified, and there is a lack of standardized access interfaces, resulting in a serious data silo problem.

[0005] Generative engine architecture: It relies on a fixed template engine (such as Word macros) or a simple rule engine, and can only generate basic content through preset marketing scenario templates (such as competitor analysis report templates), and cannot achieve dynamic semantic association and knowledge reasoning.

[0006] Application of intelligent technologies: Only basic NLP technology is used for keyword extraction, and no industry knowledge graph has been built; case matching relies on manually preset rules (such as keyword matching for "promotional activities"), and lacks semantic understanding capabilities.

[0007] Output format: Primarily supports plain text output, with a few tools capable of converting to PDF format, but lacks multimodal interactive capabilities such as chart embedding and dynamic data visualization.

[0008] II. Analysis of the shortcomings of existing technologies

[0009] (1) Low data integration efficiency

[0010] Heterogeneous data barriers: Due to differences in data formats and protocols, brand CRM systems (such as Salesforce), social media monitoring tools (Brandwatch), and industry databases (iResearch Consulting) require the development of customized interfaces for integration, resulting in data synchronization cycles generally exceeding T+1 days. This fails to meet the real-time data requirements of market analysis documents (such as sudden public opinion events).

[0011] Lack of real-time performance: Traditional systems cannot achieve real-time data capture and analysis. For example, when a sudden marketing event (such as a hot topic) occurs, document generation still relies on data synchronized overnight, causing the content to lag behind market dynamics.

[0012] (2) Insufficient intelligent content generation

[0013] Lack of knowledge reasoning ability: The system lacks industry knowledge graph support, making it unable to automatically link historical brand activity data with current market trends. For example, when a food brand launches a new product, the system cannot automatically link the "healthy snacks" market trend with the brand's "low-sugar" selling point, resulting in content planning lacking data-driven logic.

[0014] Weak semantic understanding: Traditional cosine similarity algorithms only calculate the angle between word vectors and cannot identify the hyponym relationship between "mobile phone" and "smartphone". The case matching accuracy is less than 62%, resulting in poor relevance of recommended content.

[0015] (3) Output format is limited

[0016] Existing tools only support basic text or static PDF output, which cannot meet the needs of marketing scenarios for multimodal content. For example, they cannot automatically generate documents containing visualization elements such as line charts of dissemination trends and user profile heatmaps; they lack dynamic interactive capabilities in HTML format (such as clicking on charts to drill down for details), making it difficult to adapt to the display needs of online marketing channels.

[0017] III. Necessity of Technological Improvement

[0018] As market competition intensifies, existing technologies can no longer meet the three core needs of marketing document generation:

[0019] Data-driven decision-making: It requires the integration of multi-source and multi-dimensional data from brands, markets, and users to achieve intelligent content generation based on real-time insights;

[0020] Agile response to the market: Document generation cycle needs to be shortened from the traditional 4 hours to within 10 minutes to cope with rapidly changing marketing scenarios;

[0021] Multimodal intelligent output: It needs to support multiple output formats such as PDF / HTML / Word, and integrate rich media content such as dynamic charts and interactive elements.

[0022] Therefore, there is an urgent need for an intelligent marketing document generation solution that can solve the problem of heterogeneous data fusion, improve semantic understanding capabilities, and support multimodal output. Summary of the Invention

[0023] In view of this, the purpose of this invention is to overcome the shortcomings of the prior art, such as low data integration efficiency, insufficient content intelligence, and limited output formats, and to provide an intelligent marketing document generation device and method based on multi-source data fusion. This solution achieves efficient integration of brand data, industry information, and internal cases through standardized data interfaces, enhances semantic understanding and case matching capabilities through knowledge graphs and natural language processing technologies, and enables diversified document format generation based on a multimodal output adapter. This solves the data silo problem in the prior art, shortens the document generation cycle, improves content relevance and display effects, and meets the modern marketing needs of data-driven decision-making, agile market response, and multimodal output.

[0024] To achieve the above objectives, the present invention provides a first aspect of an intelligent marketing document generation device based on multi-source data fusion, comprising:

[0025] User instruction input module: used to receive marketing document generation requests submitted by users;

[0026] NLP parsing engine module: connected to the user command input module, configured to perform natural language processing on the user command, extract structured parameters and verify the validity of the parameters;

[0027] Parameter decomposition module: connected to the NLP parsing engine module, used to decompose the parsed instructions into task parameters corresponding to external data acquisition and internal case invocation;

[0028] External data engine module: connected to the parameter decomposition module, configured to collect and process external industry data based on task parameters to generate semantically enhanced industry information;

[0029] Internal Case Engine Module: Connected to the parameter decomposition module, configured to initiate a case library search based on task parameters, filter matching marketing cases, and output them;

[0030] Multimodal output module: Connected to the external data engine module and the internal case engine module respectively, it is used to integrate industry information and marketing cases to generate and output marketing documents in multiple formats;

[0031] Data layer interface module: Connects to the external data engine module and the internal case engine module, includes a standardized API gateway, supports RESTful / SOAP dual protocol access to heterogeneous data sources, and realizes data interaction;

[0032] Knowledge graph storage unit: connected to the internal case engine module and NLP parsing engine module, using Neo4j graph database to store entity relationships, covering the association form of "brand-industry-case profile", used for semantic association and case retrieval support;

[0033] Core algorithm engine: including

[0034] Hybrid generation engine: Connected to the multimodal output module and parameter decomposition module, it integrates the Wenxin Yiyan API call module and rule engine, supports dynamic adjustment of generation parameters, and is used for intelligent construction of marketing content;

[0035] Case recommendation and evaluation engine: Connected to the internal case engine module, it deploys a TF-IDF vector index library, supports millisecond-level similarity calculation for tens of millions of cases, and provides similarity evaluation for case library retrieval;

[0036] Case library retrieval module: Connected to the internal case engine module and case recommendation and evaluation engine, it selects suitable cases from the case library based on the semantic association of the knowledge graph storage unit and the similarity calculation results of the case recommendation and evaluation engine.

[0037] Furthermore, the NLP parsing engine module employs a natural language processing model to perform word segmentation, entity recognition, and intent understanding on user commands. The natural language processing model includes a pre-trained language model.

[0038] The external data engine module includes a data acquisition unit and a data processing unit. The data acquisition unit is configured to call third-party data sources to obtain industry trend and market dynamic data. The data processing unit is configured to clean, semantically enhance, and score the importance of the acquired data.

[0039] Furthermore, when the internal case engine module initiates a case library retrieval, it invokes the TF-IDF vector index library capability of the case recommendation and evaluation engine, and combines it with the entity relationship association of the knowledge graph storage unit to achieve case filtering.

[0040] Furthermore, the semantic matching of the case library retrieval module adopts the vector similarity calculation method, which compares the user command parameters with the case feature vectors and filters cases whose similarity meets the threshold. The vector similarity calculation is implemented based on the TF-IDF vector index library of the case recommendation and evaluation engine.

[0041] A second aspect of this invention provides a method for generating intelligent marketing documents based on multi-source data fusion, the production method comprising:

[0042] S1. User instruction parsing process: Process the raw instructions to obtain structured parameters;

[0043] S2. External data acquisition process: collecting and optimizing industry data based on structured parameters;

[0044] S3. Internal case processing workflow: searching, scoring, and generating evaluations from the enterprise case library;

[0045] S4. Multimodal output process, which converts the fused data into a target format document.

[0046] Furthermore, the S1 user instruction parsing process includes: S11. Cleaning the original instruction text using regular expressions to remove redundant symbols and unify punctuation format;

[0047] S12. Extract parameters such as brand, industry level, and output format from the model to generate a structured JSON object;

[0048] S13. Verify the legitimacy of the brand-industry level based on the Neo4j knowledge graph and output the verification identifier.

[0049] Furthermore, the S2 external data acquisition process includes:

[0050] S21. Construct a search query based on structured parameters, call Baidu's big data model to obtain industry data, and deduplicate data using a similarity hashing algorithm;

[0051] S22. Use the Wenxin large model to perform TF-IDF cleaning, BERT-CLS semantic enhancement, and summary generation on the data;

[0052] S23. Sort the data using a weighted formula: score = 0.6 × semantic_sim + 0.3 × time_decay + 0.1 × authority, where:

[0053] semantic_sim is the semantic similarity (value 0 to 1), which is calculated based on the model to determine the relevance between the text and the user's needs;

[0054] time_decay is the time decay factor (value 0 to 1), which is determined based on the difference between the data release time and the current time;

[0055] The authority rating is a score for the data source's authority (ranging from 0 to 1), based on the source level settings such as industry white papers (0.7) and internal company reports (0.5); the top N data points are filtered according to the score.

[0056] Furthermore, the S3 internal case processing flow includes:

[0057] S31. Knowledge base case library construction steps: The original cases are sliced ​​into large model segments by a scheduled task, entity relationships are extracted and structured and stored in the Neo4j graph database and Milvus vector database;

[0058] S32. Semantic retrieval steps: Based on user command parameters, perform industry path filtering in Neo4j, and combine the propagation data score (propagation_score = ∑(propagation volume × e^(-0.1 × Δt)) to select candidate cases, where:

[0059] Dissemination volume refers to the dissemination data of the case on Weibo and WeChat platforms;

[0060] Δt represents the time difference between the data release time and the current time, in months.

[0061] S33. Dynamic scoring steps: The candidate cases are comprehensively scored using the formula final_score = 0.6 × semantic_sim + 0.3 × industry_sim + 0.1 × propagation_score, where:

[0062] industry_sim represents the industry matching degree (value 0-1), calculated based on the Jaccard similarity of the industry path;

[0063] semantic_sim is the semantic similarity between the case summary and the user command (values ​​from 0 to 1), calculated using the Sentence-BERT model;

[0064] S34. Explainable evaluation generation steps: Generate a CoT reasoning chain evaluation report based on the scoring dimensions, and summarize the highlights of the case.

[0065] Furthermore, the S4 multimodal output process includes:

[0066] S41. Format adaptation steps: Based on the output format, including PDF / HTML / Word, use Apache POI and pdfkit components to achieve automated conversion from Markdown to the target format;

[0067] S42. Content Integration Steps: Use the BART model to integrate external data with case content, ensuring a logical coherence score > 0.7;

[0068] S43. Metadata generation steps: Output document metadata including data source identifier and generation time.

[0069] The present invention, by adopting the above technical solution, has at least the following beneficial effects:

[0070] This invention efficiently solves the problem of integrating heterogeneous data sources by using standardized data interfaces and multi-source data fusion technology: it uses a unified API gateway to enable access to data sources with different protocols and formats, breaks down data silos, significantly improves data synchronization efficiency, and can obtain market dynamic data in real time to meet the timeliness requirements of marketing scenarios.

[0071] This invention, based on knowledge graph and natural language processing technologies, significantly enhances the intelligence level of marketing document generation: by constructing a network of relationships between brands, industries, and cases, and combining the semantic understanding capabilities of large models, it achieves cross-domain semantic association, accurately matches relevant cases and market trends, and changes the low accuracy problem of traditional keyword matching, making the document content more logical and data-supported.

[0072] This invention employs a modular architecture and parallel processing mechanism to achieve rapid generation of marketing documents: the various functional modules work together, and the entire process from parsing user instructions to document output is processed efficiently, greatly shortening the document generation time and enabling enterprises to quickly respond to market trends and changes, thereby improving marketing efficiency.

[0073] This invention leverages multimodal output technology to meet diverse marketing display needs: it supports conversion of multiple document formats and can embed dynamic charts, interactive elements, etc., breaking through the limitations of traditional plain text or static documents, adapting to multiple online and offline application scenarios, and enhancing the display effect and attractiveness of marketing content.

[0074] This invention features dynamic scalability and flexible adaptability: standardized interfaces and an extensible knowledge graph architecture facilitate the addition of new data sources and adaptation to different industry needs, eliminating the need for large-scale system reconstruction, reducing enterprise system upgrade costs, and improving the versatility of technical solutions.

[0075] This invention provides reliable support for marketing decisions through data credibility assessment and interpretability reasoning: it comprehensively scores the data to ensure data quality and generates clear case recommendation logic, making the content of marketing documents more credible and helping enterprises make accurate decisions based on reliable information. Attached Figure Description

[0076] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0077] Figure 1 This is a schematic diagram of an embodiment of the intelligent marketing document generation device of the present invention;

[0078] Figure 2 This is a flowchart of the intelligent marketing document generation method of the present invention;

[0079] Figure 3 This is a schematic diagram of the three-structure implementation of the intelligent marketing document generation device of the present invention;

[0080] Figure 4 This is a schematic diagram illustrating the construction of the knowledge base case library of this invention. Detailed Implementation

[0081] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0082] In the description of the embodiments of the present invention, unless otherwise expressly specified and limited, the terms "connected" and "linked" should be interpreted broadly. For example, they can refer to electrical connections; direct connections; indirect connections via an intermediate medium; or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.

[0083] Example 1

[0084] like Figure 1 As shown, this embodiment provides an intelligent marketing document generation device based on multi-source data fusion, including:

[0085] User instruction input module: used to receive marketing document generation requests submitted by users;

[0086] NLP parsing engine module: connected to the user command input module, configured to perform natural language processing on the user command, extract structured parameters and verify the validity of the parameters;

[0087] Parameter decomposition module: connected to the NLP parsing engine module, used to decompose the parsed instructions into task parameters corresponding to external data acquisition and internal case invocation;

[0088] External data engine module: connected to the parameter decomposition module, configured to collect and process external industry data based on task parameters to generate semantically enhanced industry information;

[0089] Internal Case Engine Module: Connected to the parameter decomposition module, configured to initiate a case library search based on task parameters, filter matching marketing cases, and output them;

[0090] Multimodal output module: Connected to the external data engine module and the internal case engine module respectively, it is used to integrate industry information and marketing cases to generate and output marketing documents in multiple formats;

[0091] Data layer interface module: Connects to the external data engine module and the internal case engine module, includes a standardized API gateway, supports RESTful / SOAP dual protocol access to heterogeneous data sources, and realizes data interaction;

[0092] Knowledge graph storage unit: connected to the internal case engine module and NLP parsing engine module, using Neo4j graph database to store entity relationships, covering the association form of "brand-industry-case profile", used for semantic association and case retrieval support;

[0093] Core algorithm engine: including

[0094] Hybrid generation engine: Connected to the multimodal output module and parameter decomposition module, it integrates the Wenxin Yiyan API call module and rule engine, supports dynamic adjustment of generation parameters, and is used for intelligent construction of marketing content;

[0095] Case recommendation and evaluation engine: Connected to the internal case engine module, it deploys a TF-IDF vector index library, supports millisecond-level similarity calculation for tens of millions of cases, and provides similarity evaluation for case library retrieval;

[0096] Large Model Scoring Module: Connected to the internal case engine module and case recommendation and evaluation engine, it is configured to call the Wenxin Large Model API to score the semantic matching degree and credibility of candidate cases, and the scoring results are used for case screening and ranking.

[0097] Case library retrieval module: Connected to the internal case engine module and case recommendation and evaluation engine, it selects suitable cases from the case library based on the semantic association of the knowledge graph storage unit and the similarity calculation results of the case recommendation and evaluation engine.

[0098] As one implementation method, the NLP parsing engine module in this embodiment uses a natural language processing model to perform word segmentation, entity recognition, and intent understanding on user commands. The natural language processing model includes a pre-trained language model.

[0099] The external data engine module includes a data acquisition unit and a data processing unit. The data acquisition unit is configured to call third-party data sources to obtain industry trend and market dynamic data. The data processing unit is configured to clean, semantically enhance, and perform importance scoring processing on the acquired data driven by a large model.

[0100] As one implementation method, when the internal case engine module of this embodiment initiates a case library retrieval, it calls the TF-IDF vector index library capability of the case recommendation and evaluation engine, and combines the entity relationship association of the knowledge graph storage unit and the dynamic scoring results of the large model scoring module to achieve case filtering.

[0101] As one implementation method, the semantic matching of the case library retrieval module in this embodiment adopts the vector similarity calculation method, which compares the user command parameters with the case feature vectors, and combines the semantic score (value 0 to 1) output by the large model scoring module to filter cases whose similarity meets the threshold. The vector similarity calculation is implemented based on the TF-IDF vector index library of the case recommendation evaluation engine.

[0102] This invention efficiently solves the problem of integrating heterogeneous data sources by using standardized data interfaces and multi-source data fusion technology: it uses a unified API gateway to enable access to data sources with different protocols and formats, breaks down data silos, significantly improves data synchronization efficiency, and can obtain market dynamic data in real time to meet the timeliness requirements of marketing scenarios.

[0103] This invention, based on knowledge graph and natural language processing technologies, significantly enhances the intelligence level of marketing document generation: by constructing a network of relationships between brands, industries, and cases, and combining the semantic understanding capabilities of a large model with the dynamic scoring mechanism of the large model scoring module, it achieves cross-domain semantic association, accurately matches relevant cases and market trends, and changes the low accuracy problem of traditional keyword matching, making the document content more logical and data-supported.

[0104] Example 2

[0105] like Figure 2 As shown, this embodiment provides a method for generating intelligent marketing documents based on multi-source data fusion. The method uses any of the apparatuses described in Embodiment 1 above, and the production method includes:

[0106] S1. User instruction parsing process: Process the raw instructions to obtain structured parameters;

[0107] S2. External data acquisition process: collecting and optimizing industry data based on structured parameters;

[0108] S3. Internal case processing workflow: searching, scoring, and generating evaluations from the enterprise case library;

[0109] S4. Multimodal output process, which converts the fused data into a target format document.

[0110] As one implementation method, the S1 user instruction parsing process in this embodiment includes: S11. Cleaning the original instruction text using regular expressions to remove redundant symbols and unify punctuation format;

[0111] S12. Extract parameters such as brand, industry level, and output format from the model to generate a structured JSON object;

[0112] S13. Verify the legitimacy of the brand-industry level based on the Neo4j knowledge graph and output the verification identifier.

[0113] As one implementation method, the S2 external data acquisition process in this embodiment includes:

[0114] S21. Construct a search query based on structured parameters, call Baidu's big data model to obtain industry data, and deduplicate data using a similarity hashing algorithm;

[0115] S22. Use the Wenxin large model to perform TF-IDF cleaning, BERT-CLS semantic enhancement, and summary generation on the data;

[0116] S23. Sort the data using a weighted formula:

[0117] score = 0.6 × semantic_sim + 0.3 × time_decay + 0.1 × authority, where:

[0118] semantic_sim is the semantic similarity (value 0 to 1), which is calculated based on the model to determine the relevance between the text and the user's needs;

[0119] time_decay is the time decay factor (value 0 to 1), which is determined based on the difference between the data release time and the current time;

[0120] The authority rating is a score for the data source's authority (ranging from 0 to 1), based on the source level settings such as industry white papers (0.7) and internal company reports (0.5); the top N data points are filtered according to the score.

[0121] like Figure 4 As shown, as one implementation method, the S3 internal case processing flow in this embodiment includes:

[0122] S31. Knowledge base case library construction steps: The original cases are sliced ​​into large model segments by a scheduled task, entity relationships are extracted and structured and stored in the Neo4j graph database and Milvus vector database;

[0123] S32. Semantic retrieval steps: Based on user command parameters, perform industry path filtering in Neo4j, and combine the propagation data score (propagation_score = ∑(propagation volume × e^(-0.1 × Δt)) to select candidate cases, where:

[0124] Dissemination volume refers to the dissemination data of the case on Weibo and WeChat platforms;

[0125] Δt represents the time difference between the data release time and the current time, in months.

[0126] S33. Dynamic scoring steps: The candidate cases are comprehensively scored using the formula final_score = 0.6 × semantic_sim + 0.3 × industry_sim + 0.1 × propagation_score, where:

[0127] industry_sim represents the industry matching degree (value 0-1), calculated based on the Jaccard similarity of the industry path;

[0128] semantic_sim is the semantic similarity between the case summary and the user command (values ​​from 0 to 1), calculated using the Sentence-BERT model;

[0129] S34. Explainable evaluation generation steps: Generate a CoT reasoning chain evaluation report based on the scoring dimensions, and summarize the highlights of the case.

[0130] As one implementation method, the S4 multimodal output process in this embodiment includes:

[0131] S41. Format adaptation steps: Based on the output format, including PDF / HTML / Word, use Apache POI and pdfkit components to achieve automated conversion from Markdown to the target format;

[0132] S42. Content Integration Steps: Use the BART model to integrate external data with case content, ensuring a logical coherence score > 0.7;

[0133] S43. Metadata generation steps: Output document metadata including data source identifier and generation time.

[0134] Implementation Three

[0135] The device in this embodiment adopts a modular layered architecture, such as... Figure 3 The flowchart of the intelligent marketing document generation system is shown below. The names of each module and their core logic are explained as follows:

[0136] I. Module Name and Function

[0137] User instructions: These are the system's "input starting point," representing the user's request to generate marketing documents (such as "generate a case study analysis report on summer promotions for beauty brands").

[0138] NLP parsing engine: Core function: Using Natural Language Processing (NLP) technology, it decomposes the semantics, intent, and parameters of user commands (such as extracting key information such as "beauty brand", "summer promotion", and "case analysis"), which is a key link in "understanding user needs".

[0139] Parameter decomposition: Purpose: To break down the requirements parsed by NLP into external data collection parameters (such as the required industry data dimensions) and internal case call parameters (such as case selection conditions), thereby achieving "requirement structuring".

[0140] External Data Engine: Function: Collect data from external data sources (such as industry databases and news platforms) according to the requirements of parameter breakdown (such as summer promotion trends in the beauty industry and competitor dynamics).

[0141] Baidu Search + Wenxin Optimization: This is a "specific implementation sub-process" of an external data engine.

[0142] First, use Baidu search to crawl public data (such as industry news and public opinion); then use the Wenxin big model (such as ERNIE) to perform semantic enhancement, deduplication, and summarization (such as extracting the core strategies of promotional cases).

[0143] Internal Case Engine: Function: Based on the requirements of parameter breakdown, it filters data from the company's internal case library (such as historical marketing plans and success stories), which is the core of "reusing internal experience".

[0144] Case library retrieval: This is a "specific implementation sub-process" of the internal case engine: using retrieval algorithms (such as vector retrieval and keyword matching) to find content that matches the user's needs from the case library (such as past summer promotion cases of beauty brands).

[0145] Large Model Scoring: Function: Use large models (such as Wenxin, GPT) to score the quality of retrieved internal cases (such as assessing the relevance, effectiveness, and reusability of cases) and filter out high-quality content.

[0146] Multimodal output: This is the system's "output endpoint": It integrates external optimization data and internal high-quality case studies to generate multi-format documents (such as PDF reports and PPT presentations) or visualizations (such as data dashboards and trend charts) to meet the needs of marketing scenarios.

[0147] II. Core Logic Link: Requirement Understanding Link: User Instructions → NLP Parsing Engine → Parameter Decomposition, realizing the transformation of "natural language requirements" into "structured task parameters", enabling the system to "understand what the user wants".

[0148] Data acquisition chain:

[0149] External: Parameter breakdown → External data engine → Baidu search + Wenxin optimization

[0150] Internal: Parameter decomposition → Internal case engine → Case library retrieval → Large model scoring

[0151] Achieve "multi-source data integration": capture real-time industry dynamics (external) and reuse historical enterprise experience (internal) to ensure data comprehensiveness.

[0152] Content generation chain: External data (optimized) + Internal case studies (scored) → Multimodal output

[0153] The core logic is "data fusion + intelligent processing" to generate documents / content that meet marketing scenarios, solving the problems of "isolated data and rigid generation" in traditional solutions.

[0154] Method and steps:

[0155] 1. User command parsing (A→B→C)

[0156] Input information: Original command text: "Please output the brand L'Oréal, the primary industry is consumer goods, the secondary industry is cosmetics, and output as a PDF."

[0157] Processing procedure:

[0158] Text cleaning: Use regular expressions to remove redundant symbols and standardize punctuation.

[0159] Entity recognition: Parameter extraction from the BERT-CRF model:

[0160] json

[0161] {

[0162] "brand":"L'Oréal",

[0163] "industry_level1":"Consumer Goods",

[0164] "industry_level2":"cosmetics",

[0165] "output_format":"pdf"

[0166] }

[0167] Intent verification: Verify the legitimacy of the brand-industry level using the Neo4j knowledge graph.

[0168] Output information:

[0169] Structured parameter objects (including brand, industry level, and output format).

[0170] Validation flag (True / False).

[0171] 2. External Data Source Engine (D→F)

[0172] Input information: Structured parameter object

[0173] Processing procedure:

[0174] Baidu's large-scale model search query: Constructing the query "L'Oréal cosmetics industry trends 2025" and retrieving the top 50 results (SimHash deduplication rate > 92%).

[0175] Wenxin large model optimization:

[0176] Data cleaning: TF-IDF filtering of ad content (retaining core marketing analytics text).

[0177] Semantic enhancement: Generate summaries (BERT-CLS extracts key sentences). Associate brand tags (calculate similarity through knowledge graph embedding vectors).

[0178] Sorting strategy: `pythonscore = 0.6 * semantic_sim + 0.3 * time_decay + 0.1 * authority`, taking the top 3 best results. This formula is used to comprehensively score and rank candidate content in search results or case studies. By weightedly fusing three core dimensions (semantic relevance (semantic_sim), timeliness (time_decay), and authority (time_decay), it ensures that the output meets user needs while also possessing data credibility and real-time relevance. For example:

[0179] Candidate content A (Baidu search results):

[0180] Semantic similarity: 0.85 (Content highly relevant to "L'Oréal Marketing Trends")

[0181] Time decay: 0.6 (Published 6 months ago)

[0182] Authority: 0.7 (Source: Industry White Paper)

[0183] Total score: 0.6 × 0.85 + 0.3 × 0.6 + 0.1 × 0.7 = 0.78

[0184] Candidate Content B (Internal Case Library):

[0185] Semantic similarity: 0.92 (Exact match for "L'Oréal cosmetics case")

[0186] Time decay: 0.9 (Published 1 month ago)

[0187] Authority: 0.5 (Source: Internal Company Report)

[0188] Total score: 0.6 × 0.92 + 0.3 × 0.9 + 0.1 × 0.5 = 0.87

[0189] Ranking results: Candidate content B (score 0.87) is preferred over content A (score 0.78) because it has better semantic matching and timeliness.

[0190] Output information: Structured external dataset (including brand information, industry trends, and marketing campaigns). Data credibility score (0-1 point).

[0191] 3. Internal case data source engine (E→G→H)

[0192] Input information: Structured parameter object, such as (brand, industry level, output format).

[0193] Processing procedure:

[0194] The internal case engine is the core module of the intelligent marketing webpage generation system. It is responsible for the structured storage, intelligent retrieval, and dynamic scoring of the company's internal case library data, ensuring that the output content is highly matched with user needs. Its core processes include: knowledge base case library construction (scheduled task), semantic retrieval (based on industry tags and dissemination data), dynamic scoring (integrating semantic matching, industry relevance, and dissemination effect), and interpretable evaluation generation (CoT inference chain).

[0195] 3.1, such as Figure 4 As shown, the knowledge base and case library are constructed.

[0196] A [Scheduled Task] --> B {Project System Case}

[0197] B-->C [A Slice of the Literary Heart's One Word Model]

[0198] C --> D [Structured Storage]

[0199] Input: Original case text (unstructured data) or structured data from the enterprise project management system.

[0200] Processing procedure:

[0201] Slices of the Wenxin large model:

[0202] Entity recognition: Extracting brand, industry level (level 1 / level 2), and marketing nodes (such as "KOL collaboration").

[0203] Relation extraction: Construct triple relations (e.g., "L'Oréal-is-a-Consumer Goods→Cosmetics").

[0204] Structured storage:

[0205] Generate Markdown format example cards (including image links and dissemination data).

[0206] Stored in the Neo4j graph database (supports industry path queries) and the Milvus vector library (supports semantic retrieval).

[0207] Example: Original case text: "During the 2023 Double Eleven shopping festival, L'Oréal achieved a 120% increase in sales and 500,000 views on Weibo through a matrix of Douyin KOLs."

[0208] Processed structured data: JSON

[0209] {

[0210] "case_id":"CASE_001",

[0211] "brand":"L'Oréal",

[0212] "industry_level1":"Consumer Goods",

[0213] "industry_level2":"cosmetics",

[0214] "case_img":"http: / / marketing-case.qiniudn.com / olay-campaign-2023.jpg",

[0215] "case_intro":"Douyin KOL matrix marketing achieves 120% sales growth",

[0216] "spread_data":{"Weibo":500000,"WeChat":200000},

[0217] "semantic_vector":[0.12,-0.35,...] / / Sentence-BERT encoded vector

[0218] }

[0219] 3.2 Semantic Retrieval (User-Command Driven)

[0220] Input: User command parameters (brand, industry level).

[0221] Processing procedure:

[0222] Industry path filtering: Search Neo4j for all cases where the secondary industry is "cosmetics". Use Jaccard similarity to verify brand-industry association (e.g., "L'Oréal - Cosmetics" path match score ≥ 0.8).

[0223] Data filtering during transmission:

[0224] Calculate the propagation effect score using Python.

[0225] propagation_score = ∑(propagation amount × e^(-0.1 × Δt)) #Δt is the data time difference (months)

[0226] Candidate cases with a score >50 are selected.

[0227] Example:

[0228] User command parameters: {"brand":"L'Oréal","industry_level2":"cosmetics"}

[0229] Search results:

[0230] Case A: L'Oréal's 2023 Double Eleven Douyin campaign (Δt = 1 month, dissemination score = 85)

[0231] Case B: L'Oréal's 2022 Xiaohongshu (Little Red Book) campaign (Δt = 6 months, dissemination score = 60)

[0232] 3.3 Dynamic Scoring (Large Model Driven)

[0233] Input: List of candidate cases.

[0234] Processing procedure:

[0235] Semantic matching score: The cosine similarity between the case summary and the user instruction is calculated using Sentence-BERT.

[0236] Example: A user queries "cosmetics marketing trends", and Case A's summary similarity score is 0.92.

[0237] Industry matching score: Jaccard similarity based on industry path (e.g., "Consumer Goods → Cosmetics" path weight 0.8).

[0238] Propagation effect score: Time decay factor calculation: e^(-0.1×Δt) (0.90 when Δt = 1 month).

[0239] Comprehensive scoring formula: Python

[0240] final_score=0.6*semantic_sim+0.3*industry_sim+0.1*propagation_score

[0241] Example:

[0242]

[0243] 3.4 Interpretable Evaluation Generation (CoT Inference Chain)

[0244] Input: Rating data for the Top 5 cases.

[0245] Processing procedure:

[0246] Template population: Generate evaluation reasons based on rating dimensions.

[0247] Logical chain verification: Ensure that the evaluation reasons are consistent with the scoring parameters.

[0248] Highlights Analysis: The main model summarizes the highlights based on the content of the case study itself.

[0249] Example output:

[0250] Case A Highlights:

[0251] Highlights: 1. A full-chain content marketing strategy covering multiple stages including product selection and traffic generation on Douyin. 2. Attracting followers and increasing brand exposure through high-quality short videos.

[0252] Reasons for evaluating Case A: 1. Industry relevance reaches 98% (cosmetics, a secondary industry); 2. Over 500,000 Weibo views in the past month (timeliness weight 0.90); 3. Includes a reusable Douyin KOL matrix strategy (semantic relevance 0.92).

[0253] Application scenario examples

[0254] User commands:

[0255] "Please export the brand L'Oréal. The primary industry category is Consumer Goods, and the secondary industry category is Cosmetics. Please export in PDF format."

[0256] Internal case engine processing flow:

[0257] The search returned Case A (TikTok ad placement) and Case B (Xiaohongshu product seeding).

[0258] After scoring, Case A scored 0.87, and Case B scored 0.65.

[0259] When generating a PDF webpage, prioritize inserting the images and strategy description from Case A.

[0260] Output effect:

[0261] Case Name xxx Case Name

[0262] Case study images

[0263] Industry: Personal Care Fast-Moving Consumer Goods - Cosmetics

[0264] Business Type: E-commerce Campaign Marketing

[0265] Background: Initial product selection & content planning, attracting followers with high-quality short videos, content traffic generation, and continuous follower retention through video editing and segmentation.

[0266] Highlights: 1. A full-chain content marketing strategy, covering multiple stages including product selection and traffic generation on Douyin. 2. Attracting followers and increasing brand exposure through high-quality short videos. Rating: "Industry relevance 98%+, timeliness 0.90".

[0267] 4. Multimodal output adapter (I)

[0268] Input information: External dataset (F output) + Internal Top 5 cases (H output)

[0269] Processing procedure:

[0270] Format adaptation rule engine:

[0271]

[0272] Content integration: Use the BART model to integrate external data with case content to ensure logical coherence.

[0273] Output information: Markdown converted to multimodal web page files (PDF / HTML / Word). File metadata (generation time, version number, data source identifier).

[0274] Experimental data support

[0275]

[0276] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. An intelligent marketing document generation device based on multi-source data fusion, characterized in that, include: User instruction input module: used to receive marketing document generation requests submitted by users; NLP parsing engine module: connected to the user command input module, configured to perform natural language processing on the user command, extract structured parameters and verify the validity of the parameters; Parameter decomposition module: connected to the NLP parsing engine module, used to decompose the parsed instructions into task parameters corresponding to external data acquisition and internal case invocation; External data engine module: connected to the parameter decomposition module, configured to collect and process external industry data based on task parameters to generate semantically enhanced industry information; Internal Case Engine Module: Connected to the parameter decomposition module, configured to initiate a case library search based on task parameters, filter matching marketing cases, and output them; Multimodal output module: Connected to the external data engine module and the internal case engine module respectively, it is used to integrate industry information and marketing cases to generate and output marketing documents in multiple formats; Data layer interface module: Connects to the external data engine module and the internal case engine module, includes a standardized API gateway, supports RESTful / SOAP dual protocol access to heterogeneous data sources, and realizes data interaction; Knowledge graph storage unit: connected to the internal case engine module and NLP parsing engine module, using Neo4j graph database to store entity relationships, covering the association form of "brand-industry-case profile", used for semantic association and case retrieval support; Core algorithm engine: including Hybrid generation engine: Connected to the multimodal output module and parameter decomposition module, it integrates the Wenxin Yiyan API call module and rule engine, supports dynamic adjustment of generation parameters, and is used for intelligent construction of marketing content; Case recommendation and evaluation engine: Connected to the internal case engine module, it deploys a TF-IDF vector index library, supports millisecond-level similarity calculation for tens of millions of cases, and provides similarity evaluation for case library retrieval; Case library retrieval module: connected to the internal case engine module and case recommendation and evaluation engine, relying on the semantic association of the knowledge graph storage unit and combined with the similarity calculation results of the case recommendation and evaluation engine, to select suitable cases from the case library; The internal case handling process includes: S31. Knowledge base case library construction steps: The original cases are sliced ​​into large model segments by a scheduled task, entity relationships are extracted and structured and stored in the Neo4j graph database and Milvus vector database; S32. Semantic retrieval steps: Based on user command parameters, perform industry path filtering in Neo4j, and combine the propagation data score (propagation_score = ∑(propagation volume × e^(-0.1 × Δt)) to select candidate cases, where: Dissemination volume refers to the dissemination data of the case on Weibo and WeChat platforms; Δt represents the time difference between the data release time and the current time, in months. S33. Dynamic scoring steps: using formula The final_score = 0.6 × semantic_sim + 0.3 × industry_sim + 0.1 × propagation_score is used to comprehensively score the candidate cases, where: industry_sim represents the industry matching degree, calculated based on Jaccard similarity of the industry path; semantic_sim is the semantic similarity between the case summary and the user command, calculated using the Sentence-BERT model; S34. Explainable evaluation generation steps: Generate a CoT reasoning chain evaluation report based on the scoring dimensions, and summarize the highlights of the case; The multimodal output process includes: S41. Format adaptation steps: Based on the output format, including PDF / HTML / Word, use Apache POI and pdfkit components to achieve automated conversion from Markdown to the target format; S42. Content Integration Steps: Use the BART model to integrate external data with case content, ensuring a logical coherence score > 0.7; S43. Metadata generation steps: Output document metadata including data source identifier and generation time.

2. The intelligent marketing document generation device based on multi-source data fusion according to claim 1, characterized in that, The NLP parsing engine module uses a natural language processing model to perform word segmentation, entity recognition, and intent understanding on user commands. The natural language processing model includes a pre-trained language model. The external data engine module includes a data acquisition unit and a data processing unit. The data acquisition unit is configured to call third-party data sources to obtain industry trend and market dynamic data. The data processing unit is configured to clean, semantically enhance, and score the importance of the collected data.

3. The intelligent marketing document generation device based on multi-source data fusion according to claim 1, characterized in that, When the internal case engine module initiates a case library retrieval, it calls the TF-IDF vector index library capability of the case recommendation and evaluation engine, and combines it with the entity relationship association of the knowledge graph storage unit to achieve case filtering.

4. The intelligent marketing document generation device based on multi-source data fusion according to claim 3, characterized in that, The semantic matching of the case library retrieval module adopts the vector similarity calculation method, which compares the user command parameters with the case feature vectors and filters cases whose similarity meets the threshold. The vector similarity calculation is implemented based on the TF-IDF vector index library of the case recommendation and evaluation engine.

5. A method for generating intelligent marketing documents based on multi-source data fusion, characterized in that: Production methods include: S1. User instruction parsing process: Process the raw instructions to obtain structured parameters; S2. External data acquisition process: collecting and optimizing industry data based on structured parameters; S3. Internal case processing workflow: searching, scoring, and generating evaluations from the enterprise case library; S4. Multimodal output process, converting fused data into a target format document; The S3 internal case processing flow includes: S31. Knowledge base case library construction steps: The original cases are sliced ​​into large model segments by a scheduled task, entity relationships are extracted and structured and stored in the Neo4j graph database and Milvus vector database; S32. Semantic retrieval steps: Based on user command parameters, perform industry path filtering in Neo4j, and combine the propagation data score (propagation_score = ∑(propagation volume × e^(-0.1 × Δt)) to select candidate cases, where: Dissemination volume refers to the dissemination data of the case on Weibo and WeChat platforms; Δt represents the time difference between the data release time and the current time, in months. S33. Dynamic scoring steps: using formula The final_score = 0.6 × semantic_sim + 0.3 × industry_sim + 0.1 × propagation_score is used to comprehensively score the candidate cases, where: industry_sim represents the industry matching degree, calculated based on Jaccard similarity of the industry path; semantic_sim is the semantic similarity between the case summary and the user command, calculated using the Sentence-BERT model; S34. Explainable evaluation generation steps: Generate a CoT reasoning chain evaluation report based on the scoring dimensions, and summarize the highlights of the case; The S4 multimodal output process includes: S41. Format adaptation steps: Based on the output format, including PDF / HTML / Word, use Apache POI and pdfkit components to achieve automated conversion from Markdown to the target format; S42. Content Integration Steps: Use the BART model to integrate external data with case content, ensuring a logical coherence score > 0.7; S43. Metadata generation steps: Output document metadata including data source identifier and generation time.

6. The method according to claim 5, characterized in that, The S1 user instruction parsing process includes: S11. Cleaning the original instruction text using regular expressions to remove redundant symbols and unify punctuation format; S12. Extract brand, industry level, and output format parameters from the model to generate a structured JSON object; S13. Verify the legitimacy of the brand-industry level based on the Neo4j knowledge graph and output the verification identifier.

7. The method according to claim 5, characterized in that, The S2 external data acquisition process includes: S21. Construct a search query based on structured parameters, call Baidu's big data model to obtain industry data, and deduplicate data using a similarity hashing algorithm; S22. Use the Wenxin large model to perform TF-IDF cleaning, BERT-CLS semantic enhancement, and summary generation on the data; S23. Sort the data using a weighted formula: score = 0.6 × semantic_sim + 0.3 × time_decay + 0.1 × authority, where: semantic_sim represents semantic similarity, which is the degree of relevance between text and user needs calculated based on the model. time_decay is a time decay factor, determined based on the difference between the data release time and the current time; The authority rating is based on the source authority of industry white papers and internal company reports. The top N data points are then selected based on their scores.

Images