Marketing system based on intelligent analysis of multi-modal document data

Abstract

The application provides a marketing system based on intelligent analysis of multi-modal literature data, which automatically identifies and analyzes various formats of scientific research literature through a literature analysis and preprocessing module, extracts structured information; an AI enhanced analysis module uses a large model to deeply mine the semantics of the literature; a data processing and integration module cleans, integrates and constructs a researcher portrait containing five dimensions of identity, interest, technology, equipment and demand prediction, and then generates a personalized marketing strategy; a cache management module optimizes the storage and retrieval efficiency of the results, and a batch processing module coordinates the parallel processing process of multiple files; the system realizes the automatic conversion from raw literature to precise marketing strategy. The application realizes deep mining and precise marketing of multi-source scientific research literature, significantly improves the information extraction accuracy and marketing conversion efficiency.

Description

Technical Field

[0001] This invention belongs to the field of Internet data acquisition technology and intelligent information analysis and processing, and specifically relates to a marketing system based on intelligent analysis of multimodal literature data. Background Technology

[0002] Internet data collection and intelligent analysis technology refers to a technical system that uses automated means to acquire data from the Internet, such as websites, apps, APIs, and sensors, and then uses machine learning, big data processing, and other technologies to clean, store, analyze, and mine the data in order to extract valuable information.

[0003] Currently, significant progress has been made in internet data collection and intelligent analysis technologies, but challenges remain in areas such as data quality, tag depth, and dynamic modeling.

[0004] In the field of data acquisition, technological development is showing a trend towards diversification and efficiency. Focusing on technologies such as web crawling and incremental crawling, these technologies have been widely applied in e-commerce, news, and other fields. Combined with dynamic IP proxies and anti-anti-crawling strategies, they have significantly improved the accuracy and efficiency of data acquisition. Meanwhile, the maturity of distributed architectures such as Scrapy-Redis and real-time stream processing technologies such as Kafka+Flink has made efficient acquisition of hundreds of millions of data points and millisecond-level real-time responses possible. However, two core problems still exist in this field: uncontrollable data quality, such as false information and chaotic formats of multi-source data, and the contradiction between real-time performance and resource costs, such as the exponential increase in server costs due to low-latency requirements.

[0005] In the field of intelligent analytics, the optimization of machine learning and deep learning algorithms has significantly improved data processing capabilities. Accuracy rates for CV / NLP tasks have approached or surpassed human levels; for example, BERT's F1 score for text classification reaches 0.92, and lightweight technologies like MobileBERT enable models to adapt to edge devices. Furthermore, the widespread adoption of AutoML tools and low-code visualization platforms like QuickBI has lowered the barrier to entry for data analysis. However, this field still faces significant challenges: the shallowness of labeling systems leads to insufficient semantic granularity; weak dynamic modeling capabilities make traditional algorithms ill-suited to shifting user interests—for example, the K-means model experiences a 15% weekly accuracy decline due to changes in data distribution; and multimodal data fusion is difficult, with limited improvement in the effectiveness of joint analysis of text and video data.

[0006] The limitations of the technology are further highlighted by the shortcomings in typical application scenarios. For example, in the supply chain of scientific research consumables, the lack of fine-grained correlation in the tagging system results in an inventory turnover rate that is 20% lower than the industry average; in cross-platform user profiling scenarios, coarse-grained tags fail to identify behavioral motivations, leading to a 12% decrease in ad click-through rates.

[0007] As the scale of deep learning model parameters expands dramatically, while the model's representational capabilities increase, the decision-making process becomes increasingly "black box" with a significant decrease in feature interpretability. This lack of interpretability severely restricts the application of AI in high-risk fields such as medical diagnosis and financial risk control, as it fails to meet the requirements of accurate results, transparent and auditable decision-making logic. In recommendation systems, new users or items are difficult to represent reliably due to sparse behavioral data, resulting in poor initial recommendation performance. However, frequent model updates to improve performance introduce instability in prediction results, increase business decision-making risks, and even disrupt the consistency of user experience. These two contradictions essentially reveal the deep tension between "performance improvement" and "reliability assurance," and between "data-driven" and "logical transparency" in the development of AI technology.

[0008] Overall, current technology is still in the stage of perceptual intelligence, and there are obvious bottlenecks in areas such as data authenticity, semantic understanding, and dynamic adaptability.

[0009] Against this technological backdrop, the limitations of existing technologies become particularly pronounced when targeting vertical, specialized fields. Especially in the specific area of ​​scientific literature data analysis and precision marketing, the existing technological architecture struggles to effectively address the following three core challenges:

[0010] First, the extraction of structured information from multi-source heterogeneous documents is incomplete and inaccurate, making deep integration difficult. Scientific research literature data exists in various formats such as PDF, XML, Word, and HTML, with different internal structures, encoding standards, and content presentation methods. Existing general collection and parsing technologies lack the ability to deeply adapt to the structure of academic documents, resulting in insufficient accuracy in extracting elements such as complex layouts, mathematical formulas, charts, and citations. This makes it difficult to achieve unified, high-quality structured integration of cross-format and cross-source data, creating "data silos."

[0011] Second, the understanding of the literature content remains at the level of keywords and superficial semantics, lacking deeper semantic insights into research intentions, technical routes, and the logic behind the use of equipment and consumables. Although existing NLP technologies can achieve high accuracy rates in text classification and entity recognition in general domains, they struggle to understand complex scientific concepts, logical connections between experimental methods, and the motivations behind technology selection in highly specialized scientific literature. This results in the inability to accurately extract high-value information from the literature, such as "the models of precision instruments required for specific experimental steps," "alternatives to key chemical reagents," or "the software toolchain upon which data analysis relies."

[0012] Third, user profiles built on shallow tags are limited and cannot support truly personalized recommendations for research products and services. Existing user profiling technologies rely heavily on superficial features such as behavioral clicks and keyword frequency, making it difficult to construct dynamic, multi-dimensional knowledge models that reflect researchers' professional capabilities, technical preferences, equipment usage history, and future demand trends. Therefore, the resulting marketing strategies are often generalized and outdated, failing to accurately match researchers' real, deep, and evolving research needs, leading to low marketing conversion rates. Summary of the Invention

[0013] The purpose of this invention is to provide a marketing system based on intelligent analysis of multimodal literature data, addressing the problems in the prior art.

[0014] Therefore, the above-mentioned objectives of the present invention are achieved through the following technical solutions:

[0015] A marketing system based on intelligent analysis of multimodal literature data, including:

[0016] The document parsing and preprocessing module uses a format recognizer to identify the format of the input documents, and a format-specific parser to perform deep content extraction and structural parsing. Then, a unified document tree structure representation is generated through a document structure parser, a metadata extractor obtains key information, and a content partitioner divides logical parts, achieving efficient and structured preprocessing of heterogeneous scientific research documents. The information is then passed to the AI-enhanced analysis module through a standardized JSON data structure output.

[0017] The AI-enhanced analysis module receives the JSON data, extracts deep semantic information from the document through a semantic chunker, a domain-specific prompt generator, a multi-LLM API connector, and a response parser, and outputs it as structured data in JSON-LD format.

[0018] The data processing and integration module cleans, verifies, and removes duplicates from the JSON-LD data. Based on the processed data, it constructs a researcher knowledge graph with multiple dimensions, including basic identity, research interests, technology stack, equipment usage, and demand prediction, through a user profile generator. It also uses a marketing strategy recommender to generate personalized product combinations and marketing plans, realizing a complete transformation from raw data to precise marketing strategies. The cleaning, verification, and structured processing extract the results.

[0019] The cache management module, connected to the data processing and integration module, is used to store and retrieve analysis results and user profiles based on content hashing.

[0020] The batch processing module is connected to the document parsing and preprocessing module, the AI-enhanced analysis module, and the cache management module, respectively, to coordinate the parallel processing and cross-correlation analysis of multiple files;

[0021] The document analysis and preprocessing module, the AI-enhanced analysis module, and the data processing and integration module are connected in series to form an end-to-end processing pipeline from original documents to marketing strategies.

[0022] While adopting the above technical solutions, the present invention may also adopt or combine the following technical solutions:

[0023] As a preferred embodiment of the present invention, the document parsing and preprocessing module includes:

[0024] The format recognizer consists of a three-level format recognition mechanism, which includes a primary judgment based on the file extension, a secondary verification based on the magic number of the file header, and a final decision based on the characteristics of the file content.

[0025] Multiple format-specific parsers, including XML file reader, PDF parser, JSON parser, HTML parser, Word document parser, TXT parser, and EPUB parser;

[0026] A document structure parser that generates a unified document tree structure representation;

[0027] Metadata extractor, extracts the title, author and DOI information of documents;

[0028] Content partitioner divides document content according to academic structure;

[0029] The document parsing and preprocessing module achieves standardized preprocessing of heterogeneous documents through three-level format recognition, multi-parser adaptation, unified structure reconstruction, metadata extraction, and academic content partitioning.

[0030] As a preferred embodiment of the present invention, the PDF parser includes:

[0031] Text, image, and table separation units based on PyMuPDF;

[0032] A scanned document processing unit integrating Tesseract-OCR;

[0033] A layout analysis unit employing computer vision algorithms is used to identify the chapter structure of a document;

[0034] Mathematical formula recognition and MathML conversion unit;

[0035] The PDF parser, through the collaborative work of the aforementioned units, achieves high-precision content and structure extraction from both native and scanned PDFs.

[0036] As a preferred embodiment of the present invention, the AI-enhanced analysis module includes:

[0037] The text segmenter employs a segmentation algorithm that protects semantic integrity.

[0038] The prompt generator contains a library of at least 200 domain-specific prompt templates. These templates are specifically designed for different parts of scientific literature and different extraction targets, and are used to guide large language models to accurately extract information about experimental instruments, experimental consumables, and software tools.

[0039] API connector, supports dynamic switching between multiple large language model APIs;

[0040] A response parser is used to convert unstructured responses into JSON-LD format;

[0041] The AI-enhanced analysis module achieves domain knowledge enhancement and standardized analysis through semantic segmentation, dedicated prompt templates, dynamic scheduling of multi-model APIs, and JSON-LD structured output.

[0042] As a preferred embodiment of the present invention, the data processing and integration module includes:

[0043] Data cleaners implement rule-based and machine learning-based noise filtering;

[0044] User profile generator, which constructs a five-dimensional profile including basic information, research interests, technology stack, device usage, and demand forecast;

[0045] The marketing strategy recommender uses a recommendation algorithm that combines collaborative filtering and knowledge graphs.

[0046] The data processing and integration module achieves precise marketing strategies through data cleaning, five-dimensional user profile construction, and collaborative filtering and knowledge graph fusion recommendation.

[0047] As a preferred technical solution of the present invention: in the five-dimensional researcher knowledge graph constructed by the user profile generator:

[0048] The technology stack dimensions include researchers' experimental methods, data analysis software, and programming languages;

[0049] The device usage dimensions include the device model, brand, and usage scenario mentioned in the literature;

[0050] The demand forecasting dimension is based on the researcher's historical technology stack, equipment usage records, and the evolution trend of their research field, using machine learning models to predict the required equipment and consumables.

[0051] As a preferred technical solution of the present invention: in the marketing strategy recommender:

[0052] The collaborative filtering algorithm is used to discover researchers within a knowledge graph, calculating the mixed similarity between the current researcher and other researcher nodes. This mixed similarity includes a linear or non-linear combination of similarity based on graph topology and similarity based on researcher multidimensional attributes. Furthermore, it identifies the product preferences of other researchers with high mixed similarity to the current researcher. Similarity based on graph topology includes, for example, based on common neighbors, while similarity based on researcher multidimensional attributes includes, for example, based on research interests and technology stack.

[0053] The knowledge graph reasoning is used to deduce potential product requirements based on the entity association paths between "technology-equipment-consumables".

[0054] As a preferred embodiment of the present invention, the cache management module includes:

[0055] A cache key generator based on content hashing;

[0056] Data storage that supports a hybrid replacement strategy of LRU and LFU;

[0057] Bloom filters accelerate search engines;

[0058] The cache management module is a high-performance hierarchical caching system that generates unique keys through content hashing, optimizes memory using a hybrid eviction policy, and prevents cache penetration with the help of Bloom filters.

[0059] As a preferred embodiment of the present invention, the batch processing module includes:

[0060] A task scheduler with dynamic resource allocation;

[0061] Parallel processor based on work-stealing algorithm;

[0062] A cross-analyzer that supports multi-dimensional correlation analysis;

[0063] The batch processing module achieves high-performance batch task processing through dynamic resource scheduling, work-stealing parallel processing, and correlation cross-analysis.

[0064] As a preferred technical solution of the present invention, it also includes a web application interactive interface, which is connected to the cache management module and provides a user operation interface.

[0065] Compared with existing technologies, the marketing system based on multimodal literature data intelligent analysis of the present invention has the following beneficial effects: The present invention realizes end-to-end intelligent transformation from original documents to executable marketing strategies by constructing an integrated technical framework that couples multimodal parsing, AI-enhanced semantic mining, dynamic knowledge graph construction and strategy generation.

[0066] Compared with the prior art, the significant effects of the technical solution of the present invention are mainly reflected in the following aspects:

[0067] 1. A multimodal document deep fusion preprocessing mechanism based on "three-level judgment - dedicated parsing - unified structure" is proposed. This mechanism accurately identifies document formats through three-level judgment based on file extension, magic number, and content features, and calls a dedicated parser (such as a PDF parser with OCR, layout analysis, and formula recognition capabilities) for deep content extraction. Subsequently, a document structure parser maps documents of different formats to a unified document tree structure representation, fundamentally solving the structuring bottleneck of heterogeneous data sources, providing a high-quality and standardized data foundation for subsequent analysis, and significantly improving the complete parsing rate of mixed-format documents.

[0068] 2. A new AI-enhanced analytics paradigm, "Semantic Chunking - Domain Hints - Multi-Model Scheduling," was established. This paradigm first employs a semantic integrity protection algorithm to chunk the text, ensuring key information remains intact. Then, a hint library containing over 200 domain-specific hint templates guides a large language model to perform refined semantic mining on different parts of the document (e.g., methods, results). Finally, an API connector supporting dynamic switching between multiple LLMs balances analytical performance and cost-effectiveness. This method is particularly suitable for accurately extracting professional information such as experimental instruments, consumables, and software tools from academic texts, achieving over 20% improvement in accuracy and recall compared to traditional NLP methods.

[0069] 3. A precise marketing decision-making model based on "five-dimensional profile - knowledge graph - integrated recommendation" was constructed. This invention breaks through the limitations of past single-dimensional user profiling technology. Based on extracted structured information, it dynamically constructs a researcher knowledge graph encompassing five dimensions: basic identity, research interests, technology stack, device usage history, and potential demand prediction. On this basis, the marketing strategy recommender employs a hybrid algorithm combining collaborative filtering and knowledge graph reasoning. It not only recommends products directly related to research content but also predicts future demand trends for researchers, thereby achieving forward-looking marketing strategies. This model improves marketing accuracy from approximately 2.3% using traditional methods to over 12.7%.

[0070] 4. A system performance optimization framework linking "caching and batch processing" was designed. A high-performance hierarchical caching system was constructed using a content hash-based cache key generator, an LRU / LFU hybrid eviction strategy, and a Bloom filter retrieval engine, significantly reducing redundant computation. Simultaneously, a parallel processor with dynamic resource allocation and a work-stealing algorithm enabled high-efficiency batch document processing and cross-reference analysis, achieving a processing speed 3-4 times faster than traditional methods, and identifying research hotspots and technical connections across documents.

[0071] In summary, through the collaborative innovation and deep integration of the above-mentioned technical aspects, this invention has successfully solved a series of technical challenges, such as data fusion of multi-source heterogeneous scientific research literature, deep semantic understanding, and generation of precise marketing strategies, and has realized the automated and intelligent transformation from massive and disordered literature data to high-value and actionable marketing insights. Attached Figure Description

[0072] Figure 1 This is a schematic diagram of the marketing system based on intelligent analysis of multimodal literature data according to the present invention. Figure 1 ;

[0073] Figure 2 This is a schematic diagram of the marketing system based on intelligent analysis of multimodal literature data according to the present invention. Figure 2 . Detailed Implementation

[0074] The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

[0075] The marketing system based on intelligent analysis of multimodal literature data of the present invention includes,

[0076] The document parsing and preprocessing module identifies the format of input documents through a three-level format recognition mechanism, performs in-depth content extraction and structure parsing using a dedicated format parser, unifies the internal representation through a document structure parser, obtains key information through a metadata extractor, and divides logical parts through a content partitioner, thereby achieving efficient and structured preprocessing of heterogeneous scientific research documents. The information is then transmitted to the AI-enhanced analysis module through a standardized JSON data structure.

[0077] The AI-enhanced analysis module receives the JSON data, extracts deep semantic information from the document through a semantic chunker, a domain-specific prompt generator, a multi-LLM API connector, and a response parser, and outputs it as structured data in JSON-LD format.

[0078] The data processing and integration module cleans, verifies, and removes duplicates from the JSON-LD data. Based on the processed data, it constructs a researcher knowledge graph with multiple dimensions, including basic identity, research interests, technology stack, equipment usage, and demand prediction, through a user profile generator. It also uses a marketing strategy recommender to generate personalized product combinations and marketing plans, realizing a complete transformation from raw data to precise marketing strategies. The cleaning, verification, and structured processing extract the results.

[0079] The cache management module, connected to the data processing and integration module, is used to store and retrieve analysis results and user profiles based on content hashing.

[0080] The batch processing module is connected to the document parsing and preprocessing module, the AI-enhanced analysis module, and the cache management module to coordinate the parallel processing and cross-correlation analysis of multiple files.

[0081] The document analysis and preprocessing module, the AI-enhanced analysis module, and the data processing and integration module are connected in series to form an end-to-end processing pipeline from original documents to marketing strategies.

[0082] The document parsing and preprocessing module includes:

[0083] The format recognizer includes a three-level format recognition mechanism: primary judgment based on file extension, secondary verification based on file header magic number, and final decision based on file content features.

[0084] Multiple format-specific parsers, including XML file reader, PDF parser, JSON parser, HTML parser, Word document parser, TXT parser, and EPUB parser;

[0085] A document structure parser that generates a unified document tree structure representation;

[0086] Metadata extractor, extracts the title, author and DOI information of documents;

[0087] Content partitioner divides document content according to academic structure;

[0088] The document parsing and preprocessing module achieves standardized preprocessing of heterogeneous documents through three-level format recognition, multi-parser adaptation, unified structure reconstruction, metadata extraction, and academic content partitioning.

[0089] The PDF parser includes:

[0090] Text, image, and table separation units based on PyMuPDF;

[0091] A scanned document processing unit integrating Tesseract-OCR;

[0092] A layout analysis unit employing computer vision algorithms is used to identify the chapter structure of a document;

[0093] Mathematical formula recognition and MathML conversion unit;

[0094] The PDF parser, through the collaborative work of the aforementioned units, achieves high-precision content and structure extraction from both native and scanned PDFs.

[0095] The AI-enhanced analysis module includes:

[0096] The text segmenter employs a segmentation algorithm that protects semantic integrity.

[0097] A prompt generator containing a library of at least 200 domain-specific prompt templates;

[0098] API connector, supports dynamic switching between multiple large language model APIs;

[0099] A response parser is used to convert unstructured responses into JSON-LD format;

[0100] The AI-enhanced analysis module achieves domain knowledge enhancement and standardized analysis through semantic segmentation, dedicated prompt templates, dynamic scheduling of multi-model APIs, and JSON-LD structured output.

[0101] The data processing and integration module includes:

[0102] Data cleaners implement rule-based and machine learning-based noise filtering;

[0103] User profile generator, which constructs a five-dimensional profile including basic information, research interests, technology stack, device usage, and demand forecast;

[0104] The marketing strategy recommender employs a hybrid recommendation model that combines collaborative filtering algorithms with knowledge graph reasoning.

[0105] The data processing and integration module achieves precise marketing strategies through data cleaning, five-dimensional user profile construction, and collaborative filtering and knowledge graph fusion recommendation.

[0106] In the five-dimensional researcher knowledge graph constructed by the user profile generator:

[0107] The technology stack dimensions include experimental methods, data analysis software, and programming languages ​​commonly used by researchers.

[0108] The device usage dimensions include the device model, brand, and usage scenario mentioned in the literature;

[0109] The demand forecasting dimension is based on the researcher's historical technology stack, equipment usage records, and the evolution trend of their research field. It uses machine learning models to predict the equipment and consumables that the researcher may need in the future.

[0110] In the marketing strategy recommender:

[0111] The collaborative filtering algorithm is used to calculate the mixed similarity between the current researcher node and other researcher nodes based on the researcher knowledge graph; the mixed similarity is a linear or nonlinear combination of a) similarity based on the graph topology (such as based on common neighbors) and b) similarity based on the researcher's multidimensional attributes (such as based on research interests and technology stack); thereby discovering the product preferences of other researchers who have high mixed similarity with the current researcher.

[0112] The knowledge graph reasoning is used to deduce potential product requirements based on the entity association paths between "technology-equipment-consumables". The cache management module includes:

[0113] A content-hash-based cache key generator is used to generate globally unique identifiers for analysis results;

[0114] Data storage that supports a hybrid LRU and LFU eviction policy is used to optimize memory usage;

[0115] A Bloom filter accelerates the search engine and is used to quickly determine whether the requested result does not exist in the cache, thus preventing cache penetration.

[0116] The cache management module is a high-performance hierarchical caching system that generates unique keys through content hashing, optimizes memory using a hybrid eviction policy, and prevents cache penetration with the help of Bloom filters.

[0117] The batch processing module includes:

[0118] A task scheduler with dynamic resource allocation dynamically allocates computing resources based on file size, parsing complexity, and the current system load.

[0119] A parallel processor based on a work-stealing algorithm achieves load balancing for multiple document parsing and AI analysis tasks;

[0120] A cross-analyzer that supports multi-dimensional correlation analysis can perform correlation analysis on the results of multiple documents after batch processing to identify common research teams, technological hotspots, and equipment usage patterns.

[0121] The batch processing module achieves high-performance batch task processing through dynamic resource scheduling, work-stealing parallel processing, and correlation cross-analysis.

[0122] It also includes a web application interface that connects to the cache management module, providing users with an interface for uploading documents, configuring parameters, visualizing analysis results, and implementing marketing strategies.

[0123] Compared with the prior art, the present invention has the following beneficial effects:

[0124] 1. Unified Multi-Format Document Processing Framework: A unified processing framework based on format recognition and a dedicated parser has been created, supporting 10 scientific literature formats including XML, PDF, JSON, HTML, Word, and EPUB. This breaks through the limitations of traditional single-format processing and greatly expands the comprehensiveness of data sources.

[0125] 2. AI-enhanced deep information extraction technology: An enhanced analysis technology based on a large language model was developed. Through a specially designed prompting engineering strategy, refined analysis was performed on different parts of the document, which significantly improved the accuracy (by more than 20%) and recall (by more than 25%) of professional information such as experimental instruments and consumables.

[0126] 3. Multi-dimensional scientific research profile construction method: Based on the extracted information, a multi-dimensional profile of scientific researchers is constructed, including research direction, equipment used, experimental consumables, software tools, etc., which improves the marketing accuracy from the traditional 2.3% to 12.7%, an increase of 5.5 times.

[0127] 4. High-efficiency parallel batch processing and cross-analysis technology: A parallel processing architecture with dynamic resource allocation and a multi-document cross-analysis algorithm were designed, which improves the processing speed to 3-4 times that of traditional methods. At the same time, it can discover deep correlations between documents and identify research hotspots and core research teams.

[0128] Example 1

[0129] like Figures 1-2 As shown, the marketing system based on multimodal literature data intelligent analysis of the present invention provides a multi-format scientific literature intelligent analysis system based on a large language model. This system addresses the limitations of existing scientific literature processing methods, such as format constraints, inaccurate information extraction, and insufficient marketing precision, by developing a complete solution. The system includes a literature parsing and preprocessing module 1, an AI-enhanced analysis module 2, a data processing and integration module 3, a cache management module 4, a web application interface 5, and a batch processing module 6.

[0130] like Figure 1 As shown, the document parsing and preprocessing module 1 and the AI-enhanced analysis module 2 exchange information through a standardized JSON data structure. The AI-enhanced analysis module 2 and the data processing and integration module 3 exchange analysis results through a structured object array. The data processing and integration module 3 and the cache management module 4 access data through key-value pair mapping. The cache management module 4 and the web application interface 5 transmit query results through a RESTful API interface. The web application interface 5 and the batch processing module 6 manage batch requests through a task queue object. The batch processing module 6 and the document parsing and preprocessing module 1 transmit batch processing tasks through a file path array and a configuration object. Simultaneously, the batch processing module 6 maintains a feedback connection with the AI-enhanced analysis module 2 and the cache management module 4, forming a complete data processing closed loop. Data transmission between modules uses standardized interface definitions to ensure the system has high scalability and modularity. In the architecture diagram, solid arrows indicate the main data flow, and dashed arrows indicate feedback or control flow.

[0131] The document parsing and preprocessing module 1 includes the following components: format recognizer 10, XML file reader 11, PDF parser 11A, JSON parser 11B, HTML parser 11C, Word document parser 11D, TXT parser 11E, EPUB parser 11F, document structure parser 12, metadata extractor 13, and content partitioner 14.

[0132] The format recognizer 10 is responsible for identifying the format type of the input document. It performs multi-level judgments based on file extension, magic number, and content features to accurately identify the file format and call the appropriate parser for each format. Specifically, format recognition uses a decision tree algorithm. First, it checks the file extension; second, it verifies the magic number in the file header; and finally, it analyzes content features. This three-level judgment ensures correct format recognition even when the file extension is modified. The recognition accuracy reaches 99.8%, significantly higher than the 92.3% of traditional single-feature recognition methods.

[0133] XML file reader 11 is responsible for reading input XML files, supports XML namespaces and tag structures from mainstream publishers such as PubMed, Elsevier, and Springer, and uses both SAX and DOM parsing methods, automatically selecting the optimal parsing strategy based on the file size;

[0134] PDF parser 11A is responsible for processing PDF documents, including the following core functions: (1) content extraction based on PyMuPDF, supporting text, image and table separation; (2) text recognition of scanned documents using Tesseract-OCR, supporting multilingual recognition; (3) layout analysis algorithm based on machine learning, recognizing the secondary structure of the article; (4) automatic recognition of citations and figures; (5) mathematical formula OCR and structured storage.

[0135] The JSON parser 11B is responsible for parsing JSON formatted scientific research data. It is suitable for API data provided by arXiv, PubMed Central, etc., including: (1) recursively parsing complex nested structures; (2) handling various Unicode encodings and special characters; (3) automatically recognizing and converting date, numeric and other fields in different formats; and (4) handling incomplete or incorrect JSON structures.

[0136] HTML parser 11C processes scientific literature in web page format and supports: (1) DOM parsing using Beautiful Soup and lxml; (2) extracting structured content based on XPath and CSS selectors; (3) handling dynamic content rendered by JavaScript; and (4) automatically identifying and extracting tables, citations, and references in HTML.

[0137] Word document parser 11D processes DOC / DOCX format scientific research documents, with functions including: (1) extracting text content and format information using python-docx and antiword; (2) identifying and preserving document style, structure and hierarchical headings; (3) extracting embedded charts and formulas; and (4) processing annotations and revisions in the document.

[0138] The TXT parser 11E processes plain text format research materials, and its functions include: (1) automatically recognizing text encoding; (2) recognizing document structure based on natural language processing technology; (3) using semantic analysis technology to distinguish between abstract, methods, results and other parts; and (4) automatically recognizing citation and reference formats.

[0139] The EPUB parser 11F processes scientific research materials in ebook format. Its functions include: (1) parsing the EPUB container structure; (2) extracting chapter content and metadata; (3) processing embedded HTML / XHTML content; and (4) extracting and associating image resources.

[0140] The document structure parser 12 parses the document tree structure based on the characteristics of various documents and adopts a unified internal representation to realize the structural mapping between different formats; the metadata extractor 13 extracts metadata such as article title, DOI, and publication date; and the content partitioner 14 divides the article content into different parts such as abstract, methods, and results.

[0141] The AI-enhanced analysis module 2 includes the following components: a text chunker 21, a prompt generator 22, an API connector 23, and a response parser 24. The text chunker 21 segments long texts into appropriately sized processing blocks; the prompt generator 22 generates targeted AI prompts based on different text types; the API connector 23 is responsible for communicating with the large language model (Gemini API); and the response parser 24 converts the text returned by the AI ​​into structured data.

[0142] The data processing and integration module 3 includes the following components: a data cleaner 31, a validator 32, a deduplication and merging unit 33, a user profile generator 34, and a marketing strategy recommender 35. The data cleaner 31 removes noise and redundant information from the analysis results; the validator 32 ensures the accuracy of the extracted information through cross-validation; the deduplication and merging unit 33 merges analysis results from multiple sources into a unified data structure; the user profile generator 34 constructs a multi-dimensional researcher profile based on the extracted information; and the marketing strategy recommender 35 generates personalized marketing suggestions based on the user profile.

[0143] User Profile Generator 34 is one of the core components of this system, responsible for transforming various types of information extracted from literature into structured researcher profiles. This component employs a multi-level tagging system and knowledge graph technology to construct a comprehensive profile encompassing the following dimensions: (1) Basic Identity Dimension: including basic information such as researcher's name, institution, title, and research field; (2) Research Interest Dimension: identifying the researcher's core research interests and development trajectory by analyzing the themes, citation patterns, and collaboration networks of their published literature; (3) Technology Application Dimension: constructing a researcher's technology preference model based on experimental methods, technical routes, and data processing methods mentioned in the literature; (4) Equipment and Consumables Dimension: establishing a researcher's equipment usage profile by extracting information on experimental equipment, reagents, and consumables from the literature; (5) Demand Prediction Dimension: predicting the researcher's potential equipment and consumables needs by combining historical research trajectories and field development trends. User Profile Generator 34 realizes the transformation from static text to dynamic user models, providing a data foundation for precision marketing.

[0144] The marketing strategy recommender 35, based on the researcher profile built by the user profile generator 34, combines a product database and a marketing knowledge base to automatically generate personalized marketing strategy suggestions. This component employs a hybrid recommendation algorithm combining rule-based and machine learning approaches, recommending the most suitable product combinations and marketing methods based on the researcher's professional background, research stage, and equipment requirements. The recommendation system considers multiple factors, including the match between products and research needs, the researcher's budget, institutional procurement cycles, and past marketing response rates, generating the optimal recommendation after comprehensive scoring. Furthermore, this component provides A / B testing capabilities, allowing the simultaneous generation of multiple marketing strategy options for small-scale testing to verify effectiveness before large-scale application.

[0145] The cache management module 4 includes the following components: a cache key generator 41, a data storage device 42, a retrieval engine 43, and a cache cleaner 44. The cache key generator 41 generates a unique identifier based on file content and parameters; the data storage device 42 saves the analysis results to the local file system; the retrieval engine 43 quickly retrieves existing results based on the cache key; and the cache cleaner 44 is responsible for cleaning up expired or invalid cache data.

[0146] The web application interface 5 includes the following components: a file uploader 51, a parameter configurator 52, a progress monitor 53, a results visualizer 54, and a history manager 55. The file uploader 51 supports single or batch uploading of scientific literature files in various formats, including XML, PDF, and JSON; the parameter configurator 52 allows users to select the type of information to be extracted; the progress monitor 53 displays the analysis progress in real time; the results visualizer 54 displays the analysis results in a structured manner; and the history manager 55 manages past analysis records and results.

[0147] The batch processing module 6 includes the following components: a task scheduler 61, a parallel processor 62, a result aggregator 63, and a cross-analyzer 64. The task scheduler 61 manages the processing order of multiple files; the parallel processor 62 processes multiple scientific literature files of different formats simultaneously; the result aggregator 63 merges multiple analysis results; and the cross-analyzer 64 identifies common elements and relationships among multiple files.

[0148] The system initialization process is as follows: Upon system startup, the configuration file is first loaded, including parameter settings for each module, API keys, caching strategies, etc.; then, each parser component is initialized, and necessary models and dictionaries are preloaded; next, communication channels between modules are established; finally, the web service is started, awaiting user requests. The initialization phase also includes environment checks, verifying the availability of all dependent libraries and external services; if unavailable, it automatically degrades to a backup solution. The system employs a lazy loading strategy, loading parsers of specific formats only when needed to optimize resource usage and startup speed.

[0149] The system security and privacy protection measures are as follows: (1) User authentication and authorization: Implement role-based access control (RBAC) to ensure that users can only access authorized functions and data; (2) Data encryption: All stored analysis results and user information are encrypted with AES-256, and the transmission process uses the TLS 1.3 protocol to ensure data security; (3) Privacy protection: Provide data desensitization options, which can partially mask sensitive information such as email addresses when extracting them; (4) Audit log: Record all system operations, including user access, file processing and data export; (5) Data retention strategy: Users can set the automatic data deletion period, and the system regularly cleans up expired data; (6) Compliance check: Built-in compliance check function for privacy regulations such as GDPR and CCPA to ensure that marketing activities comply with legal requirements.

[0150] The workflow of this invention is as follows: First, the user uploads one or more scientific literature files (supporting multiple formats such as XML, PDF, JSON, HTML, Word, TXT, EPUB, etc.) through the file uploader 51 of the Web application interactive interface 5, and selects the type of information to be extracted (such as email, experimental instruments, experimental consumables, etc.) through the parameter configurator 52.

[0151] Next, for single-file analysis, the format recognizer 10 of the document parsing and preprocessing module 1 first determines the document type. For XML files, it calls the XML file reader 11 for processing; for PDF files, it calls the PDF parser 11A for OCR processing and structure extraction; for JSON files, it calls the JSON parser 11B for parsing; for HTML files, it calls the HTML parser 11C for processing; for Word documents, it calls the Word document parser 11D for processing; for TXT files, it calls the TXT parser 11E for processing; and for EPUB files, it calls the EPUB parser 11F for processing. Afterwards, the document structure parser 12 parses the document structure based on its respective format characteristics, the metadata extractor 13 extracts the article's metadata, and the content partitioner 14 divides the text into different parts. The parsed structured data is then uniformly passed to the AI-enhanced analysis module 2.

[0152] For multi-file analysis, the task scheduler 61 of the batch processing module 6 arranges the processing order, while the parallel processor 62 processes multiple files of different formats simultaneously. Each file first has its format type determined by the format recognizer 10, and then the corresponding parser is called for processing. The subsequent process is the same as for single-file analysis. This design enables the system to efficiently handle batch tasks consisting of documents in various formats. In the specific implementation, the task scheduler 61 uses a priority queue and resource-aware algorithm to dynamically adjust the processing order and resource allocation based on file size, complexity, and the current system load, maximizing throughput while ensuring processing quality. The parallel processor 62 implements a work-stealing-based load balancing strategy to ensure that computing resources are fully utilized and to avoid situations where the processor is idle while tasks pile up.

[0153] In AI-enhanced analysis module 2, text chunker 21 divides the text into appropriately sized chunks, prompt generator 22 generates specialized prompts based on text type (e.g., method section, results section), API connector 23 sends the prompts and text to a large language model, and response parser 24 parses the returned results into structured data. Technically, text chunking employs an adaptive chunking algorithm based on semantic integrity, considering not only text length limitations but also ensuring the semantic integrity of each chunk, preventing the fragmentation of key information. Prompt generator 22 uses a template library and dynamic combination technology to generate highly specialized prompts for different document sections and extraction targets. For example, when extracting experimental instruments from the method section, the prompt includes professional guidance such as "identify all instruments and equipment used in the experimental steps, paying attention to extracting the brand, model, parameter settings, and purpose of use"; when extracting software tools from the results section, the prompt includes targeted content such as "identify the software packages used for data analysis and visualization, including version numbers and specific parameters." This specialized prompting process improves extraction accuracy by 18.7% compared to general prompts.

[0154] Subsequently, the data cleaning unit 31 of the data processing and integration module removes invalid data, the validator 32 verifies the data format, and the deduplication and merging unit 33 removes duplicate items. In actual implementation, the data cleaning unit 31 uses a combination of rule-based and machine learning methods to identify and process various anomalous data, including incomplete information, format errors, and outliers. The validator 32 applies specific verification rules for different types of data (such as email addresses, instrument names, and chemical substances) to ensure data validity. The deduplication and merging unit 33 not only identifies identical items based on string matching but also identifies information that expresses different meanings but refers to the same entity (such as "PCR instrument" and "polymerase chain reaction instrument") through fuzzy matching and semantic similarity analysis, achieving a merging accuracy of 95.3%.

[0155] Next, the user profile generator 34 transforms the cleaned and verified structured data into multi-dimensional researcher profiles. First, the system establishes basic identity tags based on extracted author and institutional information. Second, it constructs research interest profiles by analyzing document topics, keywords, and citation patterns. Then, it forms technology application preferences based on extracted experimental methods and technical routes. Next, it establishes equipment usage profiles by identifying experimental equipment and consumables mentioned in the literature. Finally, it generates demand prediction models by combining historical data and domain trends. The entire profile construction process employs an incremental learning approach, continuously refining and updating the researcher profiles as more documents are analyzed. When processing multiple documents by the same researcher, the system can automatically merge and update profile information, maintaining the timeliness and completeness of the profiles.

[0156] The Marketing Strategy Recommender 35 generates personalized marketing strategies based on user profiles, combined with a company product database and marketing knowledge base. The recommendation process consists of three stages: product matching, channel selection, and content customization. In the product matching stage, the system selects the most suitable product combination from the product database based on the researcher's technical needs and device usage habits. In the channel selection stage, the system determines the best contact channels (such as email, academic conferences, professional seminars, etc.) based on the researcher's contact method preferences and response history. In the content customization stage, the system generates professional marketing copy and technical materials based on the researcher's professional background and research stage. The entire recommendation process employs an A / B testing mechanism to continuously optimize recommendation effectiveness, resulting in a 37.8% increase in marketing conversion rate compared to traditional methods.

[0157] The processed results are handled by the cache management module 4. The cache key generator 41 generates a unique identifier, and the data storage device 42 saves the results to the local cache for fast future retrieval. The caching system employs a hierarchical storage strategy, storing frequently accessed data in memory and less frequently accessed data on disk. It also implements an adaptive expiration strategy, dynamically adjusting cache retention time based on data access frequency and recent access time. This design reduces system response time by 87% in scenarios with numerous repetitive queries, significantly improving the user experience.

[0158] Finally, the analysis results are transmitted back to the web application interface 5, where they are displayed in a user-friendly manner by the results visualizer 54. Users can view detailed results, download result files, or save them to their history. The interface design employs a responsive layout and modular components, supporting a consistent experience across different devices. The results visualizer 54 implements various visualization methods, including tables, charts, network diagrams, and heatmaps, enabling users to understand the analysis results from different perspectives. The system also provides interactive filtering and sorting functions, allowing users to adjust the result display method as needed.

[0159] For batch processing, the batch processing module 6's result aggregator 63 merges the results from multiple files, and the cross-analyzer 64 analyzes the common elements among the multiple files to generate a comprehensive report. The cross-analyzer 64 implements various advanced analysis algorithms, including graph-based researcher relationship network analysis, topic model-based research hotspot identification, time-series-based research trend analysis, and association rule-based device-consumable relationship mining. These algorithms can discover deep-seated correlations from the analysis results of multiple documents, such as identifying valuable marketing insights like "research teams using a specific brand of PCR instrument often also tend to use the same brand of DNA extraction reagents." The batch processing module 6 also feeds back the analysis results and performance data to the AI-enhanced analysis module 2 and the cache management module 4 via feedback connections to optimize prompting and cache management strategies.

[0160] This invention also implements a complete error handling and anomaly response mechanism. During the document parsing phase, the system can detect and process corrupted files, non-standard formats, and incomplete content, extracting as much useful information as possible through degradation processing strategies. In the AI ​​analysis phase, the system implements a request retry mechanism and a model degradation strategy to ensure that analysis tasks can still be completed even when API connections are unstable or response times out. During the data processing phase, the system marks and records all abnormal data, generating a detailed quality assessment report so that users can understand the reliability of the results. These mechanisms enable the system to exhibit high robustness in real-world application environments; even with 20% of the input documents containing problems, the system can still maintain an overall success rate of over 85%.

[0161] The core advantages of this invention's system are: it supports unified processing of scientific literature in multiple formats, including XML, PDF, and JSON, greatly expanding the range of data sources; it automatically extracts key information from scientific literature through a large-scale language model, including but not limited to author emails, experimental instruments, experimental consumables, chemical substances, software tools, databases, and statistical methods, significantly reducing the workload of manual extraction; simultaneously, its modular design gives the system excellent scalability, allowing the addition of new document format parsers and extraction types as needed; its automatic format recognition function intelligently determines the document type and calls the corresponding processing flow, improving the system's adaptability and user experience; its caching mechanism improves system efficiency and avoids redundant analysis; and its intuitive and user-friendly web interface provides rich interactive functions and result display methods.

[0162] While achieving the basic functions of this invention, other technical approaches can also be used to achieve similar effects. For example, regarding document format processing strategies, in addition to the dedicated parser used in this invention to process documents of different formats separately, a format conversion scheme can also be adopted. That is, all non-XML format documents (such as PDF, JSON, etc.) can be uniformly converted to XML format before subsequent processing. This method can simplify the subsequent processing flow, but may lose some format-specific information during the conversion process. When processing training data, important parts of the documents can be extracted for detailed analysis, or the complete documents can be directly handed over to the AI ​​model for overall processing. Regarding the source of document data, in addition to relying on public literature databases, a proprietary dataset can be built by searching documents locally and using key data regularization matching technology, or relevant research results information can be scraped from the official websites of various universities to build a database. In terms of batch processing mode, the system can use queued processing instead of parallel processing, processing files in priority order. This method is particularly suitable for system environments with limited resources. In addition, regarding the selection of AI models, in addition to using the Gemini API, the system can also be configured to use other large language models such as the OpenAIGPT series, Claude series, or open source models such as LLaMA, to adapt to different scenario requirements and cost considerations. Although these alternative solutions have different implementation paths, they can all meet the basic requirements of the present invention to a certain extent.

[0163] This invention presents a marketing system based on intelligent analysis of multimodal literature data. By constructing a unified multi-format literature processing framework, it successfully solves the data extraction problem of multi-source heterogeneous scientific research literature. This framework adopts a modular parser matrix to support 10 formats such as XML / PDF / JSON and a three-layer format recognition system: file extension → magic number verification → content feature analysis. Combined with a decision tree algorithm, it ensures a format recognition accuracy of 99.8%. Through a document structure parser, it establishes a general document tree model to achieve structural mapping between different formats and standardized JSON output. It also has a built-in fault tolerance mechanism. The system has a high complete parsing rate when processing mixed format documents, significantly expanding the comprehensiveness of data sources. This invention also develops AI model-enhanced multimodal analysis technology to solve the problem of low accuracy in deep information extraction. Furthermore, this invention constructs a five-dimensional profile modeling system and knowledge graph enhancement technology to solve the problem of insufficient depth in scientific research profiles, enabling synonym merging, technology evolution analysis, and demand prediction models.

[0164] The marketing system based on multimodal literature data intelligent analysis of the present invention analyzes heterogeneous literature, extracts semantics through AI-enhanced analysis, constructs a researcher knowledge graph and generates precise marketing strategies. Finally, through caching and batch processing to optimize the performance of the multimodal intelligent marketing system, the present invention achieves in-depth mining and precise marketing of multi-source scientific literature by using a unified multi-format processing framework, AI-enhanced information extraction, multi-dimensional scientific research profiling and efficient parallel analysis technology, which significantly improves the accuracy of information extraction and marketing conversion efficiency.

[0165] The above specific embodiments are used to explain and illustrate the present invention, and are only preferred embodiments of the present invention, not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A marketing system based on intelligent analysis of multi-modal document data, characterized in that: Comprising, The document analysis and preprocessing module uses a format recognizer to identify the format of the input document and uses a format-specific parser for deep content extraction and structural analysis, and then generates a unified document tree structure representation through a document structure parser. The metadata extractor obtains key information, and the content partitioner divides the logical part to achieve efficient and structured preprocessing of heterogeneous scientific literature. The information is transmitted to the AI-enhanced analysis module through the standardized JSON data structure output; The AI-enhanced analysis module receives the JSON data and extracts deep semantic information from the literature through a semantic chunker, a domain-specific prompt generator, a multi-LLMAPI connector, and a response parser, and outputs the structured data in JSON-LD format; The data processing and integration module cleans, verifies, and de-duplicates the structured data in JSON-LD format, and based on the processed data, constructs a researcher knowledge graph containing basic identity, research interest, technology stack, device usage, and demand prediction through a user portrait generator, and generates personalized product combinations and marketing strategies using a marketing strategy recommender, achieving complete conversion from raw data to precise marketing strategies. Cleaning, verification, and structured processing of extraction results; The cache management module connects the data processing and integration module to store and retrieve analysis results and user portraits based on content hash; The batch processing module connects the document analysis and preprocessing module, AI-enhanced analysis module, and cache management module to coordinate parallel processing and cross-correlation analysis of multiple files; The document analysis and preprocessing module, AI-enhanced analysis module, and data processing and integration module are connected in series to form an end-to-end processing pipeline from raw literature to marketing strategies.

2. The marketing system based on intelligent analysis of multi-modal document data as claimed in claim 1, wherein: The document analysis and preprocessing module includes: The format recognizer includes a three-level format recognition mechanism based on file extension, secondary verification based on file header magic number, and final decision based on file content characteristics; The format-specific parser includes an XML file reader, a PDF parser, a JSON parser, an HTML parser, a Word document parser, a TXT parser, and an EPUB parser; The document structure parser generates a unified document tree structure representation; The metadata extractor extracts the title, author, and DOI information of the document; The content partitioner divides the document content according to the academic structure; The document analysis and preprocessing module implements standardized processing of heterogeneous literature through three-level format recognition, multi-parser adaptation, unified structure reconstruction, metadata extraction, and academic content partitioning. 3.The marketing system based on intelligent analysis of multi-modal document data according to claim 2, wherein, The PDF parser includes: A text, image, and table separation unit based on PyMuPDF; A scanned document processing unit integrated with Tesseract-OCR; A layout analysis unit using computer vision algorithms to identify the chapter structure of the document; A mathematical formula recognition and MathML conversion unit; The PDF parser achieves high-precision content and structure extraction of native PDF and scanned PDF through the coordinated work of the above units. 4.The marketing system based on intelligent analysis of multi-modal document data according to claim 1, wherein, The AI-enhanced analysis module includes: A text chunker with a semantic integrity-protected segmentation algorithm; A prompt generator containing a library of at least 200 domain-specific prompt templates; An API connector supporting dynamic switching of multiple large language model APIs; A response parser for converting unstructured responses into JSON-LD format; The AI-enhanced analysis module implements domain knowledge enhancement and standardized analysis through semantic chunking, specialized prompt templates, dynamic scheduling of multiple model APIs, and JSON-LD structured output. 5.The marketing system based on intelligent analysis of multi-modal document data according to claim 1, wherein, The data processing and integration module includes: A data cleaner that implements rule-based and machine learning-based noise filtering; A user portrait generator that constructs a five-dimensional portrait containing basic information, research interests, technology stack, device usage, and demand prediction; A marketing strategy recommender that uses a hybrid recommendation model combining collaborative filtering algorithms and knowledge graph reasoning; The data processing and integration module achieves precise marketing strategies through data cleaning, five-dimensional user portrait construction, and collaborative filtering and knowledge graph fusion recommendations. 6.The marketing system based on intelligent analysis of multi-modal document data according to claim 5, wherein, In the five-dimensional researcher knowledge graph constructed by the user portrait generator: The technology stack dimension includes the researchers' experimental methods, data analysis software, and programming languages; The device usage dimension includes the equipment models, brands, and usage scenarios mentioned in the literature; The demand prediction dimension is based on the researcher's historical technology stack, device usage records, and the evolution trend of their research field, and predicts the required equipment and consumables through a machine learning model. 7.The marketing system based on intelligent analysis of multi-modal document data according to claim 5, wherein, In the marketing strategy recommender: The collaborative filtering algorithm is used to discover similarities with researchers in the knowledge graph, and to calculate the hybrid similarity between the current researcher and other researcher nodes, which includes linear or nonlinear combinations of similarity based on graph topology and similarity based on researcher multi-dimensional attributes; The knowledge graph reasoning is used to infer potential product demand based on the entity association paths between technology, equipment, and consumables.

8. The marketing system based on intelligent analysis of multi-modal document data as claimed in claim 1 wherein, The cache management module includes: A content hash-based cache key generator for generating globally unique identifiers for analysis results; A data storage with LRU and LFU hybrid eviction policy to optimize memory usage; A bloom filter-accelerated search engine to determine whether the requested results do not exist in the cache to prevent cache penetration; The cache management module uses a high-performance hierarchical cache system that generates unique keys based on content hashes, optimizes memory using a hybrid eviction policy, and prevents cache penetration with the help of a bloom filter. 9.The marketing system based on intelligent analysis of multi-modal document data according to claim 1, wherein, The batch processing module includes: A dynamic resource allocation task scheduler that dynamically allocates computing resources based on file size, parsing complexity, and current system load; A parallel processor based on the work-stealing algorithm to achieve load balancing for multiple literature parsing and AI analysis tasks; A cross-analyzer that supports multi-dimensional correlation analysis to perform correlation analysis on multiple literature results after batch processing to identify common research teams, technology hotspots, and equipment usage combination patterns; The batch processing module implements a high-performance batch task processing computing module through dynamic resource scheduling, work-stealing parallel processing, and cross-correlation analysis.

10. The marketing system based on intelligent analysis of multi-modal document data as claimed in claim 1 wherein: The application further comprises a web application interactive interface connected to the cache management module, which provides an operation interface for users to upload documents, configure parameters, visualize analysis results and marketing strategies.

Images