A multimodal content generation system and method combining GEO with SEO dual-channel collaborative optimization.

Abstract

This invention discloses a multimodal content generation system and method that combines GEO and SEO for dual-path collaborative optimization. This addresses the technical problem in existing technologies where information content is difficult for different types of search engines to accurately crawl, resulting in low-quality query results. The system includes: crawling raw information data and generating a preprocessed document; performing entity recognition and relation extraction on the standardized text contained within, and annotating it with structured data to generate a structured document and write it into an SEO article repository; performing semantic expansion and long-tail extension based on the structured document to generate a keyword matrix and rewriting the structured document in a question-and-answer format to obtain a GEO variant set; searching and matching credible standard data related to each text paragraph in the GEO variant set to generate an additional evidence chain summary; obtaining the target content, and adding adapted multimodal materials based on the semantics of the target content to generate a multimodal content body containing the additional evidence chain summary.

Description

Technical Field

[0001] This invention relates to the field of computer information processing technology, and in particular to a multimodal content generation system, method, electronic device, computer storage medium, and computer program product that combines GEO with SEO dual-path collaborative optimization. Background Technology

[0002] Search engines can accurately capture users' core needs, relying on core algorithms to deeply filter, intelligently match, and efficiently retrieve massive amounts of internet data, quickly extracting specific information that meets user requirements and presenting it to users in a clear and intuitive format. With its powerful information aggregation and distribution capabilities, it has become a key hub connecting users and information in the internet ecosystem, and is undoubtedly one of the main traffic entry points.

[0003] Traditional Search Engine Optimization (SEO) focuses on three core implementation dimensions: keyword density control, backlink building, and page structure optimization. It has long been a classic operational model used in the field of internet traffic acquisition. However, with the development of Artificial Intelligence (AI) technology and the rise of generative search technology, the underlying logic of the search ecosystem has shifted from "keyword matching and retrieval" to "semantic understanding and content value utilization." The adaptability shortcomings of traditional SEO have become increasingly apparent, mainly exhibiting the following core defects: The construction of content authority and the layout of keyword strategy are difficult to coordinate and balance. The imbalance between the two can easily lead to the double loss of content professionalism and search adaptability, resulting in large fluctuations in the citation rate of AI search for target content. At the same time, the coverage of long-tail keywords is insufficient in breadth and depth, making it difficult to meet the search needs of users in segmented scenarios. There is a significant disconnect between structured data standards and natural language expression. The content organization format does not meet the core needs of AI information extraction and logical sorting, which makes it impossible for AI to effectively capture core information and build a stable and complete chain of evidence, thus restricting the quality of intelligent answer generation. There are significant deficiencies in multi-dimensional adaptation capabilities and ecosystem operation capabilities. Mobile adaptation and optimization are inadequate, multimodal content compatibility is insufficient, the ecosystem of user-generated content (UGC) and expert-created content has not yet been established, and there is a lack of dynamic monitoring and cross-channel collaborative optimization mechanisms throughout the entire process. It is difficult to achieve continuous iteration and efficiency improvement in the new operational scenario that integrates generative engine optimization (GEO) and traditional search engine optimization (SEO).

[0004] Therefore, there is an urgent need for a content optimization and adaptation solution that can re-integrate and optimize existing information content so that the optimized content can be adapted to both generative search and traditional search, making it easier for search engines to crawl during the retrieval process, improving the visibility of the content, and thus improving the quality of information obtained by users in the search. Summary of the Invention

[0005] The main objective of this invention is to solve the technical problem in the prior art that information content is difficult for search engines to accurately crawl, resulting in poor visibility of actual relevant content and low quality of query results obtained by users.

[0006] The first aspect of this invention provides a multimodal content generation system that combines GEO and SEO for dual-path collaborative optimization, comprising: The data acquisition module is used to configure the content capture workflow and capture raw information data in response to data capture requests. The content understanding module is used to preprocess and recognize the content of the raw information data, and generate a preprocessed document. The SEO optimization module is used to perform entity recognition and relation extraction on the normalized text contained in the preprocessed document, perform structured data annotation based on the extracted entities and relations, generate a structured document, and write it into the SEO article repository. The GEO optimization module is used to perform semantic augmentation and long-tail expansion based on the structured document to generate a keyword matrix; rewrite the structured document in a question-and-answer format according to the keyword matrix to obtain a GEO variant set; find and match credible standard data related to each text paragraph in the GEO variant set to generate an additional evidence chain summary; obtain target content from the SEO article repository and / or the GEO variant set, and add adapted multimodal materials based on the content semantics of the target content to generate a multimodal content body containing the additional evidence chain summary.

[0007] Optionally, in a first implementation of the first aspect of the present invention, the SEO optimization module is further configured to: Based on the keyword matrix, find multimodal content bodies containing additional chain of evidence summaries that are related to the keywords of the structured document; Insert keyword-related multimodal content bodies containing summaries of additional evidence chains into the structured documents included in the SEO article repository, and provide a redirect entry to the page.

[0008] Optionally, in a second implementation of the first aspect of the present invention, the original information data includes traditional search information data and generative search information data; The data acquisition module is specifically used for: Build a workflow platform and configure a content retrieval workflow on the workflow platform; Receive and respond to intelligent content generation requests, and obtain the content generation scope of the intelligent content generation request; Traditional search techniques are used to obtain relevant traditional search results for the content generation range, while generative search techniques are used to obtain generative search results related to the content generation range. The search results that meet the first relevance threshold are selected from the traditional search results and used as traditional search information data. Obtain the referenced content of the generative search results, and filter out the search results whose relevance meets the second relevance threshold from the referenced content as generative search information data.

[0009] Optionally, in a third implementation of the first aspect of the present invention, a promotion routing module is further included, the promotion routing module being specifically used for: Obtain multiple preset promotional information items, and obtain the semantic anchor points or semantic triggering rules of each promotional information item; When the target content contains semantic anchors or when a user's browsing behavior on the multimodal content meets the semantic triggering rules, preset promotional information is inserted into the preset promotional position of the multimodal content.

[0010] Optionally, in a fourth implementation of the first aspect of the present invention, a control experiment module is further included, specifically for: Insert a short traffic-driving link into the preset promotion position of the multimodal content body and inject the embedding parameters; The promotional effect of the promotional information is statistically analyzed based on the aforementioned data collection parameters.

[0011] Optionally, in a fifth implementation of the first aspect of the present invention, a monitoring and optimization module is further included, wherein the monitoring and optimization module is specifically used for: The multimodal content body, which includes a summary of the additional chain of evidence, and the unoptimized raw information data will be published in parallel. The first content quality index of the multimodal content body containing the additional evidence chain summary and the second content quality index of the unoptimized raw information data are obtained respectively. The optimization effect of the multimodal content body containing the summary of the additional evidence chain is calculated based on the first content quality index and the second content quality index. When the optimization effect does not meet the optimization requirements, an optimization write-back instruction is triggered to regenerate the original information data corresponding to the multimodal content body containing the additional evidence chain summary.

[0012] A second aspect of this invention provides a method for generating multimodal content bodies through GEO combined with SEO dual-path collaborative optimization, comprising: Configure the content retrieval workflow to retrieve raw information data in response to data retrieval requests; The original information data is preprocessed and its content is identified to generate a preprocessed document; Entity recognition and relation extraction are performed on the normalized text contained in the preprocessed document. Based on the extracted entities and relations, structured data annotation is performed to generate a structured document and write it into the SEO article repository. Based on the structured document, semantic expansion and long-tail expansion are performed to generate a keyword matrix; The structured document is rewritten in a question-and-answer format based on the keyword matrix to obtain a set of GEO variants; Find and match credible standard data related to each text segment in the GEO variant set to generate an additional chain of evidence summary; The target content is obtained from the SEO article repository and / or the GEO variant set, and multimodal materials with corresponding content semantics are added based on the target content to generate a multimodal content body containing an additional evidence chain summary.

[0013] A third aspect of the present invention provides a multimodal content generation device that combines GEO and SEO dual-path collaborative optimization, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the multimodal content generation device that combines GEO and SEO dual-path collaborative optimization to perform the steps of the above-described multimodal content generation method that combines GEO and SEO dual-path collaborative optimization.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the steps of the above-described multimodal content generation method combining GEO and SEO dual-path collaborative optimization.

[0015] A fifth aspect of the present invention provides a computer program product comprising a computer program / instruction that, when executed by a processor, implements the steps of the multimodal content body generation method as described above, which combines GEO with SEO dual-path collaborative optimization.

[0016] The technical solution provided by this invention includes: a data acquisition module for configuring a content crawling workflow and crawling raw information data in response to data crawling requests; a content understanding module for preprocessing and content recognition of the raw information data to generate a preprocessed document; an SEO optimization module for entity recognition and relation extraction of the normalized text contained in the preprocessed document, performing structured data annotation based on the extracted entities and relations, generating a structured document, and writing it into an SEO article repository; a GEO optimization module for semantic expansion and long-tail expansion based on the structured document to generate a keyword matrix; rewriting the structured document into a question-and-answer format based on the keyword matrix to obtain a GEO variant set; searching and matching credible standard data related to each text paragraph in the GEO variant set to generate an additional evidence chain summary; obtaining target content from the SEO article repository and / or the GEO variant set, adding adapted multimodal materials based on the content semantics of the target content, and generating a multimodal content body containing an additional evidence chain summary.

[0017] This invention enables content identification and structuring of existing information, and long-tail expansion based on keywords. It re-optimizes and integrates relevant information, generating optimized content from both generative and traditional search perspectives. This allows the optimized information to be more easily crawled by different search engines, improving the visibility of important information and the quality of search results for users. Furthermore, the method, electronic device, computer-readable storage medium, and computer program product provided by this invention also solve the corresponding technical problems. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of the processing steps of the first embodiment of the multimodal content body generation method combining GEO and SEO dual-path collaborative optimization in this invention. Figure 2 This is a schematic diagram of the processing steps of the second embodiment of the multimodal content body generation method combining GEO and SEO dual-path collaborative optimization in this invention. Figure 3 This is a flowchart illustrating a specific example of the second embodiment of the multimodal content generation method combining GEO and SEO dual-path collaborative optimization in this invention. Figure 4 This is a schematic diagram of the architecture of an embodiment of the multimodal content generation system that combines GEO with SEO dual-path collaborative optimization in this invention. Figure 5 This is a schematic diagram of another embodiment of the multimodal content generation system that combines GEO with SEO dual-path collaborative optimization in this invention. Detailed Implementation

[0019] Exemplary embodiments of the invention will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limiting the invention to the embodiments set forth herein. Rather, these exemplary embodiments are provided to make the invention more comprehensive and complete, and to facilitate a full communication of the inventive concept to those skilled in the art. The same reference numerals in the drawings denote the same or similar elements, components, or parts, and therefore repeated descriptions of them will be omitted.

[0020] Subject to the technical concept of this invention, the features, structures, characteristics or other details described in a particular embodiment may be combined in one or more other embodiments in a suitable manner.

[0021] In the description of specific embodiments, the features, structures, characteristics, or other details described in this invention are intended to enable those skilled in the art to fully understand the embodiments. However, it is not excluded that those skilled in the art can practice the technical solutions of this invention without one or more of the specific features, structures, characteristics, or other details.

[0022] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0023] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0024] The terms “and / or” or “and / or” include all combinations of any one or more of the listed items.

[0025] Please see Figure 1 The first embodiment of the multimodal content generation method combining GEO and SEO dual-path collaborative optimization in this invention is as follows: In this embodiment, GEO refers to Generative Engine Optimization, and SEO refers to Search Engine Optimization. The method provided in this embodiment can achieve GEO-SEO dual-path collaborative optimization to generate multimodal content. Specific details include: S101. Configure the content retrieval workflow to retrieve raw information data in response to data retrieval requests; It is understood that the executing entity of this invention can be a multimodal content generation system that combines GEO with SEO for dual-path collaborative optimization, or it can be a terminal or a server; the specific implementation is not limited here. This embodiment of the invention will be described using a server as an example.

[0026] In this embodiment, a workflow platform needs to be pre-built, and a content scraping workflow needs to be configured on the workflow platform. When performing specific content scraping, the system receives and responds to intelligent content generation requests, and obtains the content generation scope of the intelligent content generation request. The content generation scope can be defined by the target domain, information source, and information publication time. Each time content optimization is performed, the configured content scraping workflow can be invoked to scrape the raw information data (Raw Record) within the content generation scope that needs optimization.

[0027] In this embodiment, the raw information data includes generative search information data and traditional search information data.

[0028] Specifically, the process of crawling raw information data includes: using generative search technology, crawling generative search results constructed by a generative model within the content optimization scope; filtering the referenced content of the generative search results; and selecting generative search results with high relevance to the content generation scope from the referenced content to obtain generative search information data. Alternatively, using traditional search technology, obtaining relevant traditional search results within the content generation scope; and selecting traditional search results with high relevance to the content generation scope from the traditional search results.

[0029] In one specific implementation, traditional search results can be obtained by capturing keywords from the content generated within the scope of the content, or by obtaining recent hot search terms from commonly used traditional search engines.

[0030] In one specific implementation, a relevance threshold can be preset when determining relevance, such as setting a first relevance threshold for traditional search results and a second relevance threshold for generative search results. During the filtering process, if the relevance of some traditional search results meets the first relevance threshold, they are treated as traditional search information data, and if the relevance of some generative search results meets the first relevance threshold, they are treated as generative search information data.

[0031] S102. Perform preprocessing and content recognition on the raw information data to generate a preprocessed document; After obtaining the raw information data, this step can be performed by calling an intelligent text processing model for processing and content recognition. In this embodiment, the intelligent text processing model can consist of multiple agents built based on a large language model, forming an Agent workflow, which is then invoked to cooperate in implementation.

[0032] Specifically, the preprocessing steps for the raw information data in this process include, but are not limited to, text content denoising, text formatting, multilingual translation, and watermark removal.

[0033] Following preprocessing, the process also includes calling the aforementioned intelligent text processing model to identify the semantic content of the original information data, while simultaneously refining and rewriting it to generate a preprocessed document. This preprocessed document contains standardized text, topic tags, content summaries, images, and core keywords.

[0034] S103. Perform entity recognition and relation extraction on the normalized text contained in the preprocessed document, perform structured data annotation based on the extracted entities and relations, generate a structured document, and write it into the SEO article repository. After obtaining the preprocessed document, the canonical text it contains is extracted, and an intelligent text processing model is invoked for further semantic understanding to identify the entities and relationships between them, and then extract the entities and relationships. Based on the content type of the canonical text, a tag editor is invoked to annotate it according to the entities and relationships, thereby converting the canonical text into a structured document.

[0035] When the original information data obtained in this embodiment is traditional search information data and needs to be optimized for traditional search engines, the structured document is directly called to optimize the expression of traditional search, so that the optimized and polished document content is easier for traditional search engines to crawl and write into the SEO article repository.

[0036] In this embodiment, the SEO article repository can be used to store SEO optimization articles, and it also serves as a transit database, preserving structured documents and related evidence chain summary information. When it is necessary to build GEO variants later, relevant information can be obtained from the SEO article repository as the basis for generative optimization.

[0037] S104. Based on the structured document, perform semantic expansion and long-tail expansion to generate a keyword matrix; In this embodiment, in order to enhance the search scope and enable users to obtain relevant knowledge to the maximum extent, it is also necessary to use long-tail expansion to enable the solution to cover synonymous content and variant information in different scenarios.

[0038] Specifically, after obtaining the structured document in the aforementioned steps, an intelligent text processing model is invoked to perform semantic expansion and long-tail expansion based on the standardized text, topic tags, content summaries, images, and core keywords contained in the structured document. For example, core keywords in the structured document are extracted as seeds to expand keywords; and the expanded keyword content is then processed into a question-and-answer format to obtain a keyword matrix.

[0039] When expanding keywords, the following methods are used: intelligent text processing models are employed to expand long-tail keywords from multiple perspectives such as user needs, region, season, and mindset, resulting in a long-tail keyword matrix.

[0040] S105. Rewrite the structured document using a question-and-answer format based on the keyword matrix to obtain a set of GEO variants; The structured document is rewritten using a question-and-answer format based on the generated long-tail keyword matrix. The optimized and adapted information content is saved in the form of questions and answers, resulting in multiple GEO variants. These GEO variants are then written into a GEO variant set for storage.

[0041] The question-and-answer reconstructed set refers to a collection of knowledge segments organized into a series of question-and-answer pairs to better adapt to the answer generation mode of generative search.

[0042] S106. Find and match credible standard data related to each text segment in the GEO variant set, and generate an additional chain of evidence summary; This step also includes pre-building an authoritative standards database. After obtaining the structured document, the process also includes searching and matching credible standard data related to each text paragraph in the pre-built authoritative standards database to generate an authoritative source index. Based on the authoritative source index, additional content evidence information is added to the structured document to generate an additional evidence chain summary.

[0043] S107. Obtain the target content from the SEO article repository and / or GEO variant set, add adapted multimodal materials based on the content semantics of the target content, and generate a multimodal content body containing an additional evidence chain summary.

[0044] When it is necessary to generate and publish multimodal content containing summaries of additional evidence chains, the target content is retrieved from the SEO article repository or GEO variant set according to the generation requirements. An intelligent text processing model is invoked to understand the semantics of the target content, and based on this semantics, appropriate multimodal materials such as images, text, and videos are generated and appended to the target content to generate multimodal content containing summaries of additional evidence chains.

[0045] In one specific implementation, the method further includes obtaining terminal information of different display terminals, obtaining content design style based on the terminal information, and adjusting the display effect of the multimodal content on the page according to the design style.

[0046] Based on the above optimization scheme, when users use the search engine, the optimized multimodal content is more easily captured by both generative search and traditional search engines.

[0047] The method provided in this embodiment of the invention can combine authoritative content citation, structured annotation, and question-based organization to achieve dual-channel content optimization, thereby improving the visibility of multimodal content in generative retrieval engines and traditional search engines, and providing users with information content that better meets their retrieval needs.

[0048] Please refer to Figure 2 and Figure 3 The second embodiment of the multimodal content generation method combining GEO and SEO dual-path collaborative optimization in this invention is as follows: In this embodiment, GEO refers to Generative Engine Optimization, and SEO refers to Search Engine Optimization. The method provided in this embodiment can achieve GEO-SEO dual-path collaborative optimization to generate multimodal content. Specific details include: S201. Configure the content retrieval workflow to retrieve raw information data in response to data retrieval requests; It is understood that the executing entity of this invention can be a multimodal content generation system that combines GEO with SEO for dual-path collaborative optimization, or it can be a terminal or a server; the specific implementation is not limited here. This embodiment of the invention uses a server as the executing entity for example. Specifically, the solution in this embodiment can achieve optimization effects through both Generative Engine Optimization (GEO) and traditional Search Engine Optimization (SEO).

[0049] Upon receiving a content optimization and adaptation request, the server first obtains the scope of content to be optimized. This scope can be defined by the target domain, information source, and publication time. Each time content optimization and adaptation is performed, the server retrieves the original information data within that scope that needs optimization.

[0050] When crawling raw information data, multiple sources can be used. Raw information data includes generative search information data and traditional search information data. Generative search information data can be obtained by crawling information content generated by intelligent text processing models within the content optimization scope and filtering out related articles cited by that content. Traditional search information data can be obtained by searching for related information data using preset keyword ranges. Specifically, when crawling cited articles provided by intelligent text processing models in generative search engines, the article titles can be crawled to find generative information related to the content optimization scope, and the core fragments of the top three cited articles can be obtained, while recording the source, timestamp, and evidence pointers of the cited articles; these core fragments are then used as generative search information data. When crawling related articles provided by traditional search engines, content information related to the keywords within each content optimization scope can be searched in the traditional engine based on keyword sorting, and this content information is then used as traditional search information data.

[0051] In one specific implementation, the input to this content crawling workflow can be a seed keyword list, a scheduled task, and hotspot monitoring information; wherein, the hotspot monitoring information includes RSS (Really Simple Syndication) information and current search trends. The content crawling workflow can be triggered by a schedule or by keywords, and the raw information data is obtained after crawling. While the content crawling workflow is running, crawling metrics are also monitored, including the number of crawls per day, duplication rate, crawling success rate, and crawling latency, in order to obtain the working status of the content crawling workflow based on these metrics.

[0052] The generative search engine described in this embodiment is a search engine built by integrating intelligent text processing model and retrieval-augmented generation (RAG) technology. It can understand and integrate the retrieved information through large language model (LLM) technology to generate fluent answers, rather than directly providing the retrieved information itself. The traditional search engine refers to a search tool that is based on keyword matching and provides the retrieved information content to the user.

[0053] S202. Perform preprocessing and content recognition on the raw information data to generate a preprocessed document; After acquiring the raw information data, the raw information data is stored in a text queue. This text queue can be constructed using a message bus, such as a Kafka queue. When the raw information data is available in the text queue, preprocessing and content recognition steps are triggered.

[0054] First, the text and other media files contained in the original information data are preprocessed, including but not limited to noise reduction, formatting, translation, and watermark removal. Then, an intelligent text processing model is called to identify the semantics of the original information data, while polishing and rewriting it to generate a preprocessed document. The preprocessed document includes canonical text, topic tags, content summary, watermark-removed images, and core keywords.

[0055] In this embodiment, the intelligent text processing model can be composed of multiple agents built based on a large language model to form an Agent workflow, and the Agent workflow is invoked to work together to achieve the desired results.

[0056] In this embodiment, when acquiring the original information data, it also includes recording multi-dimensional features such as search terms, clickstream, dwell time, device, and region of the original information data, preferably using 12-dimensional feature data; when performing semantic recognition, it can output explicit or latent intent tags, as well as the confidence level of the output tags.

[0057] In a preferred embodiment, this step involves performing preprocessing and content recognition. While generating the preprocessed document, the processing metrics for preprocessing and content recognition are monitored, including estimated translation accuracy, summary length distribution, and watermark removal success rate, in order to obtain the working status of preprocessing and content recognition in this step.

[0058] S203. Perform entity recognition and relation extraction on the normalized text contained in the preprocessed document, perform structured data annotation based on the extracted entities and relations, generate a structured document, and write it into the SEO article repository. After obtaining the preprocessed document, its document type can be determined based on an intelligent text processing model. These document types include, but are not limited to, FAQs, How-To instructions, product information, and articles.

[0059] After obtaining the document type, based on the different characteristics of the document type, the intelligent text processing model is invoked to perform further semantic understanding on the normalized text contained in the preprocessed document obtained in the previous steps, identify the entities contained therein and the relationships between entities, extract the entities and relationships, and invoke the corresponding tag editor based on the document type to annotate the extracted entities and relationships, thereby converting the preprocessed document into a structured document, and saving the structured document to the search knowledge base.

[0060] In one specific implementation, when converting the preprocessed document into a structured document, the Schema field is used to annotate the information in the preprocessed document. This step is automatically triggered after the preprocessed document is obtained in the previous step. While executing, it also monitors structured metrics such as annotation coverage, entity extraction F1 (harmonic mean of precision and recall), and data entry speed.

[0061] Once the structured document is obtained, an intelligent text processing model can be used to optimize the structured document for traditional search engine results, such as text polishing and word choice adjustment. The optimized content is then written into the SEO article repository to increase the probability of the content being found by traditional search engines.

[0062] In this embodiment, the SEO article repository can be used to store traditional search optimization content generated based on structured documents. It also serves as an intermediary database, preserving structured documents and related evidence chain summary information. When GEO variants need to be built subsequently, relevant information can be obtained from the SEO article repository as the basis for generative optimization.

[0063] S204. Based on the structured document, perform semantic expansion and long-tail expansion to generate a keyword matrix; In this embodiment, in order to enhance the search scope and enable users to obtain relevant knowledge to the maximum extent, it is also necessary to use semantic expansion and long-tail expansion to enable the solution to cover synonymous content and variant information in different scenarios.

[0064] Specifically, after obtaining the structured document in the aforementioned steps, an intelligent text processing model is invoked to perform semantic expansion and long-tail expansion based on the standardized text, topic tags, content summaries, images, and core keywords contained in the structured document. For example, core keywords in the structured document are extracted as seeds to expand keywords; and the expanded keyword content is then processed into a question-and-answer format to obtain a keyword matrix.

[0065] When expanding keywords, an intelligent text processing model is used to expand the keyword base from multiple perspectives such as user needs, region, season, and mindset, resulting in a long-tail keyword matrix. For each core keyword theme, preferably, 45 variations and 32 extension dimensions can be maintained.

[0066] S205. Rewrite the structured document using a question-and-answer format based on the keyword matrix to obtain a set of GEO variants; After obtaining the long-tail keyword matrix, the structured document is rewritten in a question-and-answer format based on the content of various long-tail keywords in the matrix, resulting in multiple GEO variants, which are then stored in a centralized GEO variants database.

[0067] Steps S204 and S205 can be triggered periodically after the structured document is successfully written to the SEO article repository. During the question-and-answer rewriting, the number of generated variations and long-tail keyword information will also be monitored to calculate the long-tail coverage rate in order to quantify the execution effect of steps S204 and S205.

[0068] In one specific implementation, after generating the actual multimodal content body containing an additional chain of evidence summary in this embodiment, the number of times the multimodal content body is crawled by various different search engines is also monitored to obtain the effective generative search optimization variant click-through rate in order to evaluate the execution effect of semantic augmentation and long-tail expansion in step S204.

[0069] In another specific implementation, the method further includes bidirectional mapping between the extended long-tail keywords and the structured data annotations in the aforementioned steps. Through the mapping tags of the structured documents, mutual indexing is achieved between multimodal content obtained based on Generative Search Optimization (GEO) and traditional search optimization (SEO) content. The traditional SEO content can be stored in an SEO article repository. For example, the method involves finding keywords related to each piece of SEO content; based on the keyword matrix, finding multimodal content related to the keywords of the SEO content; and inserting a link to the page containing the keyword-related multimodal content within the SEO content, so that when users view related SEO content through a search engine, they can directly view the information content of the related multimodal content through the link.

[0070] S206. Find and match credible standard data related to each text segment in the GEO variant set, and generate an additional chain of evidence summary; Specifically, this implementation pre-constructs an authoritative standards database, which contains relevant industry standards, white papers, and reports—authoritative rules within the industry—to obtain credible sources of content. After acquiring the structured document, based on the previously obtained topic tags, content summaries, and core keywords, credible standard data related to each text paragraph in the structured document is searched and matched in the authoritative standards database to generate an authoritative source index. Then, based on the authoritative source index, content evidence information is added to multiple GEO variants included in the GEO variant set.

[0071] In one specific implementation, this step also includes monitoring information such as the proportion of paragraphs enhanced with authority, the distribution of evidence credibility, and the audit pass rate, in order to monitor the authoritative citation effect of the evidence chain summary.

[0072] S207. Obtain the target content from the SEO article repository or GEO variant set; When content is approved or the scheduled publication time arrives, optimized documents from the SEO article repository or GEO variant collection are retrieved as targeted content for delivery.

[0073] S208. Based on the semantic adaptation of the target content, generate a multimodal content body containing an additional evidence chain summary; When generating specific multimodal content based on target content, this includes generating multimodal supplementary information based on optimized content data, and displaying information versions differentiated according to the terminal (such as mobile or computer) of the user who submitted the search request or regional information, such as providing images, long bar supplementary information maps, and optional video keyframe summaries.

[0074] S209. Obtain multiple preset promotional information and obtain the semantic anchor point or semantic triggering rule of each promotional information; S210. When the target content contains semantic anchors or when the user's browsing behavior on the multimodal content body meets the semantic triggering rules, insert preset promotional information into the preset promotional position of the multimodal content body. In this embodiment, the additional information also includes extended content resources such as promotional information, which can be added to the corresponding positions through semantic triggering.

[0075] In one specific implementation, attaching promotional information includes: acquiring multiple preset promotional information sets and obtaining the semantic anchor points or semantic triggering rules for each set; when the target content contains semantic anchor points or the user's browsing behavior on the multimodal content body meets the semantic triggering rules, inserting the preset promotional information into a preset promotional position in the multimodal content body. For example, this can be achieved by inserting promotional information text into native paragraphs, inserting promotional information banners on the page, or attaching promotional information overlays to videos; for example, matching corresponding promotional information when the content contains semantic descriptions such as "feature engineering / model deployment"; and jointly determining the intensity and display format of the promotional information based on the current user's reading time, scrolling depth, and historical preferences.

[0076] Furthermore, after generating the optimized and adapted multimodal content, the content is converted into HTML (Hypertext Markup Language) and other formats. During display, differentiated presentations are generated based on client type (e.g., mobile or desktop) and geographic location (e.g., different cities from which users originate). Specific adjustments may include font size, image density, and module order. The optimized and adapted content pages and multimodal materials are then pushed to official accounts, information websites, or social media platforms for distribution. The client is responsible for rendering and event reporting.

[0077] In a preferred embodiment, when a user belongs to a group sensitive to promotional content (such as the elderly, students, etc.), the promotional information or sensitive statements can be automatically downgraded or blocked, and prompts such as "Be cautious when borrowing" or "Annualized interest rate information" can be automatically added.

[0078] S211. Insert a short link to drive traffic into the preset promotion position of the multimodal content body, and inject the tracking parameters. Calculate the promotion effect of the promotion information based on the tracking parameters. Furthermore, this embodiment also includes a monitoring and optimization function, which includes monitoring and optimizing the effectiveness of promoted content and monitoring and optimizing generated content. When monitoring and optimizing the effectiveness of promoted content, a short referral link is inserted into the preset promotion position of the multimodal content body, and tracking parameters are injected; the promotion effect of the promoted information is statistically analyzed based on the tracking parameters.

[0079] S212. Publish the multimodal content body containing the additional evidence chain summary and the unoptimized original information data in parallel, conduct a control experiment, obtain the optimization effect of the multimodal content body, and when the optimization effect of the multimodal content body does not meet the optimization requirements, re-optimize and generate the content.

[0080] When monitoring and optimizing the effect of generated content, A / B testing can be used to compare and determine which steps require content optimization.

[0081] For example, after generating the multimodal content body containing the additional evidence chain summary, the process may include: publishing the multimodal content body containing the additional evidence chain summary and the unoptimized original information data in parallel; obtaining a first content quality index of the multimodal content body containing the additional evidence chain summary and a second content quality index of the unoptimized original information data, respectively; calculating the optimization effect of the multimodal content body containing the additional evidence chain summary based on the first content quality index and the second content quality index; and triggering an optimization write-back instruction when the optimization effect does not meet the optimization requirements, re-optimizing and generating the corresponding original information data of the multimodal content body containing the additional evidence chain summary, that is, re-executing the multimodal content body generation step according to the original information data captured in S201 to obtain a multimodal content body whose optimization effect meets the optimization requirements.

[0082] In other words, the solution in this embodiment can publish original content, traditional search-optimized content, and generative search-optimized content based on multimodal content in parallel, collect and detect exposure metric data for each, and adjust the aforementioned optimization steps based on the exposure metric data, automating retraining to achieve self-optimization of the solution. Specifically, the published multimodal content and the traditional search-optimized version of the content are monitored and compared to obtain the exposure metrics of the multimodal content and the exposure metrics and conversion rate data of the search-optimized version, respectively; content whose exposure metrics do not meet the exposure requirements is re-optimized and adapted; and the additional multimodal materials are adjusted or the multimodal intelligent agent is regenerated based on the conversion rate data. The exposure metrics include, but are not limited to, page views, click-through rate, conversion rate, number of times the artificial intelligence model is cited, search engine result page ranking, and number of social media reposts.

[0083] In one specific implementation, this embodiment can also perform inter-group comparisons of content with and without ads. Specific indicators include page views, click-through rate, conversion rate, number of times the artificial intelligence model is referenced, etc., and the parameters in the optimization effect-driven template, keyword matrix, and threshold write-back process are adjusted and optimized based on the indicator information to improve the effectiveness of the optimization and adaptation scheme.

[0084] In addition, when the number of citations of the artificial intelligence model cannot be obtained, the question-and-answer recall rate and brand keyword association coverage rate of external search can be used as exposure indicators.

[0085] In one specific implementation, this step also includes using a data sandbox combined with transparent traceability reports to regularly publish authoritative scores and risk audit results. Strategies are then regularly adjusted and optimized based on these scores and results, and manual review of the entire process is also supported. The system can obtain operation logs based on full-chain operational indicators monitored at each step, before and after content publication, or when high-risk events are detected. Based on the content evidence chain and compliance strategy library, it outputs audit reports, violation alerts, and regulatory adjustment suggestions.

[0086] Please continue reading. Figure 3 The following example uses information related to Artificial Intelligence (AI) as the scope of content optimization to illustrate the optimization and adaptation method in this embodiment: First, a workflow platform (such as Dify or the self-developed Deepbank-flow) is pre-built, and a web crawler workflow is configured on this platform. Raw information data is obtained through the configured web crawler workflow. Specifically, the web crawler workflow crawls publicly available content that does not involve sensitive information or require authorization.

[0087] Specifically, the process involves first acquiring news headlines related to AI from both domestic and international sources, then using web crawlers to extract articles cited by AI models (or intelligent text processing models) within these news articles, storing the content as raw news data. This data is then preprocessed through translation, polishing, and rewriting, and summaries are generated. Further optimization is then performed from multiple perspectives, including content quality, structure and semantics, multimodality and dynamics, and user intent and context, resulting in optimized resource content. This optimized content is then converted to HTML format for storage and updates, achieving content optimization and adaptation. This optimized content is more easily crawled by generative search engines, reducing the loss of relevant content.

[0088] Please continue reading. Figure 3 In another implementation, it also includes obtaining article or search keyword rankings from traditional search engines within a relevant scope, and obtaining raw information data based on the content of these articles or related keywords. After refining and rewriting the initial and content data, it can be converted into HTML format for storage and updates, thereby achieving content optimization and adaptation.

[0089] For further information, please refer to [link / reference]. Figure 3 In this embodiment, information-related images can also be automatically generated based on the image generation model, and the watermark-removed information-related images can be stored together with the relevant optimized content.

[0090] Please continue reading. Figure 3 The process involves not only crawling raw information content but also acquiring and storing AI monetization guidelines. These guidelines can be used to guide subsequent implementation of additional advertising strategies. Specifically, after obtaining the optimized content, the Agent workflow can be invoked to analyze the stored optimized articles. Based on the analysis results, scenario-based native embedding, dynamic ad adaptation, optimization of hidden traffic entry points, and compliance enhancements can be implemented through ad routing. During ad embedding, the process also includes generating ad images using an image generation model and generating corresponding short referral links.

[0091] The method provided in this invention can perform content recognition and structuring processing on existing information content. Through steps such as constructing rewrite sets and expanding long-tail keywords, it achieves automatic generation of multimodal content, generating AI-generated information content that is more easily crawled by generative search engines and traditional search engines, thus achieving dual-path optimization. The optimized and adapted information content, while satisfying the crawling requirements of traditional search engines, also increases the likelihood of being cited and crawled by generative search engines, improves the visibility of important information in both paths, optimizes and improves the quality of query results obtained by users, and provides users with information content that better meets their search needs, thereby improving content production efficiency.

[0092] The above describes the multimodal content generation method combining GEO and SEO dual-path collaborative optimization in the embodiments of the present invention. The following describes the multimodal content generation system combining GEO and SEO dual-path collaborative optimization in the embodiments of the present invention. Please refer to [link / reference]. Figure 4 An embodiment of the multimodal content generation system combining GEO and SEO dual-path collaborative optimization in this invention is as follows: The multimodal content generation system combining GEO and SEO dual-path collaborative optimization described in this embodiment includes: a data acquisition and input layer 401, a data cleaning and storage layer 402, a content generation and modal layer 403, a GEO optimization and advertising decision layer 404, and a distribution and closed-loop layer 405. These are described in detail below: (1) Data Acquisition and Input Layer 401: A crawler workflow is configured in the data acquisition and input layer 401 to crawl data information from multiple sources based on the input source.

[0093] Specifically, relevant traditional search information data is crawled from competitor SEO articles or search keywords, and keywords are obtained. These keywords are stored in the keyword storage database in the data cleaning and storage layer 402, and traditional search information data is stored in the SEO article storage repository in the data cleaning and storage layer 402. The crawled AI monetization guide and externally cited articles crawled based on domestic and foreign AI information titles are temporarily stored in the information database of the data cleaning and storage layer 402.

[0094] (2) Data cleaning and storage layer 402: After preprocessing, denoising, multilingual translation, structured annotation, and entity extraction of externally cited articles obtained from the information database, the knowledge is deposited into the SEO article repository; traditional search information data is stored in the SEO article repository in the data cleaning and storage layer 402.

[0095] (3) Content generation and modal layer 403: The content generation and modal layer 403 is equipped with an intelligent model inference core. The intelligent model inference core acquires the content stored in the information database and SEO article repository in the data cleaning and storage layer 402. In the text stream unit, it performs information translation, summary generation, and expansion of the long-tail keyword matrix. Based on the expanded long-tail keywords, it generates a variant set, polishes and rewrites it, binds additional evidence chain summaries, and enhances authority (including binding industry standards, authoritative industry information, and industry standard white papers) to generate authoritative additional evidence chain summaries.

[0096] In the visual flow unit of the content generation and modal layer 403, keywords stored in the keyword repository in the data cleaning and storage layer 402 are obtained. Based on the keywords, the image generation model deployed in the visual flow unit is called to generate images, as well as perform operations such as watermark removal and formatting.

[0097] (4) GEO optimization and advertising decision-making level 404: The GEO optimization engine is deployed in the GEO optimization and advertising decision layer 404. The GEO variants obtained in the text stream of the content generation and modal layer 403 are processed by the GEO optimization engine and sent to the optimization module to perform multiple steps such as content quality, structuring and semantic optimization, multimodal dynamic optimization and user intent scenario optimization. The optimized content is then used to generate an article, converted into HTML format and stored and updated.

[0098] After acquiring the image in the content generation and modal layer 403 image stream, the image is added as additional multimodal information to the HTML-formatted article.

[0099] The GEO optimization and advertising decision layer 404 also deploys an LLM content analysis module, which performs content analysis based on HTML format articles, sends the results to the advertising routing unit to generate corresponding advertising information and performs native embedding, and builds traffic-driving short links and tracking points.

[0100] (5) Distribution and closed-loop layer 405: The distribution and closed-loop layer 405 contains advertising image and content synthesis units. The synthesized units with added advertising information are sent to the network-wide distribution matrix and distributed to public accounts, websites, and social media.

[0101] The distribution and closed-loop layer 405 also deploys a monitoring and write-back module, which includes a security and compliance data sandbox that can be used for isolation testing before specific distribution monitoring; it also includes an A / B control test unit and a monitoring and write-back unit, which can be used to detect information such as write-back reference rate and conversion rate.

[0102] After obtaining information such as write-back reference rate and conversion rate in the monitoring and write-back unit in the distribution and closed-loop layer 405, feedback-driven retraining can be performed on the intelligent model inference core in the content generation and modality layer 403, the advertising routing unit in the GEO optimization and advertising decision layer 404, and the preset settings of the crawler workflow in the data collection and input layer 401 can be adjusted based on this information.

[0103] Furthermore, each level and the modules and units contained in this embodiment can be used to execute the multimodal content body generation method that combines GEO with SEO dual-path collaborative optimization. The specific steps for executing the method can be found in the content of the aforementioned method embodiment, and will not be repeated here.

[0104] Please see Figure 5 Another embodiment of the multimodal content generation system combining GEO and SEO dual-path collaborative optimization in this invention includes: Data acquisition module 501 is used to configure the content capture workflow and capture raw information data in response to data capture requests; Content understanding module 502 is used to preprocess and recognize the content of the original information data and generate a preprocessed document; The SEO optimization module 503 is used to perform entity recognition and relation extraction on the normalized text contained in the preprocessed document, perform structured data annotation based on the extracted entities and relations, generate a structured document, and write it into the SEO article repository. GEO optimization module 504 is used to perform semantic expansion and long-tail expansion based on the structured document to generate a keyword matrix; rewrite the structured document in a question-and-answer format according to the keyword matrix to obtain a GEO variant set; find and match credible standard data related to each text paragraph in the GEO variant set to generate an additional evidence chain summary; obtain target content from the SEO article repository and / or the GEO variant set, and add adapted multimodal materials based on the content semantics of the target content to generate a multimodal content body containing the additional evidence chain summary.

[0105] The system provided in this embodiment of the invention can combine authoritative content citation, structured annotation, and question-based organization to achieve dual-channel content optimization, thereby improving the visibility of multimodal content in generative retrieval engines and traditional search engines, and providing users with information content that better meets their retrieval needs.

[0106] In another embodiment of this application, the SEO optimization module 503 is further configured to: Based on the keyword matrix, find multimodal content bodies containing additional chain of evidence summaries that are related to the keywords of the structured document; Insert keyword-related multimodal content bodies containing summaries of additional evidence chains into the structured documents included in the SEO article repository, and provide a redirect entry to the page.

[0107] In another embodiment of this application, the original information data includes traditional search information data and generative search information data; The data acquisition module 501 is specifically used for: Build a workflow platform and configure a content retrieval workflow on the workflow platform; Receive and respond to intelligent content generation requests, and obtain the content generation scope of the intelligent content generation request; Traditional search techniques are used to obtain relevant traditional search results for the content generation range, while generative search techniques are used to obtain generative search results related to the content generation range. The search results that meet the first relevance threshold are selected from the traditional search results and used as traditional search information data. Obtain the referenced content of the generative search results, and filter out the search results whose relevance meets the second relevance threshold from the referenced content as generative search information data.

[0108] In another embodiment of this application, a promotion routing module is further included, the promotion routing module being specifically used for: Obtain multiple preset promotional information items, and obtain the semantic anchor points or semantic triggering rules of each promotional information item; When the target content contains semantic anchors or when a user's browsing behavior on the multimodal content meets the semantic triggering rules, preset promotional information is inserted into the preset promotional position of the multimodal content.

[0109] In another embodiment of this application, a control experiment module is also included, specifically used for: Insert a short traffic-driving link into the preset promotion position of the multimodal content body and inject the embedding parameters; The promotional effect of the promotional information is statistically analyzed based on the aforementioned data collection parameters.

[0110] In another embodiment of this application, a monitoring optimization module is further included, which is specifically used for: The multimodal content body, which includes a summary of the additional chain of evidence, and the unoptimized raw information data will be published in parallel. The first content quality index of the multimodal content body containing the additional evidence chain summary and the second content quality index of the unoptimized raw information data are obtained respectively. The optimization effect of the multimodal content body containing the summary of the additional evidence chain is calculated based on the first content quality index and the second content quality index. When the optimization effect does not meet the optimization requirements, an optimization write-back instruction is triggered to regenerate the original information data corresponding to the multimodal content body containing the additional evidence chain summary.

[0111] The system provided in this embodiment of the invention can perform content recognition and structuring processing on existing information content. Through steps such as constructing rewritten sets and long-tail expanded keywords, it generates AI-generated information content that is more easily crawled by generative search and traditional search engines. By generating optimized and adapted information content, the visibility of important information in both channels is improved, the quality of search results obtained by users is optimized, and information content that better meets users' retrieval needs is provided.

[0112] Based on the same inventive concept, embodiments of this specification also provide an electronic device for generating multimodal content by combining GEO with SEO dual-path collaborative optimization. The electronic device for generating multimodal content by combining GEO with SEO dual-path collaborative optimization includes: a memory and at least one processor, wherein the memory stores instructions. The at least one processor invokes the instructions in the memory to cause the GEO-SEO dual-path collaborative optimization multimodal content generation device to perform the steps of the GEO-SEO dual-path collaborative optimization multimodal content generation method as described in the foregoing method embodiments.

[0113] Based on this, the present invention also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implements the multimodal content generation method for GEO combined with SEO dual-path collaborative optimization as described in any of the above embodiments. Furthermore, it implements... Figure 1 , Figure 2 or Figure 3 The computer program for the method shown can be stored on one or more computer-readable media.

[0114] It will be readily understood by those skilled in the art that the exemplary embodiments described in this invention can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to embodiments of the present invention can be embodied in the form of a program product, which can be stored in a computer-readable storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, or network device, etc.) to execute the methods described above according to the present invention. When the computer program is executed by a data processing device, it enables the computer-readable medium to implement the methods described above, i.e.: as... Figure 1 , Figure 2 or Figure 3 The method shown.

[0115] The computer-readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example,, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0116] The computer-readable storage medium may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0117] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0118] In summary, the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that in practice, general-purpose data processing devices such as microprocessors or digital signal processors (DSPs) can be used to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0119] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the present invention is not inherently related to any specific computer, virtual device, or electronic device, and various general-purpose devices can also implement the present invention. The above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0120] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0121] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, with clear signs / information informing users of the personal information processing rules, authorization is obtained from the individual through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0122] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A multimodal content generation system that combines GEO and SEO for dual-path collaborative optimization, characterized in that, include: The data acquisition module is used to configure the content capture workflow and capture raw information data in response to data capture requests. The content understanding module is used to preprocess and recognize the content of the raw information data, and generate a preprocessed document. The SEO optimization module is used to perform entity recognition and relation extraction on the normalized text contained in the preprocessed document, perform structured data annotation based on the extracted entities and relations, generate a structured document, and write it into the SEO article repository. The GEO optimization module is used to perform semantic augmentation and long-tail expansion based on the structured document to generate a keyword matrix; rewrite the structured document in a question-and-answer format according to the keyword matrix to obtain a GEO variant set; find and match credible standard data related to each text paragraph in the GEO variant set to generate an additional evidence chain summary; obtain target content from the SEO article repository and / or the GEO variant set, and add adapted multimodal materials based on the content semantics of the target content to generate a multimodal content body containing the additional evidence chain summary.

2. The multimodal content generation system combining GEO and SEO dual-path collaborative optimization according to claim 1, characterized in that, The SEO optimization module is also used for: Based on the keyword matrix, find multimodal content bodies containing additional chain of evidence summaries that are related to the keywords of the structured document; Insert keyword-related multimodal content bodies containing summaries of additional evidence chains into the structured documents included in the SEO article repository, and provide a redirect entry to the page.

3. The multimodal content generation system combining GEO and SEO dual-path collaborative optimization according to claim 1, characterized in that, The raw information data includes traditional search information data and generative search information data; The data acquisition module is specifically used for: Build a workflow platform and configure a content retrieval workflow on the workflow platform; Receive and respond to intelligent content generation requests, and obtain the content generation scope of the intelligent content generation request; Traditional search techniques are used to obtain relevant traditional search results for the content generation range, while generative search techniques are used to obtain generative search results related to the content generation range. The search results that meet the first relevance threshold are selected from the traditional search results and used as traditional search information data. Obtain the referenced content of the generative search results, and filter out the search results whose relevance meets the second relevance threshold from the referenced content as generative search information data.

4. The multimodal content generation system combining GEO and SEO dual-path collaborative optimization according to claim 1, characterized in that, It also includes a promotion routing module, which is specifically used for: Obtain multiple preset promotional information items, and obtain the semantic anchor points or semantic triggering rules of each promotional information item; When the target content contains semantic anchors or when a user's browsing behavior on the multimodal content meets the semantic triggering rules, preset promotional information is inserted into the preset promotional position of the multimodal content.

5. The multimodal content generation system combining GEO and SEO dual-path collaborative optimization according to claim 1, characterized in that, It also includes a monitoring and optimization module, which is specifically used for: The multimodal content body, which includes a summary of the additional chain of evidence, and the unoptimized raw information data will be published in parallel. The first content quality index of the multimodal content body containing the additional evidence chain summary and the second content quality index of the unoptimized raw information data are obtained respectively. The optimization effect of the multimodal content body containing the summary of the additional evidence chain is calculated based on the first content quality index and the second content quality index. When the optimization effect does not meet the optimization requirements, an optimization write-back instruction is triggered to regenerate the original information data corresponding to the multimodal content body containing the additional evidence chain summary.

6. A multimodal content generation method combining GEO and SEO dual-path collaborative optimization, characterized in that, include: Configure the content retrieval workflow to retrieve raw information data in response to data retrieval requests; The original information data is preprocessed and its content is identified to generate a preprocessed document; Entity recognition and relation extraction are performed on the normalized text contained in the preprocessed document. Based on the extracted entities and relations, structured data annotation is performed to generate a structured document and write it into the SEO article repository. Based on the structured document, semantic expansion and long-tail expansion are performed to generate a keyword matrix; The structured document is rewritten in a question-and-answer format based on the keyword matrix to obtain a set of GEO variants; Find and match credible standard data related to each text segment in the GEO variant set to generate an additional chain of evidence summary; The target content is obtained from the SEO article repository and / or the GEO variant set, and multimodal materials with corresponding content semantics are added based on the target content to generate a multimodal content body containing an additional evidence chain summary.

7. The multimodal content generation method combining GEO and SEO dual-path collaborative optimization according to claim 6, characterized in that, After generating the multimodal content body containing a summary of the additional chain of evidence, the method further includes: Based on the keyword matrix, find multimodal content bodies containing additional chain of evidence summaries that are related to the keywords of the structured document; Insert keyword-related multimodal content bodies containing summaries of additional evidence chains into the structured documents included in the SEO article repository, and provide a redirect entry to the page.

8. The multimodal content generation method combining GEO and SEO dual-path collaborative optimization according to claim 6, characterized in that, The raw information data includes traditional search information data and generative search information data; The configuration content capture workflow, in response to data capture requests, captures raw information data including: Build a workflow platform and configure a content retrieval workflow on the workflow platform; Receive and respond to intelligent content generation requests, and obtain the content generation scope of the intelligent content generation request; Traditional search techniques are used to obtain relevant traditional search results for the content generation range, while generative search techniques are used to obtain generative search results related to the content generation range. The search results that meet the first relevance threshold are selected from the traditional search results and used as traditional search information data. Obtain the referenced content of the generative search results, and filter out the search results whose relevance meets the second relevance threshold from the referenced content as generative search information data.

9. The multimodal content generation method combining GEO and SEO dual-path collaborative optimization according to claim 6, characterized in that, After generating a multimodal content body containing an additional evidence chain summary from the multimodal material based on the content semantic additional adaptation of the target delivery content, the method further includes: Obtain multiple preset promotional information items, and obtain the semantic anchor points or semantic triggering rules of each promotional information item; When the target content contains semantic anchors or when a user's browsing behavior on the multimodal content meets the semantic triggering rules, preset promotional information is inserted into the preset promotional position of the multimodal content.

10. The multimodal content generation method combining GEO and SEO dual-path collaborative optimization according to claim 6, characterized in that, After generating the multimodal content body containing a summary of the additional chain of evidence, the method further includes: The multimodal content body, which includes a summary of the additional chain of evidence, and the unoptimized raw information data will be published in parallel. The first content quality index of the multimodal content body containing the additional evidence chain summary and the second content quality index of the unoptimized raw information data are obtained respectively. The optimization effect of the multimodal content body containing the summary of the additional evidence chain is calculated based on the first content quality index and the second content quality index. When the optimization effect does not meet the optimization requirements, an optimization write-back instruction is triggered to regenerate the original information data corresponding to the multimodal content body containing the additional evidence chain summary.

11. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the steps of the multimodal content body generation method for GEO combined with SEO dual-path collaborative optimization as described in any one of claims 6-10.

Images