Article intelligent generation method and device, equipment, medium and product

Abstract

The invention discloses an intelligent article generation method and device, equipment, a medium and a product, and the method comprises the steps: constructing initial information matched with an article generation instruction according to the obtained article generation instruction; based on the article generation instruction and the initial information, constructing an article outline; generating an initial article based on the article outline and the initial information; and optimizing the initial article to obtain the target article, so that the speciality of the generated article can be improved, and the requirements of high-requirement scenes can be met.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, equipment, medium and product for intelligent article generation. Background Art

[0002] While intelligent article generation technology has made some progress, it still faces numerous technical difficulties and limitations in practical applications. In practice, it has been found that the process of intelligent article generation often lacks comprehensive information collection and relies heavily on a single source, making it difficult to achieve in-depth information mining. This results in uneven quality and insufficient professionalism in the generated articles, making them difficult to meet the needs of demanding scenarios. Summary of the Invention

[0003] The purpose of this application is to provide a method, device, equipment, medium and product for intelligent article generation, which can improve the professionalism of the generated articles and thus meet the needs of high-demand scenarios.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a method for intelligently generating articles, comprising:

[0006] Constructing initial information matching the article generation instruction according to the obtained article generation instruction;

[0007] constructing an article outline based on the article generation instruction and the initial information;

[0008] generating an initial article based on the article outline and the initial information;

[0009] The initial article is optimized to obtain a target article.

[0010] Optionally, constructing initial information matching the article generation instruction based on the acquired article generation instruction specifically includes:

[0011] Performing a search based on the acquired article generation instruction to obtain search information matching the article generation instruction;

[0012] According to the keywords in the article generation instruction, construct an expert role model corresponding to the keywords;

[0013] generating professional information matching the article generation instruction based on the expert role model;

[0014] Inputting the article generation instruction into a pre-built dialogue simulation model to obtain dialogue information output by the dialogue simulation model;

[0015] Inputting the article generation instruction into a pre-built question generation model to obtain predicted answer information;

[0016] The search information, the professional information, the dialogue information, and the predicted answer information are used together to determine initial information that matches the article generation instruction.

[0017] Optionally, inputting the article generation instruction into a pre-built question generation model to obtain predicted answer information specifically includes:

[0018] Inputting the article generation instruction into a pre-built question generation model to obtain an initial question set output by the question generation model; wherein the initial question set includes at least one initial question;

[0019] Deleting duplicate initial questions from the initial question set to obtain a candidate initial question set;

[0020] Performing a value evaluation on each initial question in the candidate initial question set to obtain a value of each initial question;

[0021] Determine the initial question whose question value is greater than a preset value threshold as the target question;

[0022] The answer corresponding to the target question is predicted to obtain predicted answer information matching the target question.

[0023] Optionally, constructing an article outline based on the article generation instruction and the initial information specifically includes:

[0024] Analyzing the article generation instructions to construct an initial outline of the article;

[0025] Extracting entity relationships from the initial information to construct a knowledge network;

[0026] Verifying the initial outline of the article based on the knowledge network to obtain a verification result;

[0027] If the verification result indicates that the initial article outline passes the verification, the initial article outline is determined as the article outline;

[0028] If the verification result indicates that the initial outline of the article fails the verification, the initial outline of the article is optimized to obtain an optimized initial outline of the article, and the above-mentioned step of verifying the initial outline of the article based on the knowledge network to obtain a verification result is performed.

[0029] Optionally, generating an initial article based on the article outline and the initial information specifically includes:

[0030] According to the initial information, construct the article content based on the article outline;

[0031] Determining the citation information of the article content according to the initial information;

[0032] Performing consistency verification on the article content and the reference information to obtain a consistency verification result;

[0033] If the consistency verification result indicates that the article content and the reference information pass the consistency verification, the article content and the verification result are jointly determined as the initial article;

[0034] If the consistency verification result indicates that the article content and the reference information fail the consistency verification, the article content and the reference information are modified to obtain modified article content and modified reference information, and the modified article content and modified reference information are determined as the initial article.

[0035] Optionally, optimizing the initial article to obtain a target article specifically includes:

[0036] Performing a quality assessment on the initial article to obtain a quality assessment result;

[0037] If the quality assessment result indicates that the initial article quality assessment fails, optimizing the initial article to obtain an optimized initial article, and performing the above-mentioned step of performing quality assessment on the initial article to obtain a quality assessment result;

[0038] If the quality assessment result indicates that the initial article passes the quality assessment, the initial article is deduplicated to obtain the target article.

[0039] In a second aspect, the present application provides an apparatus for intelligently generating articles, comprising:

[0040] A first construction unit is configured to construct initial information matching the article generation instruction according to the acquired article generation instruction;

[0041] A second construction unit is configured to construct an article outline based on the article generation instruction and the initial information;

[0042] A generating unit, configured to generate an initial article based on the article outline and the initial information;

[0043] The optimization unit is used to optimize the initial article to obtain a target article.

[0044] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any one of the above-described methods for intelligently generating articles.

[0045] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the above-mentioned methods for intelligently generating articles.

[0046] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of any one of the above-mentioned methods for intelligently generating articles.

[0047] In a sixth aspect, the present application provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction. When the processor executes the program or instruction, the steps of the intelligent article generation method described in any one of the above are implemented.

[0048] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0049] The present application provides an intelligent article generation method, device, equipment, medium and product, which effectively improves the professionalism and domain adaptability of the generated articles by constructing initial information that is highly matched with user instructions and combining a multi-dimensional raw data processing mechanism. The system constructs a structured outline based on instruction semantics to ensure that the logical architecture of the article conforms to professional writing standards. At the same time, through deep cleaning and knowledge extraction of raw data, the authority and accuracy of the content materials are guaranteed. In the generation process, the domain knowledge graph and terminology library are introduced for content optimization, which significantly enhances the accuracy of the application of professional terminology and the depth of expression of industry characteristics, so that the generated articles not only have a rigorous logical system, but also can meet the high-quality content requirements in professional scenarios such as academic research and industry analysis, improve the professionalism of the generated articles, thereby meeting the needs of high-demand scenarios, and solving the technical pain points of traditional generation models in professional fields such as shallow expression and misuse of terminology. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0051] Figure 1A flowchart of an intelligent article generation method provided in one embodiment of the present application;

[0052] Figure 2 A schematic diagram of the functional modules of an intelligent article generation device provided in one embodiment of the present application;

[0053] Figure 3 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0055] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0056] In an exemplary embodiment, Figure 1 As shown, a method for intelligently generating articles is provided. The method is executed by a computer device, and can be executed solely by a computer device such as a terminal or a server, or can be executed jointly by a terminal and a server. In an embodiment of the present application, the method includes the following steps 101 to 104.

[0057] in:

[0058] Step 101: construct initial information matching the article generation instruction based on the acquired article generation instruction.

[0059] In an embodiment of the present application, intelligent article generation can be achieved based on a pre-built intelligent article generation model, which may include four modules: a knowledge collection module, an outline generation module, an article generation module, and an article polishing module.

[0060] In the embodiment of the present application, the acquired article generation instruction can be processed by the knowledge collection module to construct initial information matching the article generation instruction. The knowledge collection module can also include a retriever system, a role generator, a dialogue simulation system, and a question generation system.

[0061] Specifically, the search engine system intelligently integrates multi-source information, supports flexible access to multiple search engines, and features adaptive source selection strategies and information source reliability assessment mechanisms. The system utilizes multi-threaded parallel search technology, coupled with load balancing and intelligent timeout management, to achieve efficient information acquisition and result merging.

[0062] The role generator is responsible for constructing expert role models. Through in-depth domain knowledge modeling and professional background building, it achieves differentiated perspective processing and realistic role behavior simulation. The system can dynamically adjust role configuration based on the topic, and through role complementarity analysis and perspective coverage assessment, it ensures the comprehensiveness and accuracy of information collection.

[0063] The conversation simulation system utilizes advanced dialogue strategy design to achieve intelligent management of multi-round conversations, precise context maintenance, and effective topic guidance. Through rigorous interaction quality control, the system automatically adjusts conversation depth, balances information density, and incorporates comprehensive mechanisms for responding to exceptions.

[0064] The question generation system, based on deep contextual understanding, intelligently constructs high-quality question sequences. By controlling question depth and ensuring diversity, it avoids duplicate and ineffective questions. The system also features a comprehensive question optimization mechanism that assesses question value, optimizes question sequences, predicts answer quality, and dynamically adjusts responses.

[0065] As an optional implementation, step 101 may include:

[0066] Performing a search based on the acquired article generation instruction to obtain search information matching the article generation instruction;

[0067] According to the keywords in the article generation instruction, construct an expert role model corresponding to the keywords;

[0068] generating professional information matching the article generation instruction based on the expert role model;

[0069] Inputting the article generation instruction into a pre-built dialogue simulation model to obtain dialogue information output by the dialogue simulation model;

[0070] Inputting the article generation instruction into a pre-built question generation model to obtain predicted answer information;

[0071] The search information, the professional information, the dialogue information, and the predicted answer information are used together to determine initial information that matches the article generation instruction.

[0072] This implementation significantly enhances the information richness and content expertise of the article generation system by integrating a multi-dimensional information generation mechanism. Specifically, the basic information obtained through command retrieval ensures the factual accuracy of the content, the expert role model construction mechanism strengthens the deep integration of professional domain knowledge, the dialogue simulation model introduces natural interaction logic, and the question generation model achieves the connection between demand prediction and answer. Ultimately, through the collaborative integration of multi-source information, the authority and logical coherence of the generated content are guaranteed, and the scenario adaptability is enhanced, so that the output results have professional depth, natural interaction, and a wide range of demand coverage.

[0073] In an embodiment of the present application, the dialogue simulation model can conduct multiple rounds of dialogue based on the input article generation instructions, and record the dialogue information of each round of dialogue. During the dialogue process, quality control can be performed on each dialogue sentence, specifically: obtaining the current sentence to be dialogued, analyzing the dialogue depth, context relevance and relevance of the sentence to be dialogued with the article generation instructions, and obtaining a dialogue quality analysis result; if the dialogue quality analysis result indicates that the dialogue depth, context relevance and relevance of the sentence to be dialogued with the article generation instructions all meet the quality control requirements, then the sentence to be dialogued is determined as a dialogue sentence.

[0074] Specifically, the article generation instruction is input into a pre-built question generation model to obtain the predicted answer information in the following manner:

[0075] Inputting the article generation instruction into a pre-built question generation model to obtain an initial question set output by the question generation model; wherein the initial question set includes at least one initial question;

[0076] Deleting duplicate initial questions from the initial question set to obtain a candidate initial question set;

[0077] Performing a value evaluation on each initial question in the candidate initial question set to obtain a value of each initial question;

[0078] Determine the initial question whose question value is greater than a preset value threshold as the target question;

[0079] The answer corresponding to the target question is predicted to obtain predicted answer information matching the target question.

[0080] Among them, the implementation of this implementation method achieves the coordinated optimization of demand understanding and content generation by building an intelligent closed loop of question generation and answer prediction. Specifically, the potential problem set is automatically mined based on the question generation model, effectively expanding the coverage dimension of demand; the quality density of the problem set is improved by eliminating redundant information through the deduplication mechanism; the introduction of the value assessment system combined with dynamic threshold screening achieves the precise positioning of high-value problems and avoids the ineffective consumption of computing resources; and finally, through the targeted answer prediction mechanism, it ensures the strong correlation between the output information and the core needs. This process not only enhances the logical coherence of content generation through the question-driven mechanism, but also establishes a positive feedback link between deep demand mining and professional answer generation, significantly improving the system's parsing accuracy and response quality to complex instructions.

[0081] In the embodiment of the present application, a method for predicting the answer corresponding to the target question and obtaining the predicted answer information matching the target question may include:

[0082] Predict the answer corresponding to the target question and obtain the initial predicted answer information matching the target question;

[0083] Evaluate the answer quality of the initial predicted answer information to obtain an answer quality evaluation value;

[0084] The initial predicted answer information whose answer quality evaluation value is greater than or equal to a preset answer evaluation threshold is determined as the predicted answer information.

[0085] The method of evaluating the answer quality of the initial predicted answer information and obtaining the answer quality evaluation value may include:

[0086] The writing quality of the initial predicted answer information is evaluated using a pre-built writing quality assessment model to obtain a writing quality score;

[0087] Analyze the information relevance between the initial predicted answer information and the target question;

[0088] Obtaining a first weight of the writing quality score and a second weight of the information relevance;

[0089] The answer quality evaluation value is determined using the writing quality score, the first weight, the information relevance, and the second weight.

[0090] Step 102: construct an article outline based on the article generation instruction and the initial information.

[0091] In the embodiment of the present application, an outline generation module can be used to construct an article outline based on the article generation instructions and initial information. The outline generation module may include an outline generation engine, an entity analysis system, a verification mechanism, and an optimization system.

[0092] The outline generation engine utilizes advanced hierarchical structure construction technology, achieving precise topic division through a topic decomposition algorithm. It utilizes hierarchical relationship reasoning to ensure structural rationality, while maintaining overall outline coordination through structural balance control and dynamic depth adjustment. Regarding logical relationship construction, the system analyzes topic logic chains, calculates inter-chapter correlations, and ensures the rigor of the outline structure through integrity checks and logical conflict resolution.

[0093] The entity analysis system is responsible for identifying and analyzing entities related to the article's topic. Using deep learning technology, the system accurately identifies key entities, performs multi-dimensional analysis of entity attributes, and constructs a knowledge network using entity relationship extraction. Based on this foundation, the system establishes a complete entity association graph and dynamically updates and maintains the knowledge system through relationship strength calculation and network structure optimization.

[0094] A verification mechanism ensures the integrity and quality of the outline. The system comprehensively assesses the outline's quality across multiple dimensions, including topic coverage, structural integrity analysis, and hierarchical rationality verification. Furthermore, a professional quality assessment system is equipped to evaluate structural balance, check logical coherence, analyze content coverage, and automatically generate optimization suggestions.

[0095] The optimization system uses a recursive optimization strategy to dynamically improve outline quality through continuous hierarchical structure adjustment, optimized association strength, and balanced topic distribution. The system also has a comprehensive feedback optimization mechanism that processes user feedback, analyzes quality indicators, and adjusts optimization strategies to ensure continuous improvement.

[0096] In terms of core technical features, the innovation of this module is mainly reflected in the following aspects: First, the multi-level title structure design realizes the adaptive adjustment of the hierarchical depth, ensuring the balance of the structure and the optimization of logical relationships; second, the entity association analysis system constructs a dynamically updated knowledge network through advanced entity recognition technology and relationship strength calculation; third, the outline verification mechanism ensures the integrity and professionalism of the outline through a multi-dimensional verification and quality assessment system; finally, the recursive optimization strategy achieves continuous improvement of the outline quality through a continuous improvement mechanism and dynamic optimization control.

[0097] As an optional implementation, step 102 may include constructing an article outline based on the article generation instruction and the initial information:

[0098] Analyzing the article generation instructions to construct an initial outline of the article;

[0099] Extracting entity relationships from the initial information to construct a knowledge network;

[0100] Verifying the initial outline of the article based on the knowledge network to obtain a verification result;

[0101] If the verification result indicates that the initial article outline passes the verification, the initial article outline is determined as the article outline;

[0102] If the verification result indicates that the initial outline of the article fails the verification, the initial outline of the article is optimized to obtain an optimized initial outline of the article, and the above-mentioned step of verifying the initial outline of the article based on the knowledge network to obtain a verification result is performed.

[0103] Among them, the implementation of this implementation method significantly improves the logical rigor and content reliability of the article outline by building a closed-loop optimization mechanism based on knowledge verification. Specifically, the knowledge network formed based on entity relationship extraction provides a structured knowledge benchmark for outline verification, effectively identifying potential logical faults and information missing; through the iterative verification-optimization mechanism, the outline framework is automatically verified and dynamically corrected, avoiding the subjectivity and inefficiency of manual verification; at the same time, the association analysis capability of the knowledge graph can deeply explore implicit information associations to ensure that the outline not only meets the requirements of explicit instructions, but also implies the knowledge context of professional fields; the final double-loop verification system (knowledge verification-manual optimization) not only ensures generation efficiency, but also builds a complete quality defense line from demand analysis to knowledge implementation, so that the article framework has both logical self-consistency and professional authority.

[0104] In an embodiment of the present application, in the process of constructing the initial outline of the article, it is necessary to perform a thematic logical chain analysis on the constructed initial outline of the article, that is, the chapter correlation between each chapter in the initial outline of the article can be calculated, and each chapter can be re-sorted according to the chapter correlation, thereby obtaining a more reasonable initial outline of the article.

[0105] Step 103: Generate an initial article based on the article outline and the initial information.

[0106] In the embodiment of the present application, the article generation module can generate an initial article based on the article outline and initial information. The article generation module may include a content generation engine, a parallel processing system, a citation manager, and a consistency checker.

[0107] The content generation engine, based on a distributed architecture, achieves load balancing and optimized result consolidation through a scientific task decomposition strategy and dynamic resource allocation mechanism. Equipped with an advanced context management system, the system maintains global context consistency, ensures local context synchronization, and handles potential content conflicts through a comprehensive conflict resolution mechanism.

[0108] The parallel processing system implements an efficient task scheduling mechanism, maximizing processing efficiency through task priority management and resource utilization optimization. The system uses intelligent synchronization control strategies to ensure data synchronization accuracy, maintain state consistency, effectively handle concurrency conflicts, and possess comprehensive exception recovery capabilities.

[0109] The Citation Manager is responsible for tracking and managing citations throughout the entire article. By recording sources in real time, maintaining citation relationships, and standardizing citation formats, the system ensures the accuracy and effectiveness of citations. Furthermore, the system is equipped with an intelligent citation optimization mechanism that controls citation density, optimizes citation distribution, assesses citation importance, and dynamically adjusts them.

[0110] A consistency checker ensures the quality and consistency of generated content. The system comprehensively ensures content accuracy and consistency through multiple dimensions, including topic consistency checks, point consistency analysis, and factual consistency verification. Furthermore, a real-time quality control system automatically corrects errors, detects and handles anomalies, and provides quality assessment feedback.

[0111] In terms of core technical features, the innovation of this module is mainly reflected in the following aspects: First, the recursive generation architecture ensures the consistency of content generation through scientific task decomposition and context transfer mechanism; second, the parallel processing mechanism achieves a significant improvement in generation efficiency through dynamic task scheduling and optimized resource allocation; third, the citation tracking system ensures the accuracy of citations through real-time source tracking and normalization processing; finally, the content consistency assurance mechanism ensures the high quality of output content through multi-dimensional consistency checks and continuous quality monitoring.

[0112] As an optional implementation, step 103 of generating the initial article based on the article outline and the initial information may include:

[0113] According to the initial information, construct the article content based on the article outline;

[0114] Determining the citation information of the article content according to the initial information;

[0115] Performing consistency verification on the article content and the reference information to obtain a consistency verification result;

[0116] If the consistency verification result indicates that the article content and the reference information pass the consistency verification, the article content and the verification result are jointly determined as the initial article;

[0117] If the consistency verification result indicates that the article content and the reference information fail the consistency verification, the article content and the reference information are modified to obtain modified article content and modified reference information, and the modified article content and modified reference information are determined as the initial article.

[0118] Among them, the implementation of this implementation method significantly improves the academic rigor and information credibility of the article content by building a closed-loop system for content generation and verification. Specifically, the content framework is automatically constructed based on the initial information, realizing the structured organization of key knowledge points; through the precise association mechanism of cited information, it is ensured that the core arguments are supported by traceable sources; and the two-way consistency verification system effectively eliminates information mismatches and citation biases by cross-checking the logical correspondence between content statements and references; when inconsistencies are detected, the system automatically triggers the iterative correction process, and by dynamically adjusting the matching degree between content statements and citation annotations, it ultimately forms a credible text unit with a strong binding of content and source. This mechanism not only reduces the cost of manual review through automated verification, but also builds a full-process quality assurance system from knowledge retrieval to the implementation of academic norms, so that the output content meets both logical consistency and academic compliance requirements.

[0119] In an embodiment of the present application, consistency verification is performed on the article content and the reference information, and a method for obtaining a consistency verification result may include:

[0120] The article content can be tested for topic consistency to obtain topic consistency test results;

[0121] You can also perform a viewpoint consistency test on the article content to obtain the viewpoint consistency test results;

[0122] It is also possible to perform fact consistency checks on the article content and citation information to obtain fact consistency check results;

[0123] Furthermore, the subject consistency test results, the viewpoint consistency test results, and the fact consistency test results can be collectively determined as the consistency verification results.

[0124] Among them, as long as any one of the subject consistency test results, viewpoint consistency test results and fact consistency test results in the consistency verification results indicates failure, the consistency verification result means that the article content and citation information have failed the consistency verification.

[0125] Step 104: Optimize the initial article to obtain a target article.

[0126] In the embodiment of the present application, the initial article can be optimized by the article polishing module to obtain the target article. The article polishing module can include a dual model system, an optimization engine, a structure maintainer, and a deduplication system.

[0127] The dual-model system utilizes advanced collaborative optimization technology, achieving optimal editing results through a model-by-model division of labor and collaborative decision-making. The system incorporates specialized quality assessment and optimization execution models. The assessment model conducts multi-dimensional quality analysis of articles, while the optimization model executes specific improvement actions. An efficient information exchange mechanism is established between the two models, ensuring a coherent and consistent optimization process.

[0128] The optimization engine is responsible for multi-dimensional article optimization. It comprehensively improves article quality by optimizing language fluency, enhancing expression accuracy, and strengthening logical coherence. Furthermore, the system is equipped with an intelligent optimization strategy manager that dynamically adjusts optimization strategies based on the characteristics of different article types, achieving personalized polishing results.

[0129] The structure maintainer ensures the structural integrity of the article during the polishing process. By maintaining chapter structure, paragraph relationships, and controlling thematic coherence, the system ensures that the article's structure is not disrupted during the optimization process. Furthermore, it is equipped with a professional structural optimization mechanism that can optimize the article's structure while maintaining the original structure, improving overall performance.

[0130] The deduplication system is responsible for checking for duplicate content and optimizing it. Using real-time content comparison, duplication assessment, and similarity analysis, the system effectively identifies and processes duplicate content within articles. Furthermore, the system is equipped with an intelligent expression transformation mechanism that allows for diversified expression transformation of duplicate content while preserving the original meaning, enhancing the originality of the article.

[0131] In terms of core technical features, the innovation of this module is mainly reflected in the following aspects: First, the dual-model collaborative mechanism achieves the optimization of polishing effect through the close cooperation of the evaluation model and the optimization model; second, the multi-dimensional optimization strategy ensures the comprehensiveness of optimization by comprehensively considering multiple aspects such as language, logic, and structure; third, the structure maintenance mechanism ensures the integrity of the article through strict structural control and moderate optimization; finally, the intelligent deduplication system improves the originality of the article through advanced duplication detection technology and expression transformation.

[0132] As an optional implementation, step 104 may include optimizing the initial article to obtain the target article by:

[0133] Performing a quality assessment on the initial article to obtain a quality assessment result;

[0134] If the quality assessment result indicates that the initial article quality assessment fails, optimizing the initial article to obtain an optimized initial article, and performing the above-mentioned step of performing quality assessment on the initial article to obtain a quality assessment result;

[0135] If the quality assessment result indicates that the initial article passes the quality assessment, the initial article is deduplicated to obtain the target article.

[0136] Among them, the implementation of this implementation method has achieved an automated lean improvement in article quality by building a quality-driven content optimization closed loop. Specifically, based on a multi-dimensional quality assessment system, a comprehensive physical examination of the article is carried out to accurately locate potential problems such as logical loopholes, professional flaws, and citation biases; through an iterative optimization mechanism, an intelligent feedback loop of "assessment-correction-reassessment" is established to ensure that the content quality continues to approach the preset standards, effectively avoiding the subjective blind spots and inefficient rework of manual proofreading; after the quality standards are met, intelligent deduplication processing is introduced to eliminate redundant expressions at the semantic level, which not only ensures the originality of the content, but also improves the information density. This process deeply integrates quality control into the generation link, replaces traditional manual spot checks with algorithm-driven standardized assessments, and cooperates with dynamic optimization strategies. While significantly reducing the cost of manual review, it builds a full-process guarantee system from content generation to quality delivery, and ultimately outputs high-quality texts that are both professional and refined.

[0137] Through innovative technical solutions, the present invention has achieved remarkable technical effects in terms of information comprehensiveness, structural rationality, generation efficiency, content reliability, optimization intelligence and system stability.

[0138] Information Collection Effectiveness: In terms of comprehensiveness, the system achieves in-depth coverage of target topics through a multi-faceted information collection mechanism. A cross-validation mechanism ensures information accuracy, while a comprehensive information quality assurance system, encompassing multiple steps including source reliability assessment, content authenticity verification, and timeliness checks, effectively safeguards the reliability of the information base. The system's multi-expert perspectives and cross-disciplinary information integration capabilities ensure the breadth and depth of information collection, providing a solid data foundation for subsequent processing.

[0139] Structural Optimization: To ensure structural rationality, the system utilizes a multi-level heading system and a logical relationship optimization algorithm to achieve a scientifically organized article structure. The system dynamically adjusts the structural hierarchy and optimizes chapter relationships based on thematic characteristics and content requirements, ensuring a rigorous and layered structure. Through a real-time optimization response mechanism, the system can promptly adjust local structures based on content changes and user feedback while maintaining global structural consistency, achieving dynamic structural optimization and flexible adaptation.

[0140] Efficiency Improvement: Parallel processing and a distributed generation architecture significantly improve article generation speed. The system utilizes intelligent task decomposition and resource scheduling strategies to maximize processing power. Concurrent multi-task execution and dynamic load balancing ensure efficient utilization of system resources. Real-time performance monitoring and adaptive adjustments continuously optimize processing efficiency. The system's distributed architecture not only increases processing speed but also enhances scalability and fault tolerance.

[0141] Content quality: Regarding content reliability, the system has established a comprehensive citation tracking system, enabling full control over information sources. Source verification ensures the accuracy of cited content, while a professional content review mechanism ensures the quality of generated content from multiple perspectives. The system's real-time recording and maintenance mechanisms ensure the accuracy of citation relationships, while authority assessment and timeliness checks further enhance content reliability.

[0142] Optimization Results: In terms of intelligent optimization, the dual-model collaborative optimization mechanism enables intelligent article polishing. The system employs multi-dimensional optimization strategies, covering aspects such as language expression, logical coherence, and structural integrity. Through an adaptive optimization mechanism, the system dynamically adjusts optimization strategies based on the characteristics of each article, achieving personalized optimization results. Continuous learning and improvement, as well as real-time performance evaluation, ensure continuous improvement in the optimization process.

[0143] System Operational Performance: In terms of system stability, a comprehensive fault-tolerance mechanism and quality monitoring system ensure reliable operation. The system has established a multi-level exception handling mechanism to effectively address various operational anomalies. Through full-process quality tracking and multi-dimensional indicator evaluation, the system achieves continuous quality monitoring and improvement. Furthermore, module independence and interface stability maintenance ensure the long-term stable operation of the system, providing users with reliable service guarantees.

[0144] The intelligent article generation method of the present invention has broad application prospects and can play an important role in many fields such as academic paper generation, technical document writing, research report generation, news article creation, product document production and corporate content creation.

[0145] Academic Paper Generation: This system effectively supports the writing of various types of papers, including literature reviews, experimental studies, and theoretical research. For literature reviews, the system comprehensively collects and analyzes relevant research literature, constructs a systematic research framework, and generates in-depth literature analysis reports. For experimental research papers, the system standardizes the organization of experimental methods, data analysis, and discussion of results, ensuring the integrity and logic of the paper structure. For theoretical research papers, the system assists in constructing theoretical frameworks, analyzing conceptual relationships, and forming a rigorous argumentation system.

[0146] Technical Documentation: In this area, the system efficiently generates a variety of technical documentation, including API documentation, system design documentation, and user manuals. For API documentation, the system accurately describes interface functions, parameter specifications, and usage examples, ensuring the practicality and understandability of the documentation. For system design documentation, the system clearly presents the architectural design, module functionality, and technical implementation details, helping development teams better understand and maintain the system. For user manuals, the system provides a user-friendly explanation of product features, operational procedures, and precautions.

[0147] Research Report Generation: This system supports the compilation of various reports, including market research reports, technical research reports, and project evaluation reports. For market research reports, the system comprehensively analyzes market data, competitive landscape, and development trends, generating valuable market insights. For technical research reports, the system systematically analyzes the current state of technology, development directions, and application prospects, providing a basis for technological decision-making. For project evaluation reports, the system provides a multi-dimensional assessment of project feasibility, risk factors, and expected benefits.

[0148] News Article Creation: In the area of ​​news article creation, this system can assist in the writing of news reports, industry news, and special reports. For news reports, the system can quickly integrate event information and construct a clear narrative structure, ensuring the timeliness and accuracy of reports. For industry news, the system can continuously track industry dynamics, analyze development trends, and generate in-depth industry analysis articles. For special reports, the system can delve into specific topics, providing comprehensive background analysis and professional insights.

[0149] Product Documentation: This system efficiently generates a variety of documents, including product manuals, marketing copy, and training materials. For product manuals, the system details product features, usage instructions, and technical specifications, ensuring users fully understand product functionality. For marketing copy, the system creates compelling promotional content tailored to product characteristics and target audiences. For training materials, the system systematically organizes key knowledge points, designs learning paths, and produces easy-to-understand training documents.

[0150] Enterprise Content Creation: This system supports the creation of a variety of content, including corporate promotional materials, internal management documents, and external communication documents. For promotional materials, the system can highlight corporate characteristics, convey brand values, and create professional corporate image copy. For internal management documents, the system can standardize the development of rules and regulations, workflows, and operational guidelines. For external communication documents, the system can generate formal business documents and professional communication materials tailored to the needs of different occasions.

[0151] The intelligent article generation method proposed in this paper adopts a modular design and achieves automated generation of high-quality articles through collaborative processing in multiple stages. The entire implementation process includes multiple stages: information collection, outline generation, article generation, optimization and improvement, and system monitoring. Each stage has its own specific processing mechanism and quality assurance measures.

[0152] Information Collection Phase: During this phase, the system first initializes a multi-role collaboration mechanism, establishing a role system encompassing search experts, analysis experts, and verification experts. Through this parallel information retrieval mechanism, the system simultaneously acquires relevant information from multiple sources, including professional databases, academic literature, industry reports, and other sources. During information integration, the system employs a cross-validation mechanism to ensure the accuracy and completeness of collected information. Furthermore, it establishes an information traceability system to effectively track and manage information sources.

[0153] Outline Generation: During this phase, the system first conducts topic analysis and planning, using deep semantic analysis technology to accurately grasp the article's theme and core points. During structural optimization, the system utilizes intelligent algorithms to optimize the heading hierarchy, ensuring logical rationality and clarity of the hierarchy. Furthermore, the system implements a dynamic outline adjustment mechanism, adjusting the outline structure as needed based on the richness and logical relevance of the content, ensuring the scientific and practical nature of the final outline.

[0154] Article Generation Phase: During this phase, the system first activates a distributed task processing mechanism, breaking down the article generation task into multiple parallel processing units. During the content generation process, the system writes specific content for each chapter based on the outline structure, ensuring the professionalism and integrity of the content. Through a unified content management mechanism, the system effectively organizes and manages generated content, while also performing real-time consistency checks to ensure coherence and coordination across all components.

[0155] Optimization and Improvement Phase: During this phase, the system activates a dual-model collaborative processing mechanism, leveraging specialized and general models to comprehensively optimize article content. During this multi-dimensional content optimization process, the system optimizes language expression, terminology, logical structure, and other dimensions to enhance the overall quality of the article. Through quality verification and improvement mechanisms, the system continuously monitors and improves article quality, ensuring that the final output meets the expected quality standards.

[0156] System Monitoring: During this phase, the system implements a full-process monitoring mechanism, providing real-time monitoring and data collection for every step of article generation. When anomalies arise, the system uses intelligent exception handling mechanisms to quickly identify and resolve issues, ensuring stable operation. Furthermore, the system establishes a continuous optimization and improvement mechanism, using operational data and feedback to continuously optimize system performance, improve processing efficiency, and enhance output quality.

[0157] This invention offers several technological innovations in the field of intelligent article generation. Through innovative solutions, it addresses existing issues and significantly improves the quality and efficiency of article generation. Key technical innovations include a multi-model collaboration mechanism, a distributed processing architecture, an intelligent verification system, and an adaptive optimization system.

[0158] Multi-model collaboration mechanism: This invention innovatively proposes a multi-model collaboration mechanism, which achieves higher-quality article generation through the organic combination of professional models and general models. Professional models are responsible for processing professional knowledge and terminology in specific fields to ensure the professionalism and accuracy of the article; general models are responsible for optimizing the overall expression and structure of the article to improve its readability and fluency. Through a carefully designed collaborative strategy, the two types of models cooperate and complement each other during the article generation process, ensuring both the professional depth of the article and the ease of expression.

[0159] Distributed Processing Architecture: This paper utilizes an innovative distributed processing architecture to achieve efficient parallel processing of article generation tasks. The system decomposes the article generation process into multiple independent processing units, including modules for information acquisition, content generation, and quality optimization. Each module can operate independently and collaborate with each other. Through a task scheduling mechanism, the system dynamically allocates computing resources based on resource availability and processing requirements, maximizing processing efficiency. Furthermore, the system also incorporates a comprehensive data synchronization mechanism to ensure data consistency across processing units.

[0160] Intelligent Verification System: This invention incorporates an innovative intelligent verification system that ensures article quality through a multi-layered verification mechanism. The system incorporates dedicated verification nodes at every stage of the process, including information collection, content generation, and optimization and improvement. Using intelligent algorithms, it verifies content across multiple dimensions. Information verification automatically verifies its accuracy and timeliness; content verification examines the article's logic and completeness; and quality verification assesses the article's professionalism and effectiveness. This comprehensive verification system effectively ensures article quality.

[0161] Adaptive Optimization System: This paper has developed an innovative adaptive optimization system that automatically adjusts processing strategies based on the characteristics and requirements of different article types. Using machine learning algorithms, the system continuously accumulates and analyzes processing experience to form optimized models tailored to different article types. In practice, the system automatically selects the most appropriate processing strategy based on factors such as the article's topic, field, and target audience. Dynamic adjustments are made during the processing process based on real-time feedback to ensure optimal processing results.

[0162] The intelligent article generation method of this invention offers significant advantages and benefits, not only achieving innovative breakthroughs at the technical level but also bringing substantial value enhancements at the application level. These advantages and benefits are primarily reflected in technical advantages, application advantages, economic benefits, and social benefits.

[0163] Technical Advantages: This invention offers significant technical advantages, primarily in terms of automation, quality assurance, and system reliability. Regarding automation, the system utilizes intelligent algorithms to automate the entire article generation process, significantly reducing the need for manual intervention. Regarding quality assurance, the system utilizes multiple verification mechanisms and optimization strategies to ensure the professionalism and readability of generated articles. Regarding system reliability, the application of a distributed architecture and fault-tolerant mechanisms ensures the system's high stability and scalability.

[0164] Application Advantages: This invention offers significant advantages at the application level, primarily in terms of adaptability, scalability, and ease of use. Regarding adaptability, the system can handle a wide range of article generation needs, meeting the application requirements of diverse scenarios. Regarding scalability, the system's modular design facilitates functional expansion and performance optimization. Furthermore, the system provides a user-friendly interface and comprehensive operating instructions, lowering the barrier to entry.

[0165] Economic Benefits: This invention can bring significant economic benefits, mainly reflected in three aspects: efficiency improvement, quality assurance, and cost optimization. In terms of efficiency improvement, the system significantly shortens the article generation time and improves content production efficiency. In terms of quality assurance, the system's intelligent verification mechanism ensures the high quality of output content and reduces the cost of subsequent revisions. In terms of cost optimization, the system's degree of automation reduces human resource investment and achieves effective cost control.

[0166] Social Benefits: This invention has positive social benefits, primarily in the areas of knowledge dissemination, industry upgrading, and value creation. Regarding knowledge dissemination, the system promotes the efficient dissemination and sharing of professional knowledge, promoting the popularization and application of knowledge. Regarding industry upgrading, the system's innovative technologies drive technological progress in the content production industry and promote its digital transformation. Regarding value creation, the system provides organizations and individuals with efficient content production tools, creating new growth points.

[0167] Implementing steps 101 to 104 above improves the professionalism of the generated article, thereby meeting the needs of demanding scenarios and resolving technical pain points such as shallow expression and misuse of terminology in traditional generative models in professional fields. Furthermore, the present application can also ensure the authority and logical coherence of the generated content through the collaborative integration of multi-source information, while also enhancing the adaptability of scenarios, so that the output results have both professional depth, natural interaction, and a breadth of demand coverage. Furthermore, the present application can also enhance the logical coherence of content generation through a question-driven mechanism, and establish a positive feedback loop between deep demand mining and professional answer generation, significantly improving the system's parsing accuracy and response quality for complex instructions. Furthermore, the present application can also build a complete quality defense line from demand analysis to knowledge implementation, ensuring that the article framework has both logical consistency and professional authority. Furthermore, the present application can also reduce manual review costs through automated verification, and build a full-process quality assurance system from knowledge retrieval to academic standard implementation, ensuring that the output content meets both logical consistency and academic compliance requirements. Furthermore, the present application can also output high-quality text that combines professional rigor with refined expression.

[0168] Based on the same inventive concept, the present application also provides an intelligent article generation device for implementing the aforementioned intelligent article generation method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the intelligent article generation device provided below can be found in the aforementioned limitations of the intelligent article generation method and will not be further elaborated here.

[0169] In an exemplary embodiment, Figure 2 As shown, an article intelligent generation device is provided, including:

[0170] A first constructing unit 201 is configured to construct initial information matching the article generation instruction according to the acquired article generation instruction;

[0171] A second construction unit 202 is configured to construct an article outline based on the article generation instruction and the initial information;

[0172] A generating unit 203 is configured to generate an initial article based on the article outline and the initial information;

[0173] The optimization unit 204 is configured to optimize the initial article to obtain a target article.

[0174] As an optional implementation manner, the first construction unit 201 constructs the initial information matching the article generation instruction according to the acquired article generation instruction in the following manner:

[0175] Performing a search based on the acquired article generation instruction to obtain search information matching the article generation instruction;

[0176] According to the keywords in the article generation instruction, construct an expert role model corresponding to the keywords;

[0177] generating professional information matching the article generation instruction based on the expert role model;

[0178] Inputting the article generation instruction into a pre-built dialogue simulation model to obtain dialogue information output by the dialogue simulation model;

[0179] Inputting the article generation instruction into a pre-built question generation model to obtain predicted answer information;

[0180] The search information, the professional information, the dialogue information, and the predicted answer information are used together to determine initial information that matches the article generation instruction.

[0181] This implementation significantly enhances the information richness and content expertise of the article generation system by integrating a multi-dimensional information generation mechanism. Specifically, the basic information obtained through command retrieval ensures the factual accuracy of the content, the expert role model construction mechanism strengthens the deep integration of professional domain knowledge, the dialogue simulation model introduces natural interaction logic, and the question generation model achieves the connection between demand prediction and answer. Ultimately, through the collaborative integration of multi-source information, the authority and logical coherence of the generated content are guaranteed, and the scenario adaptability is enhanced, so that the output results have professional depth, natural interaction, and a wide range of demand coverage.

[0182] As an optional implementation, the first construction unit 201 inputs the article generation instruction into a pre-constructed question generation model to obtain the predicted answer information in the following manner:

[0183] Inputting the article generation instruction into a pre-built question generation model to obtain an initial question set output by the question generation model; wherein the initial question set includes at least one initial question;

[0184] Deleting duplicate initial questions from the initial question set to obtain a candidate initial question set;

[0185] Performing a value evaluation on each initial question in the candidate initial question set to obtain a value of each initial question;

[0186] Determine the initial question whose question value is greater than a preset value threshold as the target question;

[0187] The answer corresponding to the target question is predicted to obtain predicted answer information matching the target question.

[0188] Among them, the implementation of this implementation method achieves the coordinated optimization of demand understanding and content generation by building an intelligent closed loop of question generation and answer prediction. Specifically, the potential problem set is automatically mined based on the question generation model, effectively expanding the coverage dimension of demand; the quality density of the problem set is improved by eliminating redundant information through the deduplication mechanism; the introduction of the value assessment system combined with dynamic threshold screening achieves the precise positioning of high-value problems and avoids the ineffective consumption of computing resources; and finally, through the targeted answer prediction mechanism, it ensures the strong correlation between the output information and the core needs. This process not only enhances the logical coherence of content generation through the question-driven mechanism, but also establishes a positive feedback link between deep demand mining and professional answer generation, significantly improving the system's parsing accuracy and response quality to complex instructions.

[0189] As an optional implementation, the second construction unit 202 may construct the article outline based on the article generation instruction and the initial information in the following manner:

[0190] Analyzing the article generation instructions to construct an initial outline of the article;

[0191] Extracting entity relationships from the initial information to construct a knowledge network;

[0192] Verifying the initial outline of the article based on the knowledge network to obtain a verification result;

[0193] If the verification result indicates that the initial article outline passes the verification, the initial article outline is determined as the article outline;

[0194] If the verification result indicates that the initial outline of the article fails the verification, the initial outline of the article is optimized to obtain an optimized initial outline of the article, and the above-mentioned step of verifying the initial outline of the article based on the knowledge network to obtain a verification result is performed.

[0195] Among them, the implementation of this implementation method significantly improves the logical rigor and content reliability of the article outline by building a closed-loop optimization mechanism based on knowledge verification. Specifically, the knowledge network formed based on entity relationship extraction provides a structured knowledge benchmark for outline verification, effectively identifying potential logical faults and information missing; through the iterative verification-optimization mechanism, the outline framework is automatically verified and dynamically corrected, avoiding the subjectivity and inefficiency of manual verification; at the same time, the association analysis capability of the knowledge graph can deeply explore implicit information associations to ensure that the outline not only meets the requirements of explicit instructions, but also implies the knowledge context of professional fields; the final double-loop verification system (knowledge verification-manual optimization) not only ensures generation efficiency, but also builds a complete quality defense line from demand analysis to knowledge implementation, so that the article framework has both logical self-consistency and professional authority.

[0196] As an optional implementation, the generating unit 203 may generate the initial article based on the article outline and the initial information in the following manner:

[0197] According to the initial information, construct the article content based on the article outline;

[0198] Determining the citation information of the article content according to the initial information;

[0199] Performing consistency verification on the article content and the reference information to obtain a consistency verification result;

[0200] If the consistency verification result indicates that the article content and the reference information pass the consistency verification, the article content and the verification result are jointly determined as the initial article;

[0201] If the consistency verification result indicates that the article content and the reference information fail the consistency verification, the article content and the reference information are modified to obtain modified article content and modified reference information, and the modified article content and modified reference information are determined as the initial article.

[0202] Among them, the implementation of this implementation method significantly improves the academic rigor and information credibility of the article content by building a closed-loop system for content generation and verification. Specifically, the content framework is automatically constructed based on the initial information, realizing the structured organization of key knowledge points; through the precise association mechanism of cited information, it is ensured that the core arguments are supported by traceable sources; and the two-way consistency verification system effectively eliminates information mismatches and citation biases by cross-checking the logical correspondence between content statements and references; when inconsistencies are detected, the system automatically triggers the iterative correction process, and by dynamically adjusting the matching degree between content statements and citation annotations, it ultimately forms a credible text unit with a strong binding of content and source. This mechanism not only reduces the cost of manual review through automated verification, but also builds a full-process quality assurance system from knowledge retrieval to the implementation of academic norms, so that the output content meets both logical consistency and academic compliance requirements.

[0203] As an optional implementation, the optimization unit 204 optimizes the initial article to obtain the target article in the following manner:

[0204] Performing a quality assessment on the initial article to obtain a quality assessment result;

[0205] If the quality assessment result indicates that the initial article quality assessment fails, optimizing the initial article to obtain an optimized initial article, and performing the above-mentioned step of performing quality assessment on the initial article to obtain a quality assessment result;

[0206] If the quality assessment result indicates that the initial article passes the quality assessment, the initial article is deduplicated to obtain the target article.

[0207] Wherein, implementing this embodiment,

[0208] By building a quality-driven content optimization closed loop, we have achieved automated lean improvement in article quality. Specifically, based on a multi-dimensional quality assessment system, we conduct a comprehensive physical examination of articles to accurately locate potential problems such as logical loopholes, professional flaws, and citation biases; through an iterative optimization mechanism, we establish an intelligent feedback loop of "assessment-correction-reassessment" to ensure that the quality of content continues to approach the preset standards, effectively avoiding the subjective blind spots and inefficient rework of manual proofreading; after the quality standards are met, we introduce intelligent deduplication processing to eliminate redundant expressions at the semantic level, which not only ensures the originality of the content, but also improves the information density. This process deeply integrates quality control into the generation link, replaces traditional manual spot checks with algorithm-driven standardized assessments, and cooperates with dynamic optimization strategies. While significantly reducing the cost of manual review, it builds a full-process guarantee system from content generation to quality delivery, and ultimately outputs high-quality texts that are both professional and refined.

[0209] The implementation of the above-mentioned implementation method improves the professionalism of the generated articles, thereby meeting the needs of high-demand scenarios and solving technical pain points such as shallow expression and misuse of terminology in traditional generation models in professional fields. In addition, the present application can also ensure the authority and logical coherence of the generated content through the collaborative integration of multi-source information, and enhance the scene adaptation capability, so that the output results have professional depth, natural interaction and breadth of demand coverage. In addition, the present application can also enhance the logical coherence of content generation through a question-driven mechanism, and establish a positive feedback link between deep demand mining and professional answer generation, significantly improving the system's parsing accuracy and response quality for complex instructions. In addition, the present application can also build a complete quality defense line from demand analysis to knowledge implementation, so that the article framework has both logical self-consistency and professional authority. In addition, the present application can also reduce the cost of manual review through automated verification, and build a full-process quality assurance system from knowledge retrieval to academic standard implementation, so that the output content meets both logical self-consistency and academic compliance requirements. In addition, the present application can also output high-quality texts that are both professional rigor and concise in expression.

[0210] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store article intelligent generation data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an article intelligent generation method is implemented.

[0211] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0212] In an exemplary embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0213] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0214] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0215] In an exemplary embodiment, a chip is provided, which includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the steps in the above-mentioned method embodiments and achieve the same technical effects. To avoid repetition, they are not described here.

[0216] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0217] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0218] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0219] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0220] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0221] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for intelligent article generation, characterized in that: The article intelligent generation method includes: Constructing initial information matching the article generation instruction according to the obtained article generation instruction; constructing an article outline based on the article generation instruction and the initial information; generating an initial article based on the article outline and the initial information; The initial article is optimized to obtain a target article.

2. The method for intelligently generating articles according to claim 1, characterized in that: The step of constructing initial information matching the article generation instruction based on the obtained article generation instruction specifically includes: Performing a search based on the acquired article generation instruction to obtain search information matching the article generation instruction; According to the keywords in the article generation instruction, construct an expert role model corresponding to the keywords; generating professional information matching the article generation instruction based on the expert role model; Inputting the article generation instruction into a pre-built dialogue simulation model to obtain dialogue information output by the dialogue simulation model; Inputting the article generation instruction into a pre-built question generation model to obtain predicted answer information; The search information, the professional information, the dialogue information, and the predicted answer information are used together to determine initial information that matches the article generation instruction.

3. The method for intelligently generating articles according to claim 2, characterized in that: The step of inputting the article generation instruction into a pre-built question generation model to obtain predicted answer information specifically includes: Inputting the article generation instruction into a pre-built question generation model to obtain an initial question set output by the question generation model; wherein the initial question set includes at least one initial question; Deleting duplicate initial questions from the initial question set to obtain a candidate initial question set; Performing a value evaluation on each initial question in the candidate initial question set to obtain a value of each initial question; Determine the initial question whose question value is greater than a preset value threshold as the target question; The answer corresponding to the target question is predicted to obtain predicted answer information matching the target question.

4. The method for intelligently generating articles according to claim 1, characterized in that: The step of constructing an article outline based on the article generation instruction and the initial information specifically includes: Analyzing the article generation instructions to construct an initial outline of the article; Extracting entity relationships from the initial information to construct a knowledge network; Verifying the initial outline of the article based on the knowledge network to obtain a verification result; If the verification result indicates that the initial article outline passes the verification, the initial article outline is determined as the article outline; If the verification result indicates that the initial outline of the article fails the verification, the initial outline of the article is optimized to obtain an optimized initial outline of the article, and the above-mentioned step of verifying the initial outline of the article based on the knowledge network to obtain a verification result is performed.

5. The method for intelligently generating articles according to claim 1, characterized in that: Generating an initial article based on the article outline and the initial information specifically includes: According to the initial information, construct the article content based on the article outline; Determining the citation information of the article content according to the initial information; Performing consistency verification on the article content and the reference information to obtain a consistency verification result; If the consistency verification result indicates that the article content and the reference information pass the consistency verification, the article content and the verification result are jointly determined as the initial article; If the consistency verification result indicates that the article content and the reference information fail the consistency verification, the article content and the reference information are modified to obtain modified article content and modified reference information, and the modified article content and modified reference information are determined as the initial article.

6. The method for intelligently generating articles according to claim 1, characterized in that: Optimizing the initial article to obtain the target article specifically includes: Performing a quality assessment on the initial article to obtain a quality assessment result; If the quality assessment result indicates that the initial article quality assessment fails, optimizing the initial article to obtain an optimized initial article, and performing the above-mentioned step of performing quality assessment on the initial article to obtain a quality assessment result; If the quality assessment result indicates that the initial article passes the quality assessment, the initial article is deduplicated to obtain the target article.

7. An intelligent article generation device, characterized in that: The article intelligent generation device includes: A first construction unit is configured to construct initial information matching the article generation instruction according to the acquired article generation instruction; A second construction unit is configured to construct an article outline based on the article generation instruction and the initial information; A generating unit, configured to generate an initial article based on the article outline and the initial information; The optimization unit is used to optimize the initial article to obtain a target article.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for intelligently generating articles according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for intelligently generating articles according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for intelligently generating articles according to any one of claims 1 to 6 are implemented.

Images