Intelligent compilation method and system for scientific research achievement standard draft based on big data

Abstract

The invention relates to the technical field of standardized management, in particular to a scientific research achievement standard draft intelligent compilation method and system based on big data. The method comprises the following steps: converting scientific research achievements into a computable technical structure expression model by constructing a multi-dimensional semantic layering and causal relationship model, and establishing a standardized adaptation index; performing isomorphic mapping on a result structure and a current standard vector space, identifying a standard blank unit, calculating an innovation contribution index, and realizing automatic positioning and priority ranking of blank terms; constructing a clause skeleton by utilizing topological sorting and conflict detection, and generating a standard draft in combination with a structured semantic template; a multi-agent collaborative verification mechanism is constructed, expert feedback is converted into structured data, and dynamic optimization of terms is realized through credibility weighting and influence tensor; and establishing an implementation monitoring and life cycle management system, and carrying out difference analysis and adaptive iteration. According to the invention, intelligent compilation and continuous optimization of the whole process from scientific achievements to standard draft are realized.

Description

Technical Field

[0001] This invention relates to the field of standardization management technology, specifically to a method and system for intelligent drafting of scientific research achievement standards based on big data. Background Technology

[0002] With the accelerating pace of global technological innovation and the continuous expansion of scientific research output, a large number of cutting-edge technologies urgently need to be transformed into technical standards to guide industrial upgrading and regulate market order. As an important bridge connecting science and technology with industry, the quality and efficiency of standard development directly affect the depth and breadth of the transformation and application of scientific and technological achievements. Currently, scientific research results are usually presented in the form of unstructured texts such as academic papers and technical reports. The core elements contained therein, such as technical problems, solutions, and performance indicators, lack a unified quantitative expression model, making it difficult to establish an effective technical mapping relationship between them and the standard system. At the same time, the standard-setting process involves the integration of multidisciplinary knowledge and multi-entity collaboration, and has strict requirements on aspects such as technological maturity, industry applicability, and parameter replicability.

[0003] Chinese invention patent CN120597846B discloses a method and system for automatic generation and multi-dimensional review of standard documents based on a large model, relating to the field of artificial intelligence technology. The method includes: Step 1, establishing a distributed database of multi-source documents and analyzing heterogeneous texts through natural language processing to obtain a standardized knowledge network; Step 2, extracting indicator elements based on the standardized knowledge network and forming a structured parameter library through verification and validation; Step 3, constructing a template library based on the structured parameter library, analyzing user needs by combining semantic matching, and automatically generating a standard document outline.

[0004] How to select technical solutions with standardization potential from a massive amount of research results, and how to ensure the logical consistency and technical feasibility of standard clauses, are issues that need to be systematically addressed in standardization practice. With the increasing maturity of big data technology and artificial intelligence algorithms, deep deconstruction and quantitative modeling of research results through semantic parsing, knowledge graphs and causal inference, and other means, can realize the intelligent generation, logical verification and dynamic optimization of standard clauses, providing a new technical path to solve the above challenges. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the background technology by proposing an intelligent compilation method and system for drafting scientific research achievement standards based on big data.

[0006] The technical solution of this invention: a method for intelligently compiling draft standards for scientific research results based on big data, comprising the following specific implementation steps: S1. Perform multi-dimensional semantic layering processing on the research results text, identify technical problems, technical means and performance indicators, and construct a causal relationship model to form a technical structure expression model; S2. Based on the technical structure expression model, a standardized adaptation index model is constructed. The model is used to determine whether scientific research results have the conditions to be transformed into standards by quantifying the maturity of technology, public diffusion capability and parameter reproducibility. S3. Using the standardization adaptation index and the technical structure expression model as input, construct the structural vector space of the current standard clauses, identify standard blank units through the isomorphic mapping between the result structure and the current standard clause structure, and filter blank units by combining maturity and public demand and calculate the innovation contribution index. S4. Based on the identified blank units and innovation contribution index, the blank technology structure is transformed into the logical skeleton of standard clauses using the technology path map. The clause content is generated through structured semantic templates, and logical dependency sorting and conflict detection are performed to generate a standard draft. S5. Construct a multi-subject collaborative verification mechanism, encode expert and industry feedback information in a structured manner, transform opinions into calculable parameters and structural adjustment quantities through credibility weighting and influence tensor calculation, and adopt a dual-channel reinforcement update mechanism of parameters and logical structure to dynamically revise the content of the clauses. S6. Establish an implementation monitoring and lifecycle management system to monitor, analyze and evaluate the implementation effects of the draft standard in practical applications in real time, generate adaptive update strategies and control version iteration.

[0007] Preferably, step S1 specifically includes: The research results text is formatted and cleaned. Based on the weighted calculation of semantic vector similarity, contextual continuity and technical term density, the full text is divided into five categories of paragraphs: technical issues, technical means, performance indicators, applicable scenarios and boundary constraints. The frequency of explicit causal dependencies between statistical technical issues and technical methods is used to construct a weighted causal matrix by combining the strength of implicit semantic associations. A technical path graph containing nodes and causal edges is generated by threshold filtering. Extract experimental data from the performance index section, perform outlier cleanup and unit unification, calculate stability coefficient and confidence interval, mark indexes with stability higher than the set threshold as candidates for standardization and generate a set of parameter boundary intervals; Based on the causal relationship matrix, the causal connectivity and the number of index associations of the technical units are statistically analyzed. The weight of the technical units is calculated by combining the scenario coverage. The core units and auxiliary units are distinguished by the weight threshold. The core technical unit set, the auxiliary unit set, the causal relationship matrix and the index interval are integrated to generate a result structure expression model.

[0008] Preferably, step S2 specifically includes: Based on the complete number of statistical problem-method-indicator verification paths for core technology unit sets, and combined with the average causal strength and indicator stability coefficient, the overall maturity index is calculated. The ratio of the number of application scenarios involved in the statistical results to the total number of scenarios in the same field, as well as the overlap ratio of core technology units and industry common technology units, are weighted and integrated to form a public diffusion capability index. The offset between the achievement indicator interval and the reference interval in the standard database is calculated. The parameter replicability index is constructed by the relative difference between the intervals, and the indicators without reference intervals are marked as potential innovation intervals. The boundary clarification coefficient is obtained by calculating the proportion of core technical units that clearly provide information on their applicable scope and boundary constraints. The risk intensity is obtained by assessing the weight of technical units involving security and compliance restrictions based on the risk semantic recognition results. The maturity index, public diffusion capability index, parameter replicability index, boundary clarification coefficient, and risk intensity are weighted and coupled to construct a standardized adaptation index model. When the comprehensive adaptation index exceeds the set threshold, the results are judged to have standardization potential.

[0009] Preferably, step S3 specifically includes: The existing standard database is decomposed at the clause level to construct a clause structure vector containing technical units, performance parameters, testing methods and scope of application, and then summarized to form a standard structure vector space; The core technology units of the achievement are mapped to the standard structure vector space, the intersection ratio of the technology units is calculated to construct a coverage matrix, and the degree of isomorphic overlap between the structure of the achievement and the existing standards is quantified. Based on the coverage matrix, technical units with coverage intensity below the threshold are selected to form a set of structural gaps. The effectiveness of the gaps is calculated by combining the maturity index, public demand index and parameter replicability coefficient. Candidate units of blank clauses with standardization value are selected and prioritized. We construct a technology path map of the achievement and a path map of the existing standards. We calculate the path difference degree by the overlap ratio of the causal relationship set. We couple the path difference degree with the blank priority to form an innovation contribution index, which is used to evaluate the supplementary strength of the achievement to the existing standard system.

[0010] Preferably, step S4 specifically includes: For each blank technical unit, extract the functional objectives, detection methods, applicable scenarios, prerequisites, and data support attributes, and use structured semantic template functions to map them into clause expression templates that conform to standard specifications; Based on the effectiveness and innovation weight of the technology path map and blank units, a clause logical dependency matrix is ​​constructed to generate clause nodes and logical reference edges. The clause order is formed by topological sorting to ensure that the precondition clauses take priority and the detection clauses are placed later. Calculate the consistency coefficient of parameters and the consistency coefficient of detection methods between clauses, generate a conflict index, automatically identify internal and cross-clause conflicts, and provide a backtracking mechanism to adjust parameter ranges and detection methods; The feasibility function of the defined clause is obtained by weighting the maturity of the results, the consistency of the detection method, and the stability of the clause expression structure. The stability index of the defined clause is obtained by coupling the innovation contribution index and the conflict index. When the feasibility and stability index are lower than the set threshold, the template is adjusted and the logical dependency matrix is ​​recalculated. The final standard clause logical skeleton is formed through iterative optimization.

[0011] Preferably, step S5 specifically includes: The feedback information from multiple entities is constructed into a feedback structure vector that includes support level, parameter adjustment suggestions, detection method modification, risk assessment and scope of application suggestions. This vector is then mapped to a set of clause attributes using a clause field mapping function, and a directional consistency coefficient is calculated to identify disputed clauses. A credibility weight model for the subject is established, which calculates the subject's credibility by comprehensively considering years of professional experience, number of historical participations, and the proportion of opinions adopted. A three-dimensional influence tensor of clause-parameter-subject is constructed, and the comprehensive influence matrix of the clause is obtained through tensor compression. A risk amplification factor is introduced to adjust the update intensity of clauses with risks higher than the threshold. A parameter enhancement and update mechanism is adopted, which adjusts the parameter range by weighting according to the comprehensive influence matrix, optimizes the selection of detection method through the method support function, updates the logical dependency weight of the clause, and introduces an innovation protection factor to prevent high innovation clauses from being excessively weakened. Construct a global stability index and a version change rate indicator, set a convergence threshold in conjunction with the proportion of disputed clauses, and automatically freeze the version and generate a version number with a structured hash when continuous iterative changes tend to stabilize.

[0012] Preferably, step S6 specifically includes: Automatically collect data on the execution parameters, detection method effectiveness, application scenario applicability, and expert feedback of the draft standard in practical applications, transform them into structured vectors, and perform abnormal data preprocessing. For each clause, calculate the consistency of execution, compliance with parameters, effectiveness of testing methods, and coverage of applicable scope. Compare with design objectives and standard templates to identify deviations and potential risks. Analyze the impact of logical dependencies between clauses to form a systematic execution effect report. Based on the results of the difference analysis, optimization strategies are developed for generating parameter range correction, detection method optimization, and scope of application expansion for clauses with execution deviations exceeding the set threshold. Optimization schemes are also developed for generating structure optimization schemes for high-risk clauses, and priority update strategies are formulated based on the consistency of clause execution and the degree of innovation contribution. New versions are generated under the guidance of adaptive strategies, and the basis for updates and related monitoring data are recorded to achieve version traceability. Multiple versions are supported for parallel management and difference visualization. The convergence of the logical skeleton of the new version terms is evaluated to ensure that the update process is safe and controllable.

[0013] Preferably, the comprehensive fit index in the standardized fit index model is calculated by weighted coupling based on the maturity index, public diffusion capability index, parameter replicability index, boundary clarity coefficient, and risk intensity. The weight coefficients were obtained through training on historical annotated corpora and experimental data. When the overall fit index exceeds the set standardization threshold, the standard blank analysis phase begins.

[0014] Preferably, the generation of the standard clause logic skeleton includes: A clause logical dependency matrix is ​​constructed based on the technology path graph, the clause nodes are topologically sorted to determine the clause order, and the conflict index between clauses is calculated. The generated clauses are iteratively evaluated by executing a feasibility function and a clause stability index. The feasibility function is obtained by weighting the maturity of the results, the consistency of the clause detection method, and the stability of the clause expression structure. The clause stability index is obtained by coupling the innovation contribution index and the conflict index. When the feasibility or stability index falls below a set threshold, the process returns to adjusting the structured semantic template or recalculating the logical dependency matrix until the iterative convergence condition is met.

[0015] The technical solution of this invention: an intelligent drafting system for scientific research achievement standards based on big data, comprising: Memory; processor; A computer program stored in the memory and capable of running on the processor; When the processor executes the computer program, it implements the above-mentioned intelligent compilation method for drafting scientific research achievement standards based on big data.

[0016] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: This invention designs an intelligent method and system for drafting standards for scientific research achievements based on big data. This represents a fundamental transformation in the drafting of standards from research findings, shifting the traditional subjective construction process reliant on expert experience to an objective, intelligent generation process driven by data and models. This significantly improves the efficiency and scientific rigor of standardization work. By automatically parsing unstructured research text into a computable technical structure model, the system can deeply explore the inherent causal logic between technical problems, methods, and performance indicators, ensuring that the subsequently generated clauses have solid technical basis and rigorous logical consistency. The introduction of a multi-dimensional coupled standardization adaptation index provides a quantitative basis for determining whether an achievement has the potential to be transformed into a standard, effectively avoiding the blind spots in standardization work. By constructing a standard structure vector space and performing isomorphic mapping, the system can accurately identify gaps and deficiencies in the existing standard system and combine them with innovation. The contribution index assessment results provide supplementary value, thus offering forward-looking decision support for the formulation and revision of standards. During the clause generation phase, topological sorting and conflict detection mechanisms ensure the correctness of logical dependencies between clauses and the stability of the overall structure. Combined with structured semantic templates, the generated draft standards conform to both standardized formats and are executable. A multi-stakeholder collaborative verification mechanism structures and quantitatively analyzes feedback from experts, industry, and other diverse stakeholders, enabling continuous optimization of the draft standards through dynamic consensus and ensuring the credibility and applicability of the standards. Finally, the implementation monitoring and adaptive iteration capabilities throughout the entire lifecycle allow standards to evolve continuously based on actual application effects, maintaining their technological advancement and practicality. This constructs a data-driven, logically rigorous, and dynamically optimized closed-loop system for standard development and maintenance, providing strong technical support for the deep integration of technological innovation and industrial development. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for intelligent compilation of draft standards for scientific research results based on big data, as proposed in this invention. Detailed Implementation

[0018] Example 1, as Figure 1 As shown, the present invention proposes an intelligent drafting method for scientific research achievement standards based on big data, which includes the following specific implementation steps: S1. Transform the text of scientific research results into a computable technical structure expression model. Through multi-dimensional semantic layering, causal relationship modeling, performance index stability analysis, and structural compression and reconstruction, a quantitative basis for mapping scientific research results to standard clauses is realized. The specific implementation process is as follows: S11. The uploaded research results text is formatted and cleaned, and the full text is divided into five categories of paragraphs: technical issues, technical methods, performance indicators, applicable scenarios, and boundary constraints. The functional role of each paragraph is identified through weighted calculations based on semantic vector similarity, contextual continuity, and technical terminology density, forming a set of five paragraph categories. This provides a functional node foundation for subsequent causal modeling. Specifically: Upload the text of your research findings and preprocess the acquired text, including but not limited to: formatting the uploaded text (PDF, Word, LaTeX, etc.), identifying chapter titles, paragraph structure, and figure content, and extracting the main text; cleaning the text, removing references, footnotes, and duplicate descriptions to ensure the integrity of the core technical content; The full text is divided into five core functional paragraphs: technical problem paragraph (describing scientific research problems and unresolved technical bottlenecks), technical means paragraph (innovative methods, experimental schemes, and solutions), performance index paragraph (quantifying results, performance parameters, and experimental data), applicable scenario paragraph (application conditions, environmental limitations, and implementation scope), and boundary constraint paragraph (limiting conditions, dependent parameters, and special requirements). Calculate a matching score for each text segment: ; For each text segment, calculate the matching score for five categories of functions, select the category with the highest score as the category, and output the five categories of paragraph sets to form a preliminary functional matrix. in, This represents the functional role determination value of the i-th text segment; The cosine similarity between the paragraph semantic vector and the functional template vector is used to represent the similarity between the two vectors. The score represents the continuity with the preceding and following paragraphs (obtained through sentence causal words and dependency parsing); Indicates the density of technical terms (the number of domain-specific terms in each paragraph divided by the total number of words); , and These represent the weighting coefficients, which were obtained through training with historical annotated corpora and experimental data. S12. Based on the functional paragraph set, calculate the frequency of explicit causal dependencies and the strength of implicit semantic associations between technical problems and technical means, construct a weighted causal matrix, use threshold filtering to generate a technical path diagram, and connect technical problems, means, and performance indicator nodes to form the core technical path, specifically: By analyzing the relationship between technical problems, technical means, and performance indicators, a causal dependency matrix is ​​constructed to quantitatively describe the technical paths in scientific research results. ; Traverse each technical issue paragraph and technical means paragraph, and count the explicit logical connections; calculate the semantic similarity of potential implicit connections, and sum them with the explicit dependencies in a weighted manner; Set threshold ,like Then establish a matrix connection; The causal dependency matrix is ​​transformed into a directed graph, where nodes represent technical problems, technical methods, and performance indicators; edges represent... causal connections; in, This indicates the causal strength between technical problem j and technical means k; Indicates the frequency of explicit causal dependencies in the text; Indicates the strength of implicit semantic association; and These represent the normalized weights; S13. Extract performance index data, perform outlier cleanup and unit standardization, calculate stability coefficients and confidence intervals, select highly stable indicators as candidates for standardization, and generate a set of parameter boundary intervals to provide a quantifiable parameter basis for clause generation. Specifically: Extract performance index data, identify experimental data, measured values ​​and parameter ranges from the performance index paragraphs, and perform unit unification and outlier cleaning on the indexes; Calculate the stability coefficient: ; When the stability coefficient of a certain index Greater than the set threshold When this condition is met, the indicator is considered a candidate for standardization. Simultaneously calculate the indicator boundary interval: ; in, Represents the stability coefficient; Indicates the standard deviation of the indicator; This represents the mean of the indicator; represents the confidence amplification factor, set by empirical values ​​in the field; B represents the acceptable range of the indicator. S14. Calculate the weight of each technical unit by combining the causal relationship matrix and indicator intervals. Distinguish between core and auxiliary units through weight thresholds. Integrate core technical units, auxiliary units, indicator intervals, and causal relationships to generate a result structure expression model. This model can be directly mapped to the standard clause structure, realizing the quantification and verifiable transformation of scientific research results into standard drafts. Specifically: After obtaining the technology roadmap and indicator ranges, the technology structure is compressed and reconstructed to form an "achievement technology structure expression model," and the weights of the technology units are defined: ; When the weight of a certain technical unit is below the compression threshold, it is marked as an auxiliary unit; those above the threshold become core technical units; finally, the result structure expression model M is constructed: ; in, Indicates the weight of the technical unit; Indicates the degree of causal connectivity (statistically determined by the causal relationship matrix R); Indicates the number of related indicators; Indicates scene coverage; , and These represent the weighting coefficients; Represents a collection of core technology units; R represents the set of auxiliary technology units; R represents the causal relationship matrix.

[0019] S2. Based on the achievement structure expression model M formed in step S1, a multi-dimensional coupled standardized adaptation index model is established by constructing a structural maturity index, a public diffusion capability index, a parameter reproducibility index, and a risk stability constraint factor. This model is used to quantitatively determine whether research results meet the conditions for transformation into standards. The specific implementation process is as follows: S21. Based on the core technology unit set, causal strength matrix, and indicator stability coefficient output in step S1, construct a technology closed-loop completeness model. Verify the formation degree of the problem-method-indicator path through statistical analysis, and combine the average causal strength with indicator stability for coupled calculation to form an overall maturity index. This index is used to determine whether the technology has reached a standardized structural maturity stage. Specifically: For each core technology unit Define closed-loop integrity : ; Based on the reliability of the technical path reflected by the intensity values ​​in the causality matrix R, the average causal intensity is defined as follows: ; Building a maturity index And calculate the final overall maturity index. : ; in, Technical unit representation A complete path of "problem-method-indicator verification" has been formed; This represents the number of complete paths that the theory should form (derived from the causal matrix R). This indicates the causal strength calculated in step S1; Indicates the number of path connections; S22. Based on the applicable scenarios and reusability characteristics of the technical structure, calculate the industry coverage ratio and the overlap ratio of core technology units in the common structure of the industry. Combine this with the completeness of the technology closed loop for weighted fusion to form a public diffusion capability index. This index is used to measure whether the achievement has the necessity and practical demand basis for cross-scenario promotion and the formulation of public standards. Specifically: Define the breadth index of scene coverage ; Simultaneously define the reusability of the technical structure. ; The diffusion coupling index is obtained by combining the results: ; in, This indicates the number of application scenarios involved in the results; This represents the total number of scenarios in the same domain within the standard database; This represents a set of common technology units in the industry; , and Indicates the weighting coefficient; S23. Based on the stable performance index interval obtained in step S1, calculate the offset between it and the reference interval in the standard database. Construct a reproducibility index through the relative difference between the intervals, and distinguish between cases with and without a reference interval to identify feasible parameters and potential innovative parameters, specifically: Based on the set of indicator intervals B obtained in step S1, a replicability verification is performed, and the consistency offset of the indicator intervals is defined: ; And further define the replicability index: ; It should be noted that when the replicability index... When the value is close to 1, it indicates that the achievement indicator has a high degree of consistency with industry data; if a certain indicator range has no reference range at all, it is marked as a "potential innovation range" and enters the subsequent blank analysis. in, Indicates the range of performance indicators; Indicates the reference range of the standard database; S24. A standardized adaptation index model is constructed by weighting and coupling the maturity index, public diffusion capability index, parameter replicability index, boundary clarification, and risk intensity. When the comprehensive index reaches a set threshold, the model proceeds to the subsequent standard structure mapping stage, thus forming a structured, quantifiable, and verifiable standard transformation decision-making mechanism. Construct a comprehensive adaptation index: ; in, Indicates the boundary clarification coefficient; Indicates the intensity of risk; , , , and Indicates the weighting coefficient; It should be noted that the boundary clarification coefficient By statistically analyzing whether core technology units explicitly specify constraints such as applicable scope, environmental conditions, upper and lower limits of parameters, and exclusion scenarios, the system identifies and matches the boundary constraint sets in the output structure model, calculating the proportion of technology units with clearly defined boundaries to form a coefficient; risk intensity. Based on the results of risk semantic recognition, the number of technical units involved in safety, environmental protection, compliance restrictions and potential failure scenarios is counted, and a weighted evaluation is carried out in combination with their weight in the causal structure, so as to reflect the level of potential uncertainty of the results in the standardization process. When the comprehensive fit index If the set standardization threshold is exceeded, the standard blank analysis phase will begin.

[0020] S3. Using the standardized fit index and outcome structure model output from step S2 as input, construct a standard clause structure vector space. Calculate the coverage relationship through isomorphic mapping between the outcome structure and the existing standard clause structure. Based on this, identify uncovered or poorly covered structural gap units, and screen them using maturity and public demand indices. Further calculate the innovation contribution intensity through technical path difference, ultimately forming a candidate set of standard blank clauses and an innovation analysis report. The specific implementation process is as follows: S31. Decompose the existing standards at the clause level, construct clause structure vectors containing technical units, performance parameters, testing methods, and scope of application, and summarize them to form a standard structure vector space. This transforms the standard system from a collection of texts into a computable structural model, providing a unified structural coordinate system for subsequent structural mapping of results. Specifically: First, the standard database is restructured, and the structure vector of a certain standard clause is defined as follows: Then, a set of clause vectors is constructed for each standard to form a standard structure vector space; in, This refers to the set of technical units involved in clause j; This represents the set of performance parameters corresponding to clause j; This refers to the set of testing or verification methods specified in clause j; This indicates the set of scenarios or scope of application for clause j; S32. Map the set of core technical units of the achievement to the vector space of the standard clause structure, construct a coverage matrix by calculating the intersection ratio of technical units, quantify the degree of isomorphic overlap between the achievement structure and the existing standard, identify covered structures and potentially uncovered structures, and lay the foundation for gap location, specifically: In step S1, the resulting structural model M has been formed, and the structural mapping function has been defined: ; When the overlap ratio If the threshold is exceeded, it is determined that part of the structure of the result has been covered by the current standard; Further construct the mapping matrix (structure covering matrix): ; in, This represents the set of core technology units of the achievement; This represents a set of technical units for a certain standard clause; This represents the set of technical units representing result i; This represents the set of technical units in standard clause i; S33. Based on the coverage matrix, select technical units with coverage intensity below the threshold to form a structural gap set. Combine this with the maturity index, public demand index, and parameter replicability coefficient to calculate the gap priority and select candidate units of gap clauses with standardization value, specifically: After obtaining the structural coverage matrix, identify the set of technical units that are not covered or whose coverage intensity is below a threshold: ; Then, by combining the reproducibility vector and maturity parameter in step S2, the gap units are filtered, and only units with maturity and reproducibility higher than the set threshold are retained as standard blank candidate units. Further define the validity of the gap: ; When the gap is valid When the value is above the threshold, the missing cell is marked as a normalizable blank cell; Further hierarchical division of blank units: Level 1 Blank: Completely uncovered and highly mature; Level 2 gap: Partial coverage but innovative parameter range; Level 3 Gaps: Innovative approaches but low public demand; in, This represents a set of structural gap technology units; Indicates the coverage determination threshold; This represents the maturity value of the binding element unit; , and Indicates the weighting coefficient; It should be noted that, in order to avoid partial coverage being misjudged as overall coverage, three types of coverage are distinguished: complete coverage, which means that both the parameter range and the detection method match; partial coverage, which means that the technical units match but the parameter ranges are different; and semantic coverage, which means that the technical paths are similar but the structures are different. S34. Construct a technical path difference map between the achievements and existing standards, calculate the path difference degree by the overlap ratio of the causal relationship set, and couple it with the gap priority to form an innovation contribution index. This assesses the strength of the achievements' supplementation to the existing standards system and their innovative value from a structural perspective. Specifically: To further distinguish between structural deficiencies and genuine innovation, a technology path map G is constructed based on the causal relationship matrix R in step S1. c ; Define a set of paths: ; Further define the standard path set: ; Calculate path difference: ; If the degree of difference is higher than the set threshold and the maturity is higher than the set threshold, it indicates that the results have substantial innovation at the structural path level. Further define the innovation contribution index: ; in, and This represents the weighting coefficient.

[0021] S4. Based on the hierarchical blank units, innovation contribution index, and technology path map output in step S3, the blank technology structure is transformed into a logical skeleton of standard clauses. Through structured semantic template generation, logical skeleton assembly, consistency and conflict detection, and iterative optimization of stability and executability, the standard clauses are automatically generated, logically closed-loop, self-consistently optimized, and innovation-executability balanced, forming a feasible and standardized draft standard. The specific implementation process is as follows: S41. By extracting attributes such as function, parameters, detection method, applicable scenarios, prerequisites, and data support for each blank technical unit, a clause expression template conforming to the standard specification is automatically generated using a structured semantic template function, while ensuring the rationality and executability of parameter ranges. Specifically: Set up structural gap technology units This is transformed into a "structured semantic template" for clause generation, ensuring that the technical information of each blank unit can be directly mapped to standard clause fields; For each blank cell Constructing a property collection: ; Based on this, a structured semantic template function is introduced. ; in, The set of attributes representing the blank unit i; Indicates functional objectives, describing the technical function or target behavior represented by the blank unit, such as "data storage security" or "signal and noise filtering"; This refers to a set of testing / verification methods used to verify functional objectives or performance indicators, such as "experimental measurement," "simulation verification," and "historical data backtesting." Indicates the applicable scenario / constraints; This indicates the preceding dependent units, that is, the list of preceding units that the current blank unit depends on in the technical path or logical flow; This indicates the degree to which the blank cell can be supported by standardized experimental or literature data, with a value ranging from 0 to 1; This represents a structured semantic mapping generation function; It should be noted that the structured semantic mapping generation function Its core function is to transform the attribute set of blank technical units into a clause expression skeleton that conforms to the standard specifications. Based on the expression patterns, field order rules and logical dependencies in the historical standard clause structure database, it performs semantic rearrangement and hierarchical organization of functional objectives, performance parameters, detection methods and applicable conditions. At the same time, it automatically adjusts the rigor and limitation of the clause expression by combining the preceding dependencies and the strength of data support, thereby generating a standard clause structure template with standard consistency, logical closure and executability. S42. By utilizing the technology path map, the validity of blank units, and innovation weights to construct a clause logical dependency matrix, clause nodes and logical edges are generated. Topological sorting is used to form the clause order, ensuring that clauses with preconditions take priority and checking clauses with postconditions. At the same time, the logical sorting is optimized by combining innovation weights to achieve automated skeleton construction of consistent technical causality among clauses. Specifically: Based on the technology roadmap G generated in step S3 c Construct a logical framework for the clauses to ensure consistency in cause and effect, parameters, and detection methods among the clauses, thus forming an executable standard framework; Construct the clause logical dependency matrix: ; According to the clause logical dependency matrix ; The clause nodes are then topologically sorted to ensure that prerequisite technical conditions appear first, while clauses for detection and verification are placed later. ; Where S represents the clause template node; Indicates a logical reference edge; This represents the logical dependency weight matrix of clauses, which measures the strength of the logical reference of clause i to clause j. Indicates the causal strength of the technical path; This represents the validity value of the blank unit, i.e., the validity score of the blank unit, which takes a value of 0-1, taking into account maturity, common needs, and parameter consistency. This represents the innovation contribution weight, the innovation contribution index of blank cell i, which takes a value of 0-1 and is derived from the innovation contribution index in step S3; This represents the framework diagram of the clauses; This indicates the topological sorting result, specifically the order in which the clauses are arranged in the draft standard. Represents the topological sorting function; It should be noted that the topological sorting function This function is used for topological sorting of the clause logic skeleton diagram. Its core function is to determine a reasonable generation order among a set of clauses with dependencies. The function takes clause nodes and their logical reference relationships as input. First, it identifies the starting node without any preceding dependencies and outputs it first. Then, it gradually removes the associated edges of the output nodes and dynamically updates the dependency state of the remaining nodes, continuously selecting nodes with no unmet dependencies to add to the sequence. If a circular dependency is detected, a conflict backtracking mechanism is triggered to adjust the structure. Through this process, it ensures that the clause arrangement order conforms to the technical causal logic and the execution sequence. S43. By calculating the consistency of parameters, the consistency of testing methods, and the conflict index among clauses, internal or cross-clause conflicts are automatically identified, and a backtracking mechanism is provided to adjust parameter ranges or testing methods. This ensures that the generated clauses are self-consistent in logic, parameters, and testing methods, guaranteeing the scientific rigor and feasibility of the draft standard. Specifically: Perform parameter consistency checks: ; Consistency of detection methods: ; Generation Conflict Index: ; in, The parameter consistency coefficient measures the interval consistency between clause i and clause j for the same performance parameter. Indicates consistency in testing methods, measuring the proportion of overlap between the testing / verification methods used in clauses i and j; Indicates the conflict index; S44. The clauses are iteratively evaluated and optimized using the innovation contribution index, clause conflict index, and execution feasibility function. Highly innovative and executable clauses are selected, while unstable or conflicting clauses are revised, forming the final standard clause logical framework. This framework supports outputting Word, PDF, and online editable drafts, achieving a balance between innovation and feasibility. Specifically: Define the feasibility function for clause execution: ; And define the terms stability index: ; like or If the terms are returned to step S1, the template needs to be adjusted, or the logical dependency matrix needs to be recalculated. It should be noted that for the selected clauses, a complete clause logic skeleton is generated, supporting draft output formats: Word and PDF, and retaining an online editable version; in, Indicates the feasibility of enforcing clause i; This indicates the maturity of the results, originates from step S2, and measures the feasibility of the technical unit in practical applications. This indicates that the testing methods for the clauses are consistent; This indicates the stability of the clause expression structure and measures whether the template function output is consistent with historical standards. , and This represents the weighting coefficient, which can be adjusted according to system requirements to meet... ; Indicates the stability index of the terms; This represents the innovation contribution index calculated in step S3; Indicates the conflict index; This represents the feasibility threshold, the minimum score required for a clause to be implemented. This represents the stability index threshold, ensuring that the clause logic is consistent and can be standardized.

[0022] S5. Construct a multi-entity collaborative verification and feedback-enhanced evolution mechanism. This mechanism structures and encodes feedback information from multiple sources, including experts, industry units, and testing institutions. Through credibility weighting and influence tensor calculation, opinions are transformed into calculable parameters and structural adjustment quantities. A dual-channel enhancement and update mechanism using parameters and logical structure is employed to dynamically revise the clause content. Furthermore, convergence judgment and version freeze control ensure the stable evolution of the draft standard, enabling the standard generation process to possess data-driven, adaptive optimization, and traceable solidification capabilities. The specific implementation process is as follows: S51. By constructing a multi-subject feedback structure vector, support, parameter adjustment suggestions, detection method modifications, risk assessment, and scope of application suggestions are mapped to clause field data objects. A directional consistency coefficient is introduced to identify disputed clauses, thus realizing the transformation of opinions from text to structured data. Specifically: Feedback from multiple stakeholders, including expert committees, industry enterprises, testing institutions, and pilot application units, is transformed into data objects that can directly affect the clause structure parameters and logical dependency matrix. A set U of participating stakeholders is defined, and the impact of each stakeholder on the clauses is further analyzed. The feedback is defined as: ; Set the clause field mapping function: This maps feedback directly to the set of terms attributes. ; When different entities have opposite directions toward the same parameter, a direction consistency coefficient is introduced: ; like If the value is below the threshold, the clause is marked as a "disputed clause" and enters the enhanced analysis phase; in, Representing the subject The feedback structure vector for the k-th clause; This indicates the overall level of support the subject has for the clause; This represents the parameter range adjustment vector suggested by the main body, i.e., the range offset value or the upper and lower limit adjustment amount; Indicates suggestions for modifying the testing method; This indicates the entity's assessment of the risk level of the terms, i.e., the risk assessment score; Suggestions for adjusting applicable scenarios; This represents the entity's reassessment of the technology maturity level; Indicates the function mapping terms fields; Indicates the consistency coefficient of feedback direction; It should be noted that the clause field mapping function This means that structured feedback information submitted by multiple entities will be automatically matched to the specific attribute fields of the corresponding clauses. Through a preset field semantic rule base and parameter type recognition mechanism, the field positioning and consistency correction will be performed on the parameter adjustment suggestions, detection method modifications, changes in applicable scenarios and risk assessments in the feedback, so as to avoid the bias of manual classification and ensure that all feedback can accurately affect the functional objectives, performance range or logical dependency structure of the clauses, so as to realize the computable transformation of feedback data into clause structure. S52. Establish a credibility weighting model for the main body, calculate credibility by comprehensively considering professional years, historical participation frequency, and opinion adoption rate, and construct a three-dimensional influence tensor of clause-parameter-main body to form a differentiated influence matrix. Simultaneously, introduce a risk amplification factor to adjust the update intensity of high-risk clauses, achieving quantitative modeling of feedback influence. Specifically: Define the subject's credibility weight: ; Construct a three-dimensional influence tensor for the set of terms: ; The combined effect matrix of the clauses is obtained through tensor compression: ; And obtain the risk amplification factor: ; in, Indicates agreement with the terms The number of entities that submitted valid feedback; Indicates the credibility weight of the subject; , and This represents the weight coefficients of the credibility model; This represents the normalized value of the subject's years of professional experience; Indicates the number of times the subject has participated in the standard historical data; This indicates the percentage of past opinions that were adopted by the main body; This represents the three-dimensional feedback influence tensor, where k is the clause number, p is the clause field number, and u is the main body number; This indicates the overall strength of the impact of clause p; Indicates the risk amplification factor of the terms; S53. A dual-channel reinforcement mechanism of parameter update and logical dependency weight update is adopted. The parameter range is weighted and corrected according to the comprehensive influence matrix, and the selection of detection method is optimized. At the same time, the weight of the logical skeleton of the clause is adjusted, and an innovation protection factor is introduced to prevent high innovation clauses from being excessively weakened, so as to achieve structural dynamic evolution. Specifically: Setting parameters to enhance the update mechanism: ; Construction method support function: ; And update the logical structure weights: ; Introducing innovation protection factors: ; in, Indicates support level, Representative subject u supports method m, The representative subject u does not support method m; Indicates the original parameter range of the clause; This indicates the parameter range after the enhancement and update; This indicates the reinforcement learning step size, controlling the magnitude of parameter adjustment. This represents the overall support of detection method m; This indicates the logical dependency weight of the original clause; This indicates the updated logical dependency weight; This represents the logical weighting enhancement coefficient; The overall strength of support for the terms is derived from a comprehensive calculation of the field influence matrix; Indicates the innovation protection factor; The innovation contribution index of the clauses (from step S3); the degree of controversy indicates low consistency in direction or a degree of dispersion in support; S54. Construct a global stability index and version change rate indicator, and set a convergence threshold based on the proportion of disputed clauses. When continuous iterative changes tend to stabilize, automatically freeze the version and generate a version number with a structured hash, thereby achieving traceable solidification and lifecycle management control of the draft standard. Specifically: Constructing a global stability index: ; Set version change rate: ; When two consecutive iterations satisfy: Increase < , The proportion of disputed clauses If the standard draft is found to be converged, a version number will be generated. in, Indicates the total number of clauses in the current draft; This indicates the overall stability index of the terms; Indicates the version change rate; , and These represent the convergence thresholds, corresponding to: stability index change threshold, parameter change rate threshold, and disputed clause proportion threshold, respectively.

[0023] S6. By conducting real-time monitoring, quantitative analysis, and difference assessment of the implementation effects of the draft standard in actual pilot projects, industry applications, and expert reviews, an adaptive update strategy is generated. Through version iteration and lifecycle management, the draft standard is continuously optimized throughout the entire implementation cycle, forming a traceable, controllable, and dynamically evolving closed-loop management system to achieve the long-term adaptability and practicality of the standard. The specific implementation process is as follows: S61. Automatically collect multi-source data, including execution parameters, detection method effectiveness, application scenario applicability, and expert feedback, and transform it into structured vectors. Preprocess abnormal or missing data to achieve quantifiable monitoring of the clause execution status. Specifically: Automatically collect multi-source data, including equipment testing data, experimental verification results, enterprise operation feedback, and expert qualitative evaluation; The collected data is transformed into a unified structured format, such as the execution indicator vector corresponding to each clause, including but not limited to parameter deviation, method success rate, number of risk events, and coverage of applicable scenarios; Preprocessing of abnormal or missing data, through data interpolation, outlier filtering, and standardization, ensures reliable input data quality; S62. Evaluate the consistency of clause execution, parameter compliance, methodological effectiveness, and scope of application, and compare them with design objectives and standard templates to identify deviations and potential risks. Simultaneously, analyze the impact of logical dependencies between clauses, and generate a systematic execution effectiveness report to provide data support for adaptive updates. Specifically: For each clause, calculate its consistency in execution, compliance with parameters, effectiveness of testing methods, and coverage of applicable scope; Compare the monitoring data with the target values ​​and standardized templates from the design phase to identify deviations and areas for improvement; Cross-analysis of the interrelationships between clauses is conducted to identify systemic deviations caused by poor implementation of a particular clause, ensuring that the problem is not limited to a single clause. Generate a structured assessment report, including but not limited to performance deviation analysis, potential risk warnings, and optimization suggestions; S63. Based on the difference analysis results, generate optimization strategies for clause parameters, logical structure, and detection methods, prioritizing high-value and innovative clauses. Implement automatic or manual confirmation and updates through executable instructions, enabling the standard to dynamically adapt to the environment and user feedback during implementation, maintaining a balance between innovation and feasibility. Specifically: For clauses with significant execution deviations (above the set threshold), parameter adjustment suggestions are generated, including parameter range correction, detection method optimization, or scope of application expansion. Generate structural optimization solutions for high-risk or logically conflicting clauses, such as adjusting the order of clauses, modifying dependencies, or adding preconditions; Based on the consistency of clause execution and the degree of innovative contribution, a priority update strategy is developed to ensure that high-value clauses are optimized first. The update strategy is encoded into executable instructions that can be applied automatically or provided to standard administrators for verification. S64. Generate new versions under the guidance of adaptive strategies, record the basis for updates and historical data, achieve version traceability, parallel management, and difference visualization, and evaluate the logical convergence of clauses to ensure the optimization process is safe and controllable, forming a long-term iterative standard lifecycle management system, specifically: Generate a new version number, record the update content, update basis and related monitoring data to form a complete version traceability; Supports parallel management of multiple versions, including historical version backtracking, comparative analysis, and difference visualization; A convergence assessment was conducted on the logical framework of the new version's terms to ensure that the update would not produce logical conflicts or systemic deviations. It can provide users with real-time updated results through online editing, automatic generation of Word / PDF and visual logic diagrams.

[0024] Example 2: The present invention proposes an intelligent drafting system for scientific research achievement standards based on big data, which is used to execute the intelligent drafting method for scientific research achievement standards based on big data proposed in Example 1, and includes: Memory; processor; A computer program stored in the memory and capable of running on the processor; The processor executes a computer program to implement the intelligent compilation method for drafting scientific research results standards based on big data, as described in Embodiment 1 above.

[0025] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for intelligently drafting standards for scientific research results based on big data, characterized in that: The specific implementation steps include the following: S1. Perform multi-dimensional semantic layering processing on the research results text, identify technical problems, technical means and performance indicators, and construct a causal relationship model to form a technical structure expression model; S2. Based on the technical structure expression model, a standardized adaptation index model is constructed. The model is used to determine whether scientific research results have the conditions to be transformed into standards by quantifying the maturity of technology, public diffusion capability and parameter reproducibility. S3. Using the standardization adaptation index and the technical structure expression model as input, construct the structural vector space of the current standard clauses, identify standard blank units through the isomorphic mapping between the result structure and the current standard clause structure, and filter blank units by combining maturity and public demand and calculate the innovation contribution index. S4. Based on the identified blank units and innovation contribution index, the blank technology structure is transformed into the logical skeleton of standard clauses using the technology path map. The clause content is generated through structured semantic templates, and logical dependency sorting and conflict detection are performed to generate a standard draft. S5. Construct a multi-subject collaborative verification mechanism, encode expert and industry feedback information in a structured manner, transform opinions into calculable parameters and structural adjustment quantities through credibility weighting and influence tensor calculation, and adopt a dual-channel reinforcement update mechanism of parameters and logical structure to dynamically revise the content of the clauses. S6. Establish an implementation monitoring and lifecycle management system to monitor, analyze and evaluate the implementation effects of the draft standard in practical applications in real time, generate adaptive update strategies and control version iteration.

2. The intelligent compilation method for draft standards of scientific research results based on big data according to claim 1, characterized in that, Step S1 specifically includes: The research results text is formatted and cleaned. Based on the weighted calculation of semantic vector similarity, contextual continuity and technical term density, the full text is divided into five categories of paragraphs: technical issues, technical means, performance indicators, applicable scenarios and boundary constraints. The frequency of explicit causal dependencies between statistical technical issues and technical methods is used to construct a weighted causal matrix by combining the strength of implicit semantic associations. A technical path graph containing nodes and causal edges is generated by threshold filtering. Extract experimental data from the performance index section, perform outlier cleanup and unit unification, calculate stability coefficient and confidence interval, mark indexes with stability higher than the set threshold as candidates for standardization and generate a set of parameter boundary intervals; Based on the causal relationship matrix, the causal connectivity and the number of index associations of the technical units are statistically analyzed. The weight of the technical units is calculated by combining the scenario coverage. The core units and auxiliary units are distinguished by the weight threshold. The core technical unit set, the auxiliary unit set, the causal relationship matrix and the index interval are integrated to generate a result structure expression model.

3. The intelligent compilation method for draft standards of scientific research results based on big data according to claim 2, characterized in that, Step S2 specifically includes: Based on the complete number of statistical problem-method-indicator verification paths for core technology unit sets, and combined with the average causal strength and indicator stability coefficient, the overall maturity index is calculated. The ratio of the number of application scenarios involved in the statistical results to the total number of scenarios in the same field, as well as the overlap ratio of core technology units and industry common technology units, are weighted and integrated to form a public diffusion capability index. The offset between the achievement indicator interval and the reference interval in the standard database is calculated. The parameter replicability index is constructed by the relative difference between the intervals, and the indicators without reference intervals are marked as potential innovation intervals. The boundary clarification coefficient is obtained by calculating the proportion of core technical units that clearly provide information on their applicable scope and boundary constraints. The risk intensity is obtained by assessing the weight of technical units involving security and compliance restrictions based on the risk semantic recognition results. The maturity index, public diffusion capability index, parameter replicability index, boundary clarification coefficient, and risk intensity are weighted and coupled to construct a standardized adaptation index model. When the comprehensive adaptation index exceeds the set threshold, the results are judged to have standardization potential.

4. The intelligent compilation method for draft standards of scientific research results based on big data according to claim 3, characterized in that, Step S3 specifically includes: The existing standard database is decomposed at the clause level to construct a clause structure vector containing technical units, performance parameters, testing methods and scope of application, and then summarized to form a standard structure vector space; The core technology units of the achievement are mapped to the standard structure vector space, the intersection ratio of the technology units is calculated to construct a coverage matrix, and the degree of isomorphic overlap between the structure of the achievement and the existing standards is quantified. Based on the coverage matrix, technical units with coverage intensity below the threshold are selected to form a set of structural gaps. The effectiveness of the gaps is calculated by combining the maturity index, public demand index and parameter replicability coefficient. Candidate units of blank clauses with standardization value are selected and prioritized. We construct a technology path map of the achievement and a path map of the existing standards. We calculate the path difference degree by the overlap ratio of the causal relationship set. We couple the path difference degree with the blank priority to form an innovation contribution index, which is used to evaluate the supplementary strength of the achievement to the existing standard system.

5. The intelligent compilation method for draft standards of scientific research results based on big data according to claim 4, characterized in that, Step S4 specifically includes: For each blank technical unit, extract the functional objectives, detection methods, applicable scenarios, prerequisites, and data support attributes, and use structured semantic template functions to map them into clause expression templates that conform to standard specifications; Based on the effectiveness and innovation weight of the technology path map and blank units, a clause logical dependency matrix is ​​constructed to generate clause nodes and logical reference edges. The clause order is formed by topological sorting to ensure that the precondition clauses take priority and the detection clauses are placed later. Calculate the consistency coefficient of parameters and the consistency coefficient of detection methods between clauses, generate a conflict index, automatically identify internal and cross-clause conflicts, and provide a backtracking mechanism to adjust parameter ranges and detection methods; The feasibility function of the defined clause is obtained by weighting the maturity of the results, the consistency of the detection method, and the stability of the clause expression structure. The stability index of the defined clause is obtained by coupling the innovation contribution index and the conflict index. When the feasibility and stability index are lower than the set threshold, the template is adjusted and the logical dependency matrix is ​​recalculated. The final standard clause logical skeleton is formed through iterative optimization.

6. The intelligent compilation method for draft standards of scientific research results based on big data according to claim 5, characterized in that, Step S5 specifically includes: The feedback information from multiple entities is constructed into a feedback structure vector that includes support level, parameter adjustment suggestions, detection method modification, risk assessment and scope of application suggestions. This vector is then mapped to a set of clause attributes using a clause field mapping function, and a directional consistency coefficient is calculated to identify disputed clauses. A credibility weight model for the subject is established, which calculates the subject's credibility by comprehensively considering years of professional experience, number of historical participations, and the proportion of opinions adopted. A three-dimensional influence tensor of clause-parameter-subject is constructed, and the comprehensive influence matrix of the clause is obtained through tensor compression. A risk amplification factor is introduced to adjust the update intensity of clauses with risks higher than the threshold. A parameter enhancement and update mechanism is adopted, which adjusts the parameter range by weighting according to the comprehensive influence matrix, optimizes the selection of detection method through the method support function, updates the logical dependency weight of the clause, and introduces an innovation protection factor to prevent high innovation clauses from being excessively weakened. Construct a global stability index and a version change rate indicator, set a convergence threshold in conjunction with the proportion of disputed clauses, and automatically freeze the version and generate a version number with a structured hash when continuous iterative changes tend to stabilize.

7. The intelligent compilation method for draft standards of scientific research results based on big data as described in claim 6, characterized in that, Step S6 specifically includes: Automatically collect data on the execution parameters, detection method effectiveness, application scenario applicability, and expert feedback of the draft standard in practical applications, transform them into structured vectors, and perform abnormal data preprocessing. For each clause, calculate the consistency of execution, compliance with parameters, effectiveness of testing methods, and coverage of applicable scope. Compare with design objectives and standard templates to identify deviations and potential risks. Analyze the impact of logical dependencies between clauses to form a systematic execution effect report. Based on the results of the difference analysis, optimization strategies are developed for generating parameter range correction, detection method optimization, and scope of application expansion for clauses with execution deviations exceeding the set threshold. Optimization schemes are also developed for generating structure optimization schemes for high-risk clauses, and priority update strategies are formulated based on the consistency of clause execution and the degree of innovation contribution. New versions are generated under the guidance of adaptive strategies, and the basis for updates and related monitoring data are recorded to achieve version traceability. Multiple versions are supported for parallel management and difference visualization. The convergence of the logical skeleton of the new version terms is evaluated to ensure that the update process is safe and controllable.

8. The intelligent compilation method for draft standards of scientific research results based on big data according to claim 3, characterized in that, The comprehensive fit index in the standardized fit index model is calculated by weighted coupling based on the maturity index, public diffusion capability index, parameter replicability index, boundary clarity coefficient, and risk intensity. The weight coefficients were obtained through training on historical annotated corpora and experimental data. When the overall fit index exceeds the set standardization threshold, the standard blank analysis phase begins.

9. The intelligent compilation method for draft standards of scientific research results based on big data according to claim 5, characterized in that, The generation of the standard clause logic skeleton includes: A clause logical dependency matrix is ​​constructed based on the technology path graph, the clause nodes are topologically sorted to determine the clause order, and the conflict index between clauses is calculated. The generated clauses are iteratively evaluated by executing a feasibility function and a clause stability index. The feasibility function is obtained by weighting the maturity of the results, the consistency of the clause detection method, and the stability of the clause expression structure. The clause stability index is obtained by coupling the innovation contribution index and the conflict index. When the feasibility or stability index falls below a set threshold, the process returns to adjusting the structured semantic template or recalculating the logical dependency matrix until the iterative convergence condition is met.

10. A smart drafting system for scientific research achievement standards based on big data, characterized in that, include: Memory; processor; A computer program stored in the memory and capable of running on the processor; When the processor executes the computer program, it implements the intelligent compilation method for draft standards of scientific research results based on big data as described in any one of claims 1 to 9.

Images