A Standard Content Semantic Association Method and System Based on Large Models
By processing the source relationships and content data of standard documents through a large model, a multi-level technical content chain is constructed to identify and analyze technical differences. This solves the problem of accuracy in identifying technical differences in multi-level standard relationships and achieves more refined and reliable standard content association.
Patent Information
- Application Number
- CN202611401029.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-10
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies are insufficient to accurately reflect the actual changes in technical differences in multi-level standard relationships, resulting in inaccurate standard retrieval, difference analysis, and compliance judgment.
By using a standard content semantic association method based on a large model, source relationship data and technical content data of standard documents are obtained, structured processing is performed, a multi-level technical content correspondence chain is constructed, technical objects, attributes, requirements and parameters are identified, difference analysis is performed, and derived technical correspondence data is generated.
It improves the granularity and accuracy of standard content association, enhances the interpretability and traceability of technical content correspondence, reduces the risk of propagating erroneous technical parameters, and improves the credibility of standard retrieval and compliance analysis.
Smart Images

Figure CN123116197A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of standard digital technology, specifically to a method and system for semantic association of standard content based on a large model. Background Technology
[0002] With the development of standards digitization, intelligent retrieval, and large language model technology, an increasing number of standards service platforms are beginning to structure national standards, industry standards, international standards, and other standard documents. They are utilizing technologies such as large models, vector retrieval, and knowledge graphs to establish relationships between standard content, supporting standard retrieval, standard comparison, standard conformity analysis, and content correspondence analysis between different standard systems. In addition to citation, substitution, and revision relationships, different standards may also have adoption relationships such as equivalent adoption and modified adoption. When the adopted standard itself originates from other standards, multi-level standard source relationships are further formed.
[0003] In multi-level standard adoption relationships, the adoption relationship between two adjacent standards usually only reflects the overall degree of correspondence between the two standards, and cannot directly reflect the final differences in specific technical content between standards at different levels. For example, if standard B modifies a certain technical parameter relative to standard C, standard A, when formed based on standard B, may modify that technical parameter again. The later modification may amplify the previous difference, reduce the previous difference, or restore the original technical content. Similar situations may also occur during continuous changes in technical requirements and constraints.
[0004] In existing standard association processing, relationships such as equivalent adoption, modified adoption, reference, and substitution are typically stored as relationship attributes between standards, and retrieval, association, or multi-hop expansion are performed based on these relationships. However, the same standard relationship may correspond to different actual technical changes. Relying solely on relationship tags or connections between standards is insufficient to accurately reflect the actual changes in technical differences across multi-level standard relationships. For example, if an intermediate standard modifies a pressure limit from 5MPa to 6MPa, and a subsequent standard reverts to 5MPa, if the system only determines the persistence of technical differences based on the two-level "modified adoption" relationship, it may incorrectly use the 6MPa requirement as the final association result and further apply it to standard retrieval, standard difference analysis, product standard matching, or compliance assessment.
[0005] Therefore, in a standard system with continuous standard source relationships, existing technologies struggle to accurately distinguish the relationship between adjacent standards and the actual technical differences between the first and last standards in the chain. This can easily lead to the transmission of technical differences that have been reduced, restored, or further changed as valid differences, thereby affecting the accuracy of results such as standard retrieval, standard difference analysis, product standard matching, and standard compliance judgment. Summary of the Invention
[0006] To address the shortcomings of existing technologies, this invention provides a standard content semantic association method based on a large model, which can effectively solve the problems mentioned in the background technology.
[0007] To achieve the above objectives, the present invention is implemented through the following technical solution: a standard content semantic association method based on a large model, including acquiring multiple standard files, collecting standard source relationship data and standard technical content data corresponding to each standard file, and performing structured processing on the standard technical content data according to the large model to obtain technical content unit data corresponding to each standard file;
[0008] Based on the standard source relationship data, the technical content unit data corresponding to adjacent standard documents are matched to obtain technical content correspondence data. Based on the technical content correspondence data, the technical content unit data corresponding to adjacent standard documents are analyzed for differences to obtain technical deviation data.
[0009] A multi-level technical content correspondence chain is constructed based on the technical content correspondence data, and the technical deviation data in the multi-level technical content correspondence chain is combined and processed to obtain the derived technical correspondence status.
[0010] Based on the data on the corresponding status and technical content of derived technologies, the data on the corresponding relationships between the standard documents at the head of the chain and the standard documents at the tail of the chain is generated. Based on the data on the corresponding relationships between derived technologies, standard content association and standard difference analysis are performed to obtain the standard content association results.
[0011] Furthermore, the method for collecting the standard source relationship data and standard technical content data corresponding to each standard document is as follows:
[0012] Multiple standard documents are obtained, and the standard number, standard name, adoption number, and adoption degree of each standard document are extracted. Based on the standard number, adoption number, and adoption degree, the source correspondence between each standard document is established to obtain standard source relationship data.
[0013] The main text of each standard document is parsed, and the main text is segmented line by line according to the chapter number and clause number to obtain the clause content data corresponding to each standard document. The clause content data is then used as the standard technical content data.
[0014] Furthermore, the method for obtaining the technical content unit data corresponding to each standard document is as follows:
[0015] The standard technical content data is input into the large model one by one, and the technical objects, technical attributes, technical requirements, constraints and technical parameters in each standard technical content data are identified to obtain structured technical feature data.
[0016] Structured technical feature data are merged according to technical objects and technical attributes. Structured technical feature data with the same technical objects and technical attributes are grouped into the same technical content unit. The technical objects, technical attributes, technical requirements, constraints, technical parameters, standard numbers, chapter numbers, and clause numbers corresponding to the technical content unit are recorded to obtain the technical content unit data corresponding to each standard document.
[0017] Furthermore, the method for matching the technical content unit data corresponding to adjacent standard documents is as follows:
[0018] Based on the standard source relationship data, adjacent standard documents with direct source relationships are identified, and the technical content unit data corresponding to the adjacent standard documents are extracted respectively;
[0019] The technical objects and technical attributes in the technical content unit data of adjacent standard documents are compared item by item. Technical content units with the same technical objects and technical attributes are paired, and the standard number, chapter number and clause number corresponding to the paired technical content units are recorded to obtain the technical content correspondence data.
[0020] Furthermore, the method for obtaining the technical deviation data is as follows:
[0021] Based on the technical content correspondence data, extract the paired technical content unit data, and compare the technical requirements, constraints and technical parameters in the paired technical content unit data item by item.
[0022] Extract inconsistencies from technical requirements, constraints, and technical parameters, and associate these inconsistencies with their corresponding standard numbers, chapter numbers, and clause numbers to obtain technical deviation data.
[0023] Furthermore, the method for constructing a multi-level technical content correspondence chain based on technical content correspondence data is as follows:
[0024] Based on the technical content correspondence data and in accordance with the order of standard documents determined by the standard source relationship data, the technical content units that have been paired in the previous adjacent standard document are sequentially connected with the corresponding technical content units in the next adjacent standard document to obtain a multi-level technical content correspondence chain.
[0025] Furthermore, the method for combining and processing the technical deviation data in the multi-level technical content correspondence chain is as follows:
[0026] Based on the multi-level technical content correspondence chain, the technical deviation data between each adjacent technical content unit is extracted sequentially, and the technical deviation data is arranged in the order of the standard documents.
[0027] The technical deviation data in the corresponding chains of multi-level technical content are read sequentially according to the standard documents. The technical requirements, constraints and technical parameters in the corresponding technical content units of the standard documents at the end of the chain are compared with the technical requirements, constraints and technical parameters in the corresponding technical content units of the standard documents at the beginning of the chain. When the corresponding content in the standard documents at the end of the chain is restored to the corresponding content in the standard documents at the beginning of the chain, the corresponding technical deviation is determined to be in an offset state. When the corresponding content in the standard documents at the end of the chain is still inconsistent with the corresponding content in the standard documents at the beginning of the chain, the corresponding technical deviation is determined to be in a retained state.
[0028] The offsetting or reserved states corresponding to the technical requirements, constraints, and technical parameters are determined separately. When the technical requirements, constraints, and technical parameters are all offsetting states, the derived technical corresponding state of the multi-level technical content correspondence chain is determined to be a consistent state. When at least one of the technical requirements, constraints, and technical parameters is a reserved state, the derived technical corresponding state of the multi-level technical content correspondence chain is determined to be a differing state.
[0029] Furthermore, the method for generating the derived technology correspondence data between the chain head standard file and the chain tail standard file is as follows:
[0030] Based on the technical content correspondence data, extract the technical content unit data corresponding to the standard document at the beginning of the chain and the technical content unit data corresponding to the standard document at the end of the chain in the multi-level technical content correspondence chain, and obtain the standard number, chapter number and clause number corresponding to the standard document at the beginning of the chain and the standard document at the end of the chain.
[0031] Based on the corresponding status of the derived technology, the technical content unit data corresponding to the standard document at the head of the chain is associated with the technical content unit data corresponding to the standard document at the tail of the chain, and the offsetting or retention status of the technical requirements, constraints and technical parameters is recorded.
[0032] By associating the standard number, chapter number, clause number, and derived technology correspondence status of the technical requirements, constraints, and technical parameters corresponding to the standard documents at the beginning and end of the chain, the derived technology correspondence data between the standard documents at the beginning and end of the chain is obtained.
[0033] Furthermore, the method for performing standard content association and standard difference analysis based on derived technology correspondence data is as follows:
[0034] Based on the derived technology correspondence data, extract the interrelated technical content unit data in the standard documents at the beginning and end of the chain, and establish the standard content association between the standard documents at the beginning and end of the chain according to the standard number, chapter number and clause number;
[0035] Read the technical requirements, constraints and technical parameters from the interrelated technical content unit data respectively, and compare them with the corresponding offset and retention states in the derived technology correspondence data; mark the technical requirements, constraints or technical parameters in the offset state as consistent content, and mark the technical requirements, constraints or technical parameters in the retention state as different content;
[0036] By summarizing the consistent and differing content between the standard documents at the beginning and end of the chain, as well as the corresponding standard numbers, chapter numbers, and clause numbers, the standard content association results are obtained.
[0037] A second aspect of the present invention provides a standard content semantic association system based on a large model, comprising:
[0038] The structured processing module is used to acquire multiple standard documents, collect standard source relationship data and standard technical content data corresponding to each standard document, and perform structured processing on the standard technical content data according to the large model to obtain the technical content unit data corresponding to each standard document.
[0039] The deviation analysis module, based on the standard source relationship data, matches the technical content unit data corresponding to adjacent standard documents to obtain technical content correspondence data. Based on the technical content correspondence data, it performs difference analysis on the technical content unit data corresponding to adjacent standard documents to obtain technical deviation data.
[0040] The deviation combination module is used to construct a multi-level technical content correspondence chain based on the technical content correspondence data, and to combine the technical deviation data in the multi-level technical content correspondence chain to obtain the derived technical correspondence status.
[0041] The correlation analysis module is used to generate the correlation data between the standard files at the head of the chain and the standard files at the tail of the chain based on the corresponding status of the derived technology and the correlation data of the technology content. Based on the correlation data of the derived technology, the module performs standard content correlation and standard difference analysis to obtain the standard content correlation results.
[0042] The present invention has the following beneficial effects:
[0043] This invention further extends standard relationship analysis to the level of specific technical content. By structurally identifying and hierarchically mapping the technical objects, attributes, requirements, constraints, and parameters in continuous standard source relationships, it can obtain the technical content that has actually changed between adjacent standards. This avoids directly equating standard-level relationship tags with specific technical content differences, improves the refinement and accuracy of semantic associations of standard content, and enhances the interpretability and traceability of the correspondence between technical content of standards at different levels.
[0044] This invention constructs a multi-level technical content correspondence chain to continuously combine and process technical deviations formed between adjacent standards. Based on changes in technical requirements, constraints, and technical parameters between the first and last standard documents in the chain, it determines whether the technical deviations are offset or retained. This effectively identifies situations where technical differences generated by intermediate standards are further amplified, reduced, or restored in subsequent standards. Compared to simply passing information based on multi-level standard relationships, this invention avoids intermediate technical deviations that have already been restored by subsequent standards from continuing to be output as valid differences, improving the accuracy of the final technical status judgment under multi-level standard source relationships and enhancing the reliability of technical difference analysis in complex standard relationships.
[0045] This invention further associates the corresponding status of derived technologies with the standard numbers, chapter numbers, clause numbers, and specific technical content of the head and tail standard documents in the chain. This forms a locationable and verifiable derived technology correspondence data and standard content association result. The final association result not only indicates whether there are technical differences between the head and tail standards, but also clarifies the specific location where the differences occur in technical requirements, constraints, or parameters. This improves the reliability of association results in standard retrieval, standard comparison, product standard matching, and standard compliance analysis, reduces the risk of erroneous technical parameters or requirements propagating along the standard relationship chain, and enhances the ability of the standard digitization system to perform accurate semantic association and difference analysis of multi-level standard source relationships. Attached Figure Description
[0046] Figure 1 This is a schematic diagram of the method flow of the present invention.
[0047] Figure 2 This is a schematic diagram of the system module connections of the present invention. Detailed Implementation
[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0049] Please see Figure 1 As shown, the first aspect of this invention provides a technical solution: a standard content semantic association method based on a large model, comprising:
[0050] Multiple standard documents are acquired, and the standard source relationship data and standard technical content data corresponding to each standard document are collected. The standard technical content data is then structured according to the large model to obtain the technical content unit data corresponding to each standard document.
[0051] Specifically, the method for obtaining multiple standard documents and collecting the standard source relationship data and standard technical content data corresponding to each standard document is as follows:
[0052] Multiple standard documents are obtained, and the standard number, standard name, adoption number, and adoption degree of each standard document are extracted. Based on the standard number, adoption number, and adoption degree, the source correspondence between each standard document is established to obtain standard source relationship data.
[0053] It should be added that the standard number and standard name can be extracted from the cover or the basic information area of the standard document, and the adoption number and adoption degree can be extracted from the field recording the adoption relationship, the preface, or the corresponding basic information area of the standard document; for standard documents that have already saved the basic information of the standard in a structured form, the corresponding fields can be read directly.
[0054] The specific process for obtaining standard source relationship data is as follows: compare the standard number and adoption number of each standard document one by one. When the adoption number of one standard document is consistent with the standard number of another standard document, the two standard documents are determined to have a direct source relationship. Connect the standard documents in sequence according to the direct source relationship between each standard document, and record the source correspondence between each standard document in combination with the degree of adoption, so as to obtain the standard source relationship data.
[0055] The source correspondence records at least the standard numbers of two standard documents with a direct source relationship and their corresponding adoption levels, and determines the source direction according to the standard number pointed to by the adoption number; when the adoption number of the first standard document corresponds to the standard number of the second standard document, and the adoption number of the second standard document corresponds to the standard number of the third standard document, a continuous standard source relationship is formed between the first standard document, the second standard document and the third standard document in sequence.
[0056] In this embodiment, multiple standard documents with pairwise adoption relationships are organized into a multi-level standard source relationship with a clear sequence, so that subsequent technical deviations can be processed continuously according to the actual source order, rather than directly comparing any two standard documents in an unordered manner.
[0057] The main text of each standard document is parsed, and the main text is segmented line by line according to the chapter number and clause number to obtain the clause content data corresponding to each standard document. The clause content data is then used as the standard technical content data.
[0058] During the specific analysis, the chapter number, clause number, and corresponding text text in the standard document are read. The adjacent chapter number or clause number is used as the dividing boundary of the current text text, and the standard number, chapter number, and clause number corresponding to the clause content data after the division are retained. When there are multiple consecutive text segments for the same clause, the consecutive text segments belonging to the same clause number are merged into the clause content data corresponding to that clause.
[0059] By retaining the correspondence between clause content and standard number, chapter number, and clause number, the specific location of technical content in the original standard document can be continuously traced during subsequent large model identification, technical content matching, and difference output processes.
[0060] Specifically, the method for structuring the standard technical content data according to the large model to obtain the technical content unit data corresponding to each standard document is as follows:
[0061] The standard technical content data is input into the large model one by one, and the technical objects, technical attributes, technical requirements, constraints, and technical parameters in each piece of standard technical content data are identified to obtain structured technical feature data. Among them, the technical requirements include the performance requirements, operational requirements, and test method requirements specified in the standard technical content for the technical object.
[0062] In this embodiment, the large-scale model used is an existing large-scale language model. This model is capable of semantic understanding, information recognition, and structured information output based on the input text content. Its specific model structure, training method, and model parameters can all be implemented using existing technologies, and this invention does not impose specific limitations on them. This embodiment mainly utilizes the large-scale model to identify the technical objects, technical attributes, technical requirements, constraints, and technical parameters in standard technical content data.
[0063] The specific process for obtaining structured technical feature data is as follows: each piece of standard technical content data is input into the large model as an independent input, and the large model is required to output the recognition results in the fixed field order of technical object, technical attribute, technical requirement, constraint condition and technical parameter; when the corresponding field does not appear in the standard technical content data, the corresponding field is set to empty, and the recognition results of each field are saved according to the clause number of the standard technical content data to obtain the structured technical feature data.
[0064] In one specific embodiment, a standard technical content data point is "When the equipment is operating normally, the working pressure should not exceed 5MPa". After inputting this standard technical content data into the large model, it is output according to fixed fields: the technical object is "equipment", the technical attribute is "working pressure", the technical requirement is "not greater than", the constraint condition is "when operating normally", and the technical parameter is "5MPa". The above identification results are associated with the corresponding standard number, chapter number, and clause number to form the structured technical feature data corresponding to that clause.
[0065] Structured technical feature data are merged according to technical objects and technical attributes. Structured technical feature data with consistent technical objects and attributes are grouped into the same technical content unit. The technical object, technical attribute, technical requirement, constraint, technical parameter, standard number, chapter number, and clause number corresponding to each technical content unit are recorded, resulting in the technical content unit data for each standard document. A technical object refers to the object addressed by the technical requirements in the standard's technical content; a technical attribute refers to the attribute used to characterize the specific technical features of the technical object; a technical requirement refers to the performance requirements, operational requirements, or test method requirements specified for the technical object and its technical attributes; a constraint is used to characterize the operating state, scope of application, or limiting conditions corresponding to the fulfillment of the technical requirements or technical parameters; and a technical parameter refers to the parameter content used to quantify or limit the technical attributes.
[0066] Structured technical feature data are consolidated within the same standard document. When multiple structured technical feature data in the same standard document have the same technical object and technical attribute, the corresponding technical requirements, constraints, technical parameters, and their respective chapter and clause numbers are recorded together in the same technical content unit to preserve the specific content of the same technical object and technical attribute appearing in different clause positions.
[0067] This process uses technical objects and technical attributes as stable merging criteria across clauses, ensuring that subsequent matching of adjacent standards targets the same technical matter. At the same time, it retains technical requirements, constraints, and technical parameters as content for difference analysis, thereby avoiding the omission of technical changes that need to be compared by using the changed technical requirements themselves as merging criteria.
[0068] Based on the standard source relationship data, the technical content unit data corresponding to adjacent standard documents are matched to obtain technical content correspondence data. Based on the technical content correspondence data, the difference analysis of the technical content unit data corresponding to adjacent standard documents is performed to obtain technical deviation data.
[0069] By first identifying the technical content units belonging to the same technical matter in adjacent standards, and then conducting a difference analysis only on the paired technical requirements, constraints, and technical parameters, the issues of "whether the technical content corresponds" and "whether the corresponding content has changed" can be processed separately, reducing the possibility of mistaking textual differences between different technical matters as differences in standard adoption.
[0070] Specifically, based on standard source relationship data, the method for matching the technical content unit data corresponding to adjacent standard documents to obtain technical content correspondence data is as follows:
[0071] Based on the standard source relationship data, adjacent standard documents with direct source relationships are identified, and the technical content unit data corresponding to the adjacent standard documents are extracted respectively.
[0072] Specifically, adjacent matching is performed only once for two standard files with a direct source relationship in the standard source relationship data. The preceding and following standard files are saved according to the source direction determined by the standard source relationship data, ensuring that the corresponding relationship data of each group of technical content maintains the sequential order consistent with the multi-level standard source relationship. This process limits the scope of technical content matching and avoids losing technical changes that occur in intermediate standards due to direct matching across intermediate standards.
[0073] The technical objects and technical attributes in the technical content unit data of adjacent standard documents are compared item by item. Technical content units with the same technical objects and technical attributes are paired, and the standard number, chapter number and clause number corresponding to the paired technical content units are recorded to obtain the technical content correspondence data.
[0074] The specific process for obtaining the technical content correspondence data is as follows: When there are multiple technical content units with the same technical objects and the same technical attributes in adjacent standard documents, the technical requirements and constraints in each technical content unit are compared, and the paired technical content units are determined according to the correspondence of the technical requirements and constraints; after the pairing is completed, the standard number, chapter number and clause number corresponding to the paired technical content units are recorded to obtain the technical content correspondence data.
[0075] In a specific embodiment, a technical content unit in standard document A corresponds to the technical object "equipment," the technical attribute "working pressure," the technical requirement "not greater than," and the technical parameter "5MPa." Similarly, a technical content unit in standard document B corresponds to the technical object "equipment," the technical attribute "working pressure," the technical requirement "not greater than," and the technical parameter "6MPa." Since both technical content units have the technical object "equipment" and the technical attribute "working pressure," the two technical content units are paired, and the standard number, chapter number, and clause number corresponding to each technical content unit are recorded, forming technical content correspondence data.
[0076] Specifically, based on the technical content correspondence data, the method for obtaining technical deviation data by performing a difference analysis on the technical content unit data corresponding to adjacent standard documents is as follows:
[0077] Based on the technical content correspondence data, extract the paired technical content unit data, and compare the technical requirements, constraints and technical parameters in the paired technical content unit data item by item.
[0078] For numerical content in technical parameters, the technical parameters in the corresponding technical content unit of the previous standard document are denoted as follows: The technical parameters in the corresponding technical content unit of the later standard document are denoted as Calculate the technical parameter deviations between adjacent technical content units:
[0079] ;
[0080] in, Indicates the first The adjacent standard documents and the first The deviation of technical parameters of corresponding technical content units between adjacent standard documents; Indicates the first The technical parameters in the corresponding technical content unit of each standard document; Indicates the first The technical parameters in the technical content unit corresponding to each standard document.
[0081] when When, it is determined that there is no deviation in the technical parameters of adjacent technical content units; when At that time, The corresponding changes in technical parameters are recorded as technical deviation data.
[0082] For technical requirements and constraints, consistency markers are used to represent technical deviations:
[0083] ;
[0084] in, Indicates the first The first standard document and the first Between the first standard documents Technical deviation status of the content and These respectively represent the technical requirements or constraints in adjacent technical content units; This indicates that there is no technical deviation. This indicates a technical discrepancy.
[0085] When technical parameters include numerical values and corresponding constraints, the corresponding constraints are compared if the technical parameter values are the same; if either the technical parameter values or the corresponding constraints are inconsistent, the corresponding content is treated as technical deviation data.
[0086] Extract inconsistencies from technical requirements, constraints, and technical parameters, and associate these inconsistencies with their corresponding standard numbers, chapter numbers, and clause numbers to obtain technical deviation data.
[0087] The specific process for obtaining technical deviation data is as follows: For inconsistent technical requirements, constraints, or technical parameters, record the content of the corresponding technical content unit in the previous standard document and the corresponding technical content unit in the subsequent standard document, and associate them with the corresponding standard number, chapter number, and clause number to obtain technical deviation data.
[0088] In a specific embodiment, the technical object in the first standard document's corresponding technical content unit is "equipment," the technical attribute is "working pressure," the technical requirement is "not greater than," the constraint is "under normal operation," and the technical parameter is "5MPa." The technical object in the second standard document's corresponding technical content unit is also "equipment," the technical attribute is "working pressure," the technical requirement is "not greater than," the constraint is "under normal operation," and the technical parameter is "6MPa." After comparing the two technical content units item by item, the technical requirements and constraints are consistent, but the technical parameters are inconsistent. According to the technical parameter deviation calculation formula, the technical parameter deviation is 1MPa. Therefore, the change from "5MPa to 6MPa," along with the corresponding standard number, chapter number, and clause number, is recorded as technical deviation data.
[0089] In this embodiment, by matching technical content units between adjacent standard documents with direct source relationships and performing item-by-item difference analysis on the corresponding technical requirements, constraints, and technical parameters, the overall adoption relationship between standard documents can be further refined into the change relationship between specific technical contents. This accurately determines the technical contents that have actually changed between adjacent standards, avoiding the direct judgment of technical differences based solely on standard-level relationship labels such as modified adoption or equivalent adoption, thus improving the precision and accuracy of identifying technical differences between adjacent standards.
[0090] A multi-level technical content correspondence chain is constructed based on the technical content correspondence data, and the technical deviation data in the multi-level technical content correspondence chain is combined and processed to obtain the derived technical correspondence status.
[0091] Specifically, the method for constructing a multi-level technical content correspondence chain based on technical content correspondence data is as follows:
[0092] Based on the technical content correspondence data and in accordance with the order of standard documents determined by the standard source relationship data, the technical content units that have been paired in the previous adjacent standard document are sequentially connected with the corresponding technical content units in the next adjacent standard document to obtain a multi-level technical content correspondence chain.
[0093] The specific process of obtaining the multi-level technical content correspondence chain is as follows: according to the order of standard documents determined by the standard source relationship data, find the next technical content unit in the previous set of technical content correspondence relationship data, and find the previous technical content unit that is the same as the technical content unit in the next set of technical content correspondence relationship data; when the standard number, chapter number and clause number corresponding to the two are consistent, connect the corresponding technical content units in sequence to obtain the multi-level technical content correspondence chain.
[0094] In a specific embodiment, the technical content unit corresponding to clause C-5.1 in standard document C is paired with the technical content unit corresponding to clause B-5.2 in standard document B. The technical content unit corresponding to clause B-5.2 in standard document B is also paired with the technical content unit corresponding to clause A-6.1 in standard document A. Since the next technical content unit in the previous set of technical content correspondence data and the previous technical content unit in the next set of technical content correspondence data are both technical content units corresponding to clause B-5.2 in standard document B, the technical content units corresponding to clauses C-5.1, B-5.2, and A-6.1 are connected sequentially to form a multi-level technical content correspondence chain.
[0095] Specifically, the method for combining and processing the technical deviation data in the multi-level technical content correspondence chain to obtain the derived technical correspondence state is as follows:
[0096] Based on the multi-level technical content correspondence chain, the technical deviation data between each adjacent technical content unit is extracted sequentially, and the technical deviation data is arranged in the order of the standard documents.
[0097] The technical deviation data in the corresponding chains of multi-level technical content are read sequentially according to the standard documents. The technical requirements, constraints and technical parameters in the corresponding technical content units of the standard documents at the end of the chain are compared with the technical requirements, constraints and technical parameters in the corresponding technical content units of the standard documents at the beginning of the chain. When the corresponding content in the standard documents at the end of the chain is restored to the corresponding content in the standard documents at the beginning of the chain, the corresponding technical deviation is determined to be in an offset state. When the corresponding content in the standard documents at the end of the chain is still inconsistent with the corresponding content in the standard documents at the beginning of the chain, the corresponding technical deviation is determined to be in a retained state.
[0098] The offset state refers to the state in which the technical requirements, constraints, or technical parameters corresponding to the tail standard document are restored to the corresponding content of the head standard document after the technical content has changed in the multi-level standard source relationship. The retention state refers to the state in which the technical requirements, constraints, or technical parameters corresponding to the tail standard document are still inconsistent with the corresponding content of the head standard document after multiple levels of change.
[0099] For numerical technical parameters, the deviations of technical parameters in the corresponding chains of multi-level technical content are combined according to the order of the standard documents to obtain:
[0100] ;
[0101] in, This indicates the total deviation of technical parameters between the standard documents at the beginning and end of the chain corresponding to multi-level technical content. This indicates the number of standard files in the chain corresponding to multi-level technical content. This indicates the technical parameters in the corresponding technical content unit of the standard document at the head of the chain; This indicates the technical parameters in the corresponding technical content unit of the standard document at the end of the chain.
[0102] when When =0, the technical deviation corresponding to the technical parameter is determined to be in a canceled state; when At that time, the technical deviations corresponding to the technical parameters are determined to be retained, and the changes in technical parameters corresponding to the total deviation of the technical parameters are taken as the final changes in technical parameters between the standard documents at the head of the chain and the standard documents at the tail of the chain. The total deviation of the technical parameters is used to characterize the final degree of change after the combination of the technical parameter deviations between adjacent technical content units in the chain corresponding to the multi-level technical content.
[0103] In one specific embodiment, the multi-level technical content correspondence chain sequentially includes standard file C, standard file B, and standard file A. The technical parameter in the technical content unit corresponding to standard file C is 5 MPa, the technical parameter in the technical content unit corresponding to standard file B is 6 MPa, and the technical parameter in the technical content unit corresponding to standard file A is 5 MPa. The technical parameter deviation from standard file C to standard file B is 1 MPa, the technical parameter deviation from standard file B to standard file A is -1 MPa, and the total technical parameter deviation is 0. Therefore, the technical deviation corresponding to this technical parameter is determined to be in a canceled state.
[0104] For technical requirements and constraints, the content corresponding to the standard document at the head of the chain will be denoted as follows: Record the content corresponding to the standard document at the end of the chain as The derived technology corresponds to the following state representation:
[0105] ;
[0106] in, This indicates the corresponding state of the derived technology that corresponds to the technical requirements or constraints. This indicates that the technical discrepancies in the corresponding chains of multi-level technical content are ultimately offset; This indicates that technical deviations in the chain corresponding to multi-level technical content are ultimately retained.
[0107] In a specific embodiment, if the technical requirement corresponding to the standard document at the beginning of the chain is "not greater than", the technical requirement corresponding to the standard document in the middle is "less than", and the technical requirement corresponding to the standard document at the end of the chain is again "not greater than", then the technical requirements of the standard document at the beginning of the chain and the standard document at the end of the chain are consistent, and the technical deviation corresponding to the technical requirement is determined to be in an offset state; when the technical requirement corresponding to the standard document at the end of the chain is still "less than", the technical deviation corresponding to the technical requirement is determined to be in a reserved state.
[0108] The offsetting or reserved states corresponding to the technical requirements, constraints, and technical parameters are determined separately. When the technical requirements, constraints, and technical parameters are all offsetting states, the derived technical corresponding state of the multi-level technical content correspondence chain is determined to be a consistent state. When at least one of the technical requirements, constraints, and technical parameters is a reserved state, the derived technical corresponding state of the multi-level technical content correspondence chain is determined to be a differing state.
[0109] In this embodiment, by sequentially connecting the corresponding technical content units in adjacent standards according to the standard source relationship to form a multi-level technical content correspondence chain, and continuously combining and processing the technical deviation data at each level, it can reflect the continuous change process of the same technical content in multiple standard source levels, and determine whether the technical deviation generated by the intermediate standard is ultimately offset or continues to be retained after subsequent standard modification, thereby avoiding the continued transmission of intermediate technical deviations that have been recovered or reduced to the end of the chain, and improving the accuracy and reliability of the final technical status judgment under the multi-level standard source relationship.
[0110] Based on the data on the corresponding status and technical content of derived technologies, the data on the corresponding relationships between the standard documents at the head of the chain and the standard documents at the tail of the chain is generated. Based on the data on the corresponding relationships between derived technologies, standard content association and standard difference analysis are performed to obtain the standard content association results.
[0111] Specifically, the method for generating the derived technology correspondence data between the chain head standard file and the chain tail standard file based on the derived technology correspondence status and technology content correspondence data is as follows:
[0112] Based on the technical content correspondence data, extract the technical content unit data corresponding to the standard document at the head of the chain and the technical content unit data corresponding to the standard document at the tail of the chain in the multi-level technical content correspondence chain, and obtain the standard number, chapter number and clause number corresponding to the standard document at the head of the chain and the standard document at the tail of the chain.
[0113] Based on the corresponding status of the derived technology, the technical content unit data corresponding to the standard document at the head of the chain is associated with the technical content unit data corresponding to the standard document at the tail of the chain, and the offsetting or retention status of the technical requirements, constraints and technical parameters is recorded.
[0114] By associating the standard number, chapter number, clause number, and derived technology correspondence status of the technical requirements, constraints, and technical parameters corresponding to the standard documents at the beginning and end of the chain, the derived technology correspondence data between the standard documents at the beginning and end of the chain is obtained.
[0115] In a specific embodiment, the head standard document of the multi-level technical content correspondence chain is standard document C, and the tail standard document is standard document A. The technical content unit corresponding to clause C-5.1 in the head standard document and the technical content unit corresponding to clause A-6.1 in the tail standard document have a corresponding relationship, and the corresponding technical requirements, constraints and technical parameters are all in a state of offsetting. Then, the standard number, chapter number and clause number of standard document C, the standard number, chapter number and clause number of standard document A, and the above offsetting states are associated to form the derived technical correspondence data between standard document C and standard document A.
[0116] Specifically, the method for obtaining standard content association results by performing standard content association and standard difference analysis based on the derived technology correspondence data is as follows:
[0117] Based on the derived technology correspondence data, extract the interrelated technical content unit data in the standard documents at the beginning and end of the chain, and establish the standard content association between the standard documents at the beginning and end of the chain according to the standard number, chapter number, and clause number.
[0118] Read the technical requirements, constraints and technical parameters from the interrelated technical content unit data respectively, and compare them with the corresponding offset and retention states in the derived technology correspondence data; mark the technical requirements, constraints or technical parameters in the offset state as consistent content, and mark the technical requirements, constraints or technical parameters in the retention state as different content.
[0119] By summarizing the consistent and differing content between the standard documents at the beginning and end of the chain, as well as the corresponding standard numbers, chapter numbers, and clause numbers, the standard content association results are obtained.
[0120] In a specific embodiment, the technical object corresponding to clause C-5.1 in the standard document at the beginning of the chain and clause A-6.1 in the standard document at the end of the chain is "equipment", and the technical attribute is "working pressure". The derived technology correspondence data shows that the technical requirements, constraints and technical parameters are all in a state of offsetting. Therefore, the technical requirements, constraints and technical parameters are all marked as consistent content. The consistent content, as well as the standard number, chapter number and clause number corresponding to clause C-5.1 and clause A-6.1, are summarized to form the standard content association result.
[0121] In this embodiment, by associating the derived technology's corresponding status with the technical content units, standard numbers, chapter numbers, and clause numbers corresponding to the standard documents at the beginning and end of the chain, a head-to-tail standard correspondence relationship with specific technical content and positional basis can be formed. This accurately distinguishes consistent content from the ultimately retained differences, enabling the standard content association results to be traced back to specific standard clauses and intermediate technical changes. This improves the verifiability and traceability of the standard content association results and enhances the credibility of standard retrieval, standard comparison, and standard compliance analysis results.
[0122] A second aspect of the present invention provides a standard content semantic association system based on a large model, comprising:
[0123] The structured processing module is used to acquire multiple standard documents, collect standard source relationship data and standard technical content data corresponding to each standard document, and perform structured processing on the standard technical content data according to the large model to obtain the technical content unit data corresponding to each standard document.
[0124] The deviation analysis module, based on the standard source relationship data, matches the technical content unit data corresponding to adjacent standard documents to obtain technical content correspondence data. Based on the technical content correspondence data, it performs difference analysis on the technical content unit data corresponding to adjacent standard documents to obtain technical deviation data.
[0125] The deviation combination module is used to construct a multi-level technical content correspondence chain based on the technical content correspondence data, and to combine the technical deviation data in the multi-level technical content correspondence chain to obtain the derived technical correspondence status.
[0126] The correlation analysis module is used to generate the correlation data between the standard files at the head of the chain and the standard files at the tail of the chain based on the corresponding status of the derived technology and the correlation data of the technology content. Based on the correlation data of the derived technology, the module performs standard content correlation and standard difference analysis to obtain the standard content correlation results.
[0127] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0128] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.
Claims
1. A standard content semantic association method based on a large model, characterized in that, include: Multiple standard documents are acquired, and standard source relationship data and standard technical content data corresponding to each standard document are collected. The standard technical content data is then structured according to the large model to obtain the technical content unit data corresponding to each standard document. Based on the standard source relationship data, the technical content unit data corresponding to adjacent standard documents are matched to obtain technical content correspondence data. Based on the technical content correspondence data, the technical content unit data corresponding to adjacent standard documents are analyzed for differences to obtain technical deviation data. A multi-level technical content correspondence chain is constructed based on the technical content correspondence data, and the technical deviation data in the multi-level technical content correspondence chain is combined and processed to obtain the derived technical correspondence status. Based on the data on the corresponding status and technical content of derived technologies, the data on the corresponding relationships between the standard documents at the head of the chain and the standard documents at the tail of the chain is generated. Based on the data on the corresponding relationships between derived technologies, standard content association and standard difference analysis are performed to obtain the standard content association results.
2. The standard content semantic association method based on a large model according to claim 1, characterized in that, The method for collecting standard source relationship data and standard technical content data corresponding to each standard document is as follows: Multiple standard documents are obtained, and the standard number, standard name, adoption number, and adoption degree of each standard document are extracted. Based on the standard number, adoption number, and adoption degree, the source correspondence between each standard document is established to obtain standard source relationship data. The main text of each standard document is parsed, and the main text is segmented line by line according to the chapter number and clause number to obtain the clause content data corresponding to each standard document. The clause content data is then used as the standard technical content data.
3. The standard content semantic association method based on a large model according to claim 2, characterized in that, The method for obtaining the technical content unit data corresponding to each standard document is as follows: The standard technical content data is input into the large model one by one, and the technical objects, technical attributes, technical requirements, constraints and technical parameters in each standard technical content data are identified to obtain structured technical feature data. Structured technical feature data are merged according to technical objects and technical attributes. Structured technical feature data with the same technical objects and technical attributes are grouped into the same technical content unit. The technical objects, technical attributes, technical requirements, constraints, technical parameters, standard numbers, chapter numbers, and clause numbers corresponding to the technical content unit are recorded to obtain the technical content unit data corresponding to each standard document.
4. The standard content semantic association method based on a large model according to claim 2, characterized in that, The method for matching the technical content unit data corresponding to adjacent standard documents is as follows: Based on the standard source relationship data, adjacent standard documents with direct source relationships are identified, and the technical content unit data corresponding to the adjacent standard documents are extracted respectively; The technical objects and technical attributes in the technical content unit data of adjacent standard documents are compared item by item. Technical content units with the same technical objects and technical attributes are paired, and the standard number, chapter number and clause number corresponding to the paired technical content units are recorded to obtain the technical content correspondence data.
5. The standard content semantic association method based on a large model according to claim 4, characterized in that, The method for obtaining technical deviation data is as follows: Based on the technical content correspondence data, extract the paired technical content unit data, and compare the technical requirements, constraints and technical parameters in the paired technical content unit data item by item. Extract inconsistencies from technical requirements, constraints, and technical parameters, and associate these inconsistencies with their corresponding standard numbers, chapter numbers, and clause numbers to obtain technical deviation data.
6. The standard content semantic association method based on a large model according to claim 4, characterized in that, The method for constructing a multi-level technical content correspondence chain based on technical content correspondence data is as follows: Based on the technical content correspondence data and in accordance with the order of standard documents determined by the standard source relationship data, the technical content units that have been paired in the previous adjacent standard document are sequentially connected with the corresponding technical content units in the next adjacent standard document to obtain a multi-level technical content correspondence chain.
7. The standard content semantic association method based on a large model according to claim 6, characterized in that, The method for combining and processing technical deviation data in the multi-level technical content correspondence chain is as follows: Based on the multi-level technical content correspondence chain, the technical deviation data between each adjacent technical content unit is extracted sequentially, and the technical deviation data is arranged in the order of the standard documents. The technical deviation data in the corresponding chains of multi-level technical content are read sequentially according to the standard documents. The technical requirements, constraints and technical parameters in the corresponding technical content units of the standard documents at the end of the chain are compared with the technical requirements, constraints and technical parameters in the corresponding technical content units of the standard documents at the beginning of the chain. When the corresponding content in the standard documents at the end of the chain is restored to the corresponding content in the standard documents at the beginning of the chain, the corresponding technical deviation is determined to be in an offset state. When the corresponding content in the standard documents at the end of the chain is still inconsistent with the corresponding content in the standard documents at the beginning of the chain, the corresponding technical deviation is determined to be in a retained state. The offsetting or reserved states corresponding to the technical requirements, constraints, and technical parameters are determined separately. When the technical requirements, constraints, and technical parameters are all offsetting states, the derived technical corresponding state of the multi-level technical content correspondence chain is determined to be a consistent state. When at least one of the technical requirements, constraints, and technical parameters is a reserved state, the derived technical corresponding state of the multi-level technical content correspondence chain is determined to be a differing state.
8. The standard content semantic association method based on a large model according to claim 7, characterized in that, The method for generating the derived technology correspondence data between the chain head standard file and the chain tail standard file is as follows: Based on the technical content correspondence data, extract the technical content unit data corresponding to the standard document at the beginning of the chain and the technical content unit data corresponding to the standard document at the end of the chain in the multi-level technical content correspondence chain, and obtain the standard number, chapter number and clause number corresponding to the standard document at the beginning of the chain and the standard document at the end of the chain. Based on the corresponding status of the derived technology, the technical content unit data corresponding to the standard document at the head of the chain is associated with the technical content unit data corresponding to the standard document at the tail of the chain, and the offsetting or retention status of the technical requirements, constraints and technical parameters is recorded. By associating the standard number, chapter number, clause number, and derived technology correspondence status of the technical requirements, constraints, and technical parameters corresponding to the standard documents at the beginning and end of the chain, the derived technology correspondence data between the standard documents at the beginning and end of the chain is obtained.
9. The standard content semantic association method based on a large model according to claim 8, characterized in that, The method for performing standard content association and standard difference analysis based on the derived technology correspondence data is as follows: Based on the derived technology correspondence data, extract the interrelated technical content unit data in the standard documents at the beginning and end of the chain, and establish the standard content association between the standard documents at the beginning and end of the chain according to the standard number, chapter number and clause number; Read the technical requirements, constraints and technical parameters from the interrelated technical content unit data respectively, and compare them with the corresponding offset and retention states in the derived technology correspondence data; mark the technical requirements, constraints or technical parameters in the offset state as consistent content, and mark the technical requirements, constraints or technical parameters in the retention state as different content; By summarizing the consistent and differing content between the standard documents at the beginning and end of the chain, as well as the corresponding standard numbers, chapter numbers, and clause numbers, the standard content association results are obtained.
10. A standard content semantic association system based on a large model, characterized in that, include: The structured processing module is used to acquire multiple standard documents, collect standard source relationship data and standard technical content data corresponding to each standard document, and perform structured processing on the standard technical content data according to the large model to obtain the technical content unit data corresponding to each standard document. The deviation analysis module, based on the standard source relationship data, matches the technical content unit data corresponding to adjacent standard documents to obtain technical content correspondence data. Based on the technical content correspondence data, it performs difference analysis on the technical content unit data corresponding to adjacent standard documents to obtain technical deviation data. The deviation combination module is used to construct a multi-level technical content correspondence chain based on the technical content correspondence data, and to combine the technical deviation data in the multi-level technical content correspondence chain to obtain the derived technical correspondence status. The correlation analysis module is used to generate the correlation data between the standard files at the head of the chain and the standard files at the tail of the chain based on the corresponding status of the derived technology and the correlation data of the technology content. Based on the correlation data of the derived technology, the module performs standard content correlation and standard difference analysis to obtain the standard content correlation results.
