Standard information comparison method and system based on semantic recognition

By constructing standard information knowledge graphs and semantic recognition technology, the problem of low data accuracy in standard information comparison is solved, accurate information segmentation and semantic matching are achieved, and the accuracy and efficiency of standard information comparison are improved.

CN119026604AInactive Publication Date: 2024-11-26CHINA NAT INST OF STANDARDIZATION
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411087791.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-09
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, semantic recognition systems have problems with low data accuracy in the standard information comparison process, mainly due to the diversity and ambiguity of the text expression of standard information, as well as the complexity and noise of natural language.

Method used

By constructing a standard information knowledge graph, performing information annotation, obtaining the standard information to be compared in real time, and performing information segmentation and feature extraction, combining semantic recognition technology, obtaining semantic recognition indexes to ensure the degree of matching the standard information to be compared and the standard information annotation labels. , improve the accuracy of the comparison.

Benefits of technology

The accuracy of standard information comparison is improved, and the degree of information segmentation and semantic matching is numerical evaluation, ensuring the accuracy and efficiency of the comparison process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026604A_ABST
    Figure CN119026604A_ABST
Patent Text Reader

Abstract

The invention discloses a standard information comparison method and system based on semantic recognition, and relates to the technical field of electric digital data processing. The standard information comparison method based on semantic recognition comprises the following steps: constructing a knowledge graph; extracting features; and performing semantic recognition and comparison. The method comprises the following steps: collecting historical standard information, constructing a standard information knowledge graph, performing information labeling to obtain a standard information labeling label, performing information segmentation on to-be-compared standard information obtained in real time to obtain to-be-compared standard information blocks, and performing feature extraction to obtain to-be-compared standard information features. And finally, carrying out information semantic recognition on the to-be-compared standard information features until an obtained semantic recognition index conforms to a semantic comparison index, and dividing the semantic recognition index into a standard information labeling label conforming to the semantic comparison index, thereby improving the comparison accuracy of the standard information in the technical field of engineering. The problem of low accuracy of standard information data in a standard information comparison process in the prior art is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of electronic digital data processing, and in particular to a standard information comparison method and system based on semantic recognition. Background Art

[0002] With the rapid advancement of artificial intelligence and natural language processing technologies, standard information comparison methods based on semantic recognition are gaining significant improvements. The development of these technologies enables us to more accurately understand and analyze the deep semantics behind the text. The rise of deep learning, especially the introduction of the Transformer model, has greatly improved the system's ability to process language information. Through large-scale data training, the model can automatically learn and infer the complex semantic relationships between words, phrases, and sentences. At the same time, the construction and application of knowledge graphs have also added a new dimension to standard information comparison. By connecting entities and concepts in a structured form, the system can more accurately understand professional terms and their relevance in specific fields. The application of adaptive technology enables the comparison system to be adjusted according to the specific needs of different fields, further improving the accuracy and applicability of matching. With the advancement of automation technology, the comparison process has become more efficient and real-time. The system can process large-scale text data and quickly update the comparison results, providing effective support for standard information management.

[0003] Existing semantic recognition systems achieve efficient text understanding and processing through a series of advanced technologies and algorithms. These systems usually include key steps such as word segmentation, part-of-speech tagging, syntactic analysis, and semantic analysis to deeply understand the meaning and context of the text. Specifically, they use classification algorithms such as support vector machines, naive Bayes, and decision trees to classify texts; and use deep learning models such as recurrent neural networks, long short-term memory networks, and Transformers to capture long-distance dependencies and semantic features in texts. In addition, word embedding technology is also widely used to convert vocabulary into representations in high-dimensional vector space to capture the semantic relationship between words. The combination of these algorithms and technologies enables semantic recognition systems to accurately understand users' intentions and needs, providing strong support for applications in the field of natural language processing.

[0004] For example, a method for searching files based on semantic recognition is disclosed in the invention patent with the publication number CN116662366B, comprising: S1, obtaining user input information, calling a semantic recognition interface to process the input information to obtain processing information; S2, performing a progressive pre-search on the processing information, stopping the progressive pre-search when the number of files searched reaches a threshold, and jumping to S3; if the number of files searched still does not reach the threshold until the progressive pre-search is completed, jumping to S5; S3, extracting guidance information from the searched files, combining the processing information with the guidance information and performing a progressive secondary search, stopping the progressive secondary search when the number of files searched reaches a threshold, and jumping to S4; if the number of files searched still does not reach the threshold until the progressive secondary search is completed, jumping to S5; S4, displaying the processing information, guidance information and search results, determining whether to continue, if not, jumping to S5, if continuing, obtaining the user's secondary input information, calling the semantic recognition interface to process the secondary input information, obtaining secondary processing information and re-executing S2; S5, outputting the search results.

[0005] For example, the semantic recognition method and semantic recognition system disclosed in the invention patent announcement with announcement number: CN107291685B include: determining multiple reference words related to a keyword, and a local text adjacent to the keyword; calculating the probability that the reference word belongs to the local text, and recording it as a first probability; when it is detected that the first probability of the first reference word is greater than or equal to a preset probability, determining the semantics related to the first reference word as the semantics of the keyword; when it is detected that the first probability of all reference words is less than the preset probability, determining the global text to which the keyword belongs; calculating the probability that the reference word belongs to the global text, and recording it as a second probability; determining the reference word with the largest second probability as the second reference word, and determining the semantics related to the second reference word as the semantics of the keyword.

[0006] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:

[0007] In the prior art, the standard information text expressions collected by semantic recognition are diverse and ambiguous, and the natural language itself is complex. The presence of noise in the standard information text data also increases the inconsistency of the standard information text data, resulting in the problem of low accuracy of the standard information data during the standard information comparison process. Summary of the invention

[0008] The embodiments of the present application solve the problem of low accuracy of standard information data in the standard information comparison process in the prior art by providing a standard information comparison method and system based on semantic recognition, and achieve an improvement in the comparison accuracy of standard information in the field of engineering technology.

[0009] An embodiment of the present application provides a standard information comparison method based on semantic recognition, comprising the following steps: S1, collecting historical standard information and constructing a standard information knowledge graph, and at the same time performing information annotation to obtain standard information annotation labels; S2, acquiring the standard information to be compared in real time, performing information segmentation on the standard information to be compared to obtain standard information blocks to be compared, and performing feature extraction on the standard information blocks to be compared to obtain standard information features to be compared; S3, performing information semantic recognition on the standard information features to be compared until the acquired semantic recognition index meets the semantic comparison index, and dividing the standard information to be compared into standard information annotation labels that meet the semantic comparison index, wherein the semantic recognition index is used to describe the degree of conformity between the standard information to be compared and the standard information annotation labels in the standard information knowledge graph.

[0010] Furthermore, the specific steps of the information segmentation are: normalizing the format of the standard information to be compared to obtain the first standard information to be compared; identifying the key information of the first standard information to be compared, and segmenting the first standard information to be compared into information blocks; automatically verifying the segmented information blocks to obtain a segmentation accuracy index. If the segmentation accuracy index meets the reference verification index, the segmented information blocks are recorded as standard information blocks to be compared of the standard information to be compared. The segmentation accuracy index is used to describe the accuracy of the information segmentation of the standard information to be compared.

[0011] Furthermore, the method for obtaining the segmentation accuracy index is as follows: obtaining the automated verification index of the standard information to be compared from the automated verification process, and obtaining the reference verification total time of the automated verification, wherein the automated verification index includes the verification coverage, the verification environment setup time, the verification execution time, and the verification failure rate; obtaining the segmentation accuracy index of the information segmentation of the standard information to be compared according to the obtained automated verification index and the reference verification total time; the segmentation accuracy index is calculated using the following formula:

[0012]

[0013] Where, e is a natural constant, f is the number of the standard information to be compared, f = 1, 2, ..., F, F is the total number of standard information to be compared, JY f A represents the segmentation accuracy index of the fth standard information to be compared, f is the verification coverage of the fth standard information to be compared, The time to set up the verification environment for the fth standard information to be compared. is the verification execution time of the fth standard information to be compared, is the total reference verification time for the fth standard information to be compared, L f is the verification failure rate of the fth standard information to be compared, and δ is the lowest point of the verification coverage.

[0014] Furthermore, the specific method of feature extraction is as follows: special characters and punctuation marks are removed from the standard information block to be compared and the uppercase and lowercase formats are unified to obtain the first standard information block to be compared; the features of the information block to be evaluated of the first standard information block to be compared are extracted and feature evaluation is performed to obtain a feature typical evaluation index, and the feature typical evaluation index is compared with the feature threshold range. If the corresponding feature threshold range is met, the corresponding feature of the information block to be evaluated is recorded as the standard information feature to be compared, and the feature typical evaluation index is used to describe the degree of matching between the feature of the information block to be evaluated and the standard information block to be compared.

[0015] Furthermore, the method for obtaining the characteristic typicality evaluation index is as follows: obtain characteristic typicality indicators of the characteristics of the information block to be evaluated, and obtain characteristic typicality coefficients according to the characteristic typical indicators, wherein the characteristic typical indicators include characteristic information gain, characteristic information value and characteristic variance expansion degree, and the characteristic typical coefficients include characteristic information gain coefficient, characteristic information value coefficient and characteristic variance expansion degree coefficient; obtain the characteristic typicality evaluation index of the characteristics of the information block to be evaluated according to the characteristic typical coefficient.

[0016] Furthermore, the specific steps of the information semantic recognition are as follows: Step 1, decomposing the standard information features to be compared into standard information phrase texts, and matching them with the reference standard information phrase texts in the standard information annotation tags to obtain the phrase text matching index, comparing the phrase text matching index with the phrase text matching threshold, if the phrase text matching threshold is met, proceeding to Step 2, the phrase text matching index is used to describe the degree of text matching between the standard information phrase text and the standard information annotation tags; Step 2, extracting the standard information semantics in the standard information features to be compared, and matching them with the reference standard information semantics in the standard information annotation tags to obtain the semantic matching index, comparing the semantic matching index with the semantic matching threshold, if the semantic matching threshold is met, proceeding to Step 3, the semantic matching index is used to describe the degree of semantic matching between the standard information semantics of the standard information to be compared and the standard information annotation tags; Step 3, obtaining the reference values ​​corresponding to the phrase text matching index and the semantic phrase matching index and their corresponding reference deviations, and thereby obtaining the semantic recognition index of the standard information to be compared.

[0017] Furthermore, the method for obtaining the semantic matching index is as follows: semantically divide the standard information semantics of the standard information to be compared to obtain standard information semantic words and standard information semantic sentences, and obtain the semantic matching vector of the standard information semantics, and at the same time obtain the reference semantic matching vector, the semantic matching vector includes a semantic matching word vector and a semantic matching sentence vector, the reference semantic matching vector includes a reference semantic matching word vector and a reference semantic matching sentence vector, and thereby obtain the semantic matching vector coefficient of the standard information to be compared, the semantic matching vector coefficient includes a semantic matching word vector coefficient and a semantic matching sentence vector coefficient; according to the obtained semantic matching vector coefficient, the semantic matching index of the standard information to be compared is obtained.

[0018] Furthermore, the semantic recognition index is calculated using the following formula:

[0019]

[0020] In the formula, SB f represents the semantic recognition index of the fth standard information to be compared, WB f is the phrase text matching index of the fth standard information to be compared, WB0 is the reference phrase text matching index, ΔWB is the phrase text matching index reference deviation, YY f is the semantic matching index of the fth standard information to be compared, YY0 is the reference semantic matching index, ΔYY is the reference deviation of the semantic matching index, and σ is the reference relative deviation of the phrase text matching index.

[0021] An embodiment of the present application provides a standard information comparison system based on semantic recognition, including: a knowledge graph construction module, a feature extraction module and a semantic recognition and comparison module; wherein the knowledge graph construction module is used to collect historical standard information and construct a standard information knowledge graph, and at the same time perform information annotation to obtain standard information annotation labels; the feature extraction module is used to obtain the standard information to be compared in real time, perform information segmentation on the standard information to be compared to obtain standard information blocks to be compared, and perform feature extraction on the standard information blocks to be compared to obtain standard information features to be compared; the semantic recognition and comparison module is used to perform information semantic recognition on the standard information features to be compared until the acquired semantic recognition index meets the semantic comparison index, and divide the standard information to be compared into standard information annotation labels that meet the semantic comparison index, and the semantic recognition index is used to describe the degree of conformity between the standard information to be compared and the standard information annotation labels in the standard information knowledge graph.

[0022] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0023] 1. By collecting historical standard information and constructing a standard information knowledge graph, and then annotating the information to obtain standard information annotation tags, the standard information to be compared obtained in real time is segmented to obtain standard information blocks to be compared, and feature extraction is performed to obtain standard information features to be compared, and finally the standard information features to be compared are semantically recognized, thereby achieving accurate comparison of standard information in the field of engineering technology, thereby improving the comparison accuracy of standard information in the field of engineering technology, and effectively solving the problem of low accuracy of standard information data in the standard information comparison process in the prior art.

[0024] 2. The segmentation accuracy index of the information segmentation of the standard information to be compared is obtained by acquiring the automated verification index of the standard information to be compared and the total reference verification time of the automated verification, and the accuracy of the information segmentation of the standard information to be compared is evaluated by the obtained segmentation accuracy index, thereby realizing the quantification of the accuracy of the information segmentation of the standard information to be compared, and further realizing a more accurate evaluation of the accuracy of the information segmentation of the standard information to be compared.

[0025] 3. By obtaining the semantic matching vector of the standard information semantics and the reference semantic matching vector, the semantic matching vector coefficient of the standard information to be compared and the semantic matching index of the standard information to be compared are obtained, and the semantic matching degree between the standard information semantics of the standard information to be compared and the standard information annotation label is evaluated by the obtained semantic matching index, thereby realizing the quantification of the semantic matching degree between the standard information semantics of the standard information to be compared and the standard information annotation label, and further realizing a more accurate evaluation of the semantic matching degree between the standard information semantics of the standard information to be compared and the standard information annotation label. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A flow chart of a standard information comparison method based on semantic recognition provided in an embodiment of the present application;

[0027] Figure 2 A statistical graph showing changes in the semantic matching index provided in the embodiment of the present application;

[0028] Figure 3 A schematic diagram of the structure of a standard information comparison system based on semantic recognition provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] The embodiments of the present application solve the problem of low accuracy of standard information data in the standard information comparison process in the prior art by providing a standard information comparison method and system based on semantic recognition. The method collects historical standard information and constructs a standard information knowledge graph, then performs information annotation to obtain standard information annotation labels, and then obtains the standard information to be compared in real time, performs information segmentation on the standard information to be compared to obtain standard information blocks to be compared, and performs feature extraction on the standard information blocks to be compared to obtain standard information features to be compared. Finally, information semantic recognition is performed on the standard information features to be compared until the acquired semantic recognition index meets the semantic comparison index, and the standard information to be compared is divided into standard information annotation labels that meet the semantic comparison index, thereby improving the comparison accuracy of standard information in the field of engineering technology.

[0030] The technical solution in the embodiment of the present application is to solve the problem of low accuracy of standard information data in the above-mentioned standard information comparison process, and the overall idea is as follows:

[0031] By collecting historical standard information and constructing a standard information knowledge graph, information annotation is performed to obtain standard information annotation tags, and then the standard information to be compared obtained in real time is segmented to obtain standard information blocks to be compared, and feature extraction is performed to obtain standard information features to be compared. Finally, information semantics recognition and division are performed on the standard information features to be compared, thereby improving the comparison accuracy of standard information in the field of engineering technology.

[0032] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0033] like Figure 1 As shown, it is a flow chart of the standard information comparison method based on semantic recognition provided by an embodiment of the present application, and the method includes the following steps: S1, collecting historical standard information and constructing a standard information knowledge graph, and at the same time performing information annotation to obtain standard information annotation labels; S2, acquiring the standard information to be compared in real time, performing information segmentation on the standard information to be compared to obtain standard information blocks to be compared, and performing feature extraction on the standard information blocks to be compared to obtain standard information features to be compared; S3, performing information semantic recognition on the standard information features to be compared until the acquired semantic recognition index meets the semantic comparison index, and dividing the standard information to be compared into standard information annotation labels that meet the semantic comparison index, and the semantic recognition index is used to describe the degree of conformity between the standard information to be compared and the standard information annotation labels in the standard information knowledge graph.

[0034] In this embodiment, the historical standard information is the preset standard within a preset time interval, and the standard information annotation tag is used to divide the historical standard information into different categories. The standard information annotation tag generally includes a technical standard tag, an engineering specification tag and a design specification tag, which can be specifically refined according to actual conditions; the standard information knowledge graph is used to store the historical standard information according to the standard information annotation tag; the standard information to be compared is the engineering technical standard obtained in real time, and the engineering technical standard includes technical standards, engineering specifications and design specifications, such as the "Technical Specifications for Structural Concrete Performance" formulated by the China Concrete and Cement Products Association, which is used for the design and construction of concrete structures, and the "General Specifications for Steel Structures" issued by the China Steel Structure Association, which is applicable to the design of steel structure buildings and bridges; the standard information block to be compared is the sub-information obtained after information segmentation; the standard information feature to be compared is used to vectorize the standard information block to be compared; the semantic comparison index is set by the preset personnel, and the tag feedback is used to put the standard information to be compared that meets the semantic comparison index into the corresponding standard information annotation tag; after the standard information to be compared is divided into the standard information annotation tag that meets the semantic comparison index, tag feedback is also required; the comparison accuracy of standard information in the field of engineering technology is improved.

[0035] Furthermore, the construction process of the standard information knowledge graph is as follows: extract historical standard information entities from the historical standard information and label the historical standard information entities with unique historical standard information identifiers; extract entity relationships between historical standard information entities from the historical standard information and convert unstructured and semi-structured data in the historical standard information entities and entity relationships into structured data; convert historical standard information entities and entity relationships into nodes and edges of the standard information knowledge graph to obtain the standard information knowledge graph and store it in a graph database, which is used to store and manage historical standard information in the standard information knowledge graph.

[0036] In this embodiment, the extraction of historical standard information entities requires the use of a named entity recognition (NER) model for entity recognition, the input of which is historical standard information, and the output of which is a historical standard information entity; a historical standard information entity is a basic unit used to describe standard information within a preset time interval, and a historical standard information identifier is a code that uniquely identifies, distinguishes, and references a historical standard information entity within a preset time interval; based on the identified entities and relationships, a structured data schema is defined, corresponding fields are designed for each entity and relationship to ensure that the data can be fully and accurately represented, and the identified entities and relationships are mapped to corresponding fields of the structured data to achieve conversion of unstructured and semi-structured data into structured data, the mapping rules of which come from a database, such as converting text data into comma-separated values ​​(CSV) and object notation (JavaScript Object Notation, JSON); entity relationships include entity associations, entity effects and entity influences; use graph theory to transform historical standard information entities and entity relationships into nodes and edges of the standard information knowledge graph, that is, historical standard information entities are nodes of the standard information knowledge graph, and entity relationships are edges of the standard information knowledge graph, and the standard information knowledge graph is obtained and stored in the graph database; the accuracy of constructing the standard information knowledge graph is improved, and the comparison accuracy of standard information in the field of engineering technology is further improved.

[0037] Furthermore, the specific steps of information segmentation are: normalizing the format of the standard information to be compared to obtain the first standard information to be compared; identifying the key information of the first standard information to be compared, and segmenting the first standard information to be compared into information blocks; automatically verifying the segmented information blocks to obtain a segmentation accuracy index. If the segmentation accuracy index meets the reference verification index, the segmented information blocks are recorded as standard information blocks to be compared of the standard information to be compared, otherwise, the information segmentation is performed again. The segmentation accuracy index is used to describe the accuracy of the information segmentation of the standard information to be compared.

[0038] In this embodiment, format normalization is used to make the encoding method of the standard information to be compared consistent, and the first standard information to be compared is the information obtained after format normalization; it is necessary to combine the natural language processing library to divide the first standard information to be compared into information blocks, and the specific steps are: use the word_tokenize function of NLTK to segment the text, segment the text into a list of words or phrases, and define regular expression rules to identify noun phrases in the text, and then use the defined regular expression analyzer to segment the text after part-of-speech tagging, and finally traverse the tree structure of each sentence to extract the noun phrases as information blocks; key information includes information title, information list item and key data, and the information block is the part of the first standard information to be compared with complete semantics and functions; the information segmentation efficiency of the standard information to be compared is improved, and the comparison accuracy of standard information in the field of engineering technology is further improved.

[0039] Furthermore, the method for obtaining the segmentation accuracy index is as follows: extract the automated verification index of the standard information to be compared from the automated verification process, and obtain the reference verification total time of the automated verification, the automated verification index includes the verification coverage, the verification environment setup time, the verification execution time and the verification failure rate; obtain the segmentation accuracy index of the information segmentation of the standard information to be compared according to the obtained automated verification index and the reference verification total time; the segmentation accuracy index is calculated using the following formula:

[0040]

[0041] Where, e is a natural constant, f is the number of the standard information to be compared, f = 1, 2, ..., F, F is the total number of standard information to be compared, JY f A represents the segmentation accuracy index of the fth standard information to be compared, f is the verification coverage of the fth standard information to be compared, The time to set up the verification environment for the fth standard information to be compared. is the verification execution time of the fth standard information to be compared, is the total reference verification time for the fth standard information to be compared, L f is the verification failure rate of the fth standard information to be compared, and δ is the lowest point of the verification coverage.

[0042] In this embodiment, the verification coverage rate indicates the proportion of information blocks actually verified in the automated verification process to all information blocks, the verification environment construction time indicates the time required from the start of the verification environment construction to the time when the environment supports the verification work, the verification execution time indicates the time required from the start of the verification process to the end of the verification process, the verification failure rate indicates the proportion of information blocks that failed the verification in the automated verification process to all information blocks, and the reference verification total time is set by the preset personnel; the lowest extreme point of the verification coverage rate is set by the preset professional personnel, generally 0.8, and can be adjusted according to the actual application scenario; when the verification coverage rate is higher than its lowest extreme point, that is, the proportion of information blocks actually verified in the automated verification process to all information blocks is higher than 80%, the segmentation accuracy index is meaningful, and when the verification coverage rate is lower than its lowest extreme point, the deviation of the segmentation accuracy index is large due to the fact that the information blocks actually verified are too low, so it has no practical significance; the accuracy of information segmentation of the standard information to be compared is digitized, and the comparison accuracy of standard information in the field of engineering technology is further improved.

[0043] Furthermore, the specific method of feature extraction is as follows: special characters and punctuation marks are removed from the standard information block to be compared and the uppercase and lowercase formats are unified to obtain the first standard information block to be compared; the features of the information block to be evaluated of the first standard information block to be compared are extracted and the features are evaluated to obtain a feature typical evaluation index, and the feature typical evaluation index is compared with the feature threshold range. If the corresponding feature threshold range is met, the corresponding feature of the information block to be evaluated is recorded as the feature of the standard information to be compared, otherwise the feature of the information block to be evaluated is obtained again. The feature typical evaluation index is used to describe the degree of matching between the feature of the information block to be evaluated and the standard information block to be compared, and the feature threshold range is preset by personnel.

[0044] In this embodiment, the unified case format is to change the uppercase format in the standard information block to be compared into the lowercase format; the feature threshold range is used to determine whether the feature of the information block to be evaluated can accurately represent the first standard information block to be compared. If the obtained feature typical evaluation index is within the feature threshold range, it indicates that the feature of the information block to be evaluated has the ability to represent the first standard information block to be compared. If it is not within the feature threshold range, it indicates that the feature of the information block to be evaluated does not have the representative ability and needs to be discarded. The improvement of the matching degree of feature extraction of the standard information block to be compared is achieved, and the comparison accuracy of standard information in the field of engineering technology is further improved.

[0045] Furthermore, a method for obtaining a characteristic typicality evaluation index is as follows: obtaining characteristic typicality indicators of the characteristics of the information block to be evaluated, obtaining characteristic typicality coefficients according to the characteristic typical indicators, the characteristic typical indicators including characteristic information gain, characteristic information value and characteristic variance expansion degree, and the characteristic typical coefficients including characteristic information gain coefficient, characteristic information value coefficient and characteristic variance expansion degree coefficient; the characteristic information gain indicates the degree of reduction in uncertainty of the first band of comparison information blocks after the characteristics of the information block to be evaluated are introduced, the characteristic information value indicates the degree of influence of the characteristics of the information block to be evaluated on the target variable, and the characteristic variance expansion degree indicates the degree of multicollinearity between the characteristics of the information block to be evaluated; the characteristic information gain coefficient is the difference between the characteristic information gain and its reference value, the characteristic information value coefficient is the difference between the characteristic information value and its reference value, and the characteristic variance expansion degree coefficient is the difference between the characteristic variance expansion degree and its reference value; according to the characteristic typicality coefficient, the characteristic typicality evaluation index of the characteristics of the information block to be evaluated is obtained.

[0046] In this embodiment, the characteristic typicality evaluation index is calculated using the following formula:

[0047]

[0048] Where g is the number of the information block feature to be evaluated, g = 1, 2, ..., G, G is the total number of information block features to be evaluated, DX g is the characteristic typical evaluation index of the feature of the g-th information block to be evaluated, C g is the characteristic information gain coefficient of the g-th information block feature to be evaluated, ΔC is the reference deviation of the characteristic information gain coefficient, J g is the characteristic information value coefficient of the feature of the g-th information block to be evaluated, ΔJ is the reference deviation of the characteristic information value coefficient, α g is the feature variance expansion coefficient of the g-th information block feature to be evaluated, and Δα is the feature variance expansion reference deviation.

[0049] Specifically, the expression of the feature information gain coefficient is Among them, C ′ g is the feature information gain of the g-th information block feature to be evaluated, is the reference feature information gain of the g-th information block feature to be evaluated; the expression of the feature information value coefficient is Among them, J ′ g is the characteristic information value of the feature of the g-th information block to be evaluated, is the reference feature information value of the g-th information block feature to be evaluated; the expression of the feature variance expansion coefficient is Among them, α ′ g is the feature variance expansion degree of the g-th information block feature to be evaluated, is the reference feature variance expansion of the g-th information block feature to be evaluated; uncertainty is usually quantified by information entropy, and information gain is the difference between the original information entropy of the data set and the conditional information entropy of the data set under given feature conditions; the degree of influence is calculated by comparing the ratio of black and white samples hit by the information block feature to be evaluated with the ratio of total black and white samples, and the value is between 0.3 and 0.5; the numerical matching degree between the information block feature to be evaluated and the standard information block to be compared is realized, which further improves the comparison accuracy of standard information in the field of engineering technology.

[0050] It should be understood that the reference feature information gain is obtained by analyzing and counting the feature information gains of the features of the information block to be evaluated and taking the average value, the reference feature information value is obtained by analyzing and counting the feature information values ​​of the features of the information block to be evaluated and taking the average value, and the reference feature variance expansion is obtained by analyzing and counting the feature variance expansion of the features of the information block to be evaluated and taking the average value.

[0051] Furthermore, the specific steps of information semantic recognition are as follows: Step 1, decompose the standard information features to be compared into standard information phrase texts, and match them with the reference standard information phrase texts in the standard information annotation labels according to a preset first correspondence to obtain a phrase text matching index, compare the phrase text matching index with the phrase text matching threshold, if the phrase text matching threshold is met, proceed to step 2, otherwise replace the standard information annotation labels for rematching until the obtained phrase text matching index meets the phrase text matching threshold, the phrase text matching index is used to describe the degree of text matching between the standard information phrase text and the standard information annotation labels, the standard information phrase text represents the words and phrases decomposed from the standard information features to be compared, and the phrase text matching threshold is set by the preset personnel.

[0052] The standard information to be compared is in text form, so the features of the standard information to be compared are also in text form.

[0053] Decomposing the standard information features to be compared into standard information phrase texts is performed according to pre-set rules, and the pre-set rules are rules in a preset database.

[0054] The preset first corresponding relationship is a pre-mapped set in a preset database, and the standard information phrase text is input and mapped to obtain a corresponding output, and the phrase text matching index is obtained according to the mapping value matching.

[0055] The phrase text matching threshold is an interval. Meeting the phrase text matching threshold means that the phrase text matching index is within the interval of the phrase text matching threshold, and not meeting the threshold means that the phrase text matching index is not within the interval.

[0056] The method for obtaining the phrase text matching index is as follows: obtaining the phrase text matching data of the standard information to be compared from the standard information phrase text, the phrase text matching data including the phrase text position and the phrase text density, and obtaining the corresponding reference value and reference deviation according to the phrase text matching data; obtaining the phrase text matching index of the standard information to be compared according to the obtained phrase text matching data, the corresponding reference value and the reference deviation; the phrase text matching index is calculated using the following formula:

[0057]

[0058] Where, e is a natural constant, f is the number of the standard information to be compared, f = 1, 2, ..., F, F is the total number of standard information to be compared, h is the number of the standard information phrase text, h = 1, 2, ..., H, H is the total number of standard information phrase texts, WB f is the phrase text matching index of the fth standard information to be compared, is the phrase text position of the hth standard information phrase text of the fth standard information to be compared, is the reference phrase text position of the hth standard information phrase text of the fth standard information to be compared, ΔQ is the phrase text position reference deviation, W f is the phrase text density of the fth standard information to be compared, is the reference phrase text density of the fth standard information to be compared, ΔW f It is the phrase text density reference deviation of the h-th standard information phrase text of the f-th standard information to be compared.

[0059] Specifically, let the phrase text be offset Phrase text partial density is the phrase text offset position of the hth standard information phrase text of the fth standard information to be compared, W' f is the phrase text partial density of the fth standard information to be compared, then the phrase text matching index can be simplified as: When H=1, the statistical table of changes in the phrase text matching index is shown in Table 1:

[0060] Table 1 Statistics of changes in phrase text matching index

[0061]

[0062]

[0063] It can be seen from the above table that the phrase text matching index cannot be directly obtained through a single data, but needs to be obtained through a comprehensive analysis of the phrase text bias position and the phrase text bias density. For example, the phrase text bias position of the first group of data is greater than the phrase text bias position in the fourth group of data, but the final phrase text matching index is smaller than that of the fourth group because the phrase text bias density also has an impact, so a comprehensive analysis is required.

[0064] Step 2: Extract the standard information semantics from the standard information features to be compared, and match them with the reference standard information semantics in the standard information annotation label according to a preset second correspondence to obtain a semantic matching index. Compare the semantic matching index with the semantic matching threshold. If the semantic matching threshold is met, proceed to step 3. Otherwise, replace the standard information annotation label for rematching until the obtained semantic matching index meets the semantic matching threshold. The semantic matching index is used to describe the degree of semantic matching between the standard information semantics of the standard information to be compared and the standard information annotation label. The standard information semantics represents the meaning of the standard information features to be compared extracted from the standard information features to be compared. The semantic matching threshold is set by the preset personnel.

[0065] The frequency of occurrence of words in the standard information features to be compared is counted as a semantic representation of the standard information features to be compared, that is, the extracted standard information semantics.

[0066] The semantics of the reference standard information refers to the parts of speech in the reference standard information, including nouns, verbs, adjectives, etc.

[0067] The preset second corresponding relationship is a pre-mapped set in a preset database, the standard information semantics is input and mapped, a corresponding output is obtained, and a semantic matching index is obtained according to the mapping value matching.

[0068] The semantic matching threshold is an interval. If the semantic matching threshold is met, the semantic matching index is within the interval of the semantic matching threshold, and if it is not within the interval, it is not met.

[0069] Step three, obtaining reference values ​​corresponding to the phrase text matching index and the semantic phrase matching index and their corresponding reference deviations, and based on these, obtaining the semantic recognition index of the standard information to be compared.

[0070] In this embodiment, accurate recognition of the information semantics of the standard information to be compared is achieved, further improving the comparison accuracy of the standard information in the field of engineering technology.

[0071] Furthermore, a method for obtaining a semantic matching index is as follows: semantically divide the standard information semantics of the standard information to be compared to obtain standard information semantic words and standard information semantic sentences, and obtain a semantic matching vector of the standard information semantics, and at the same time obtain a reference semantic matching vector, the standard information semantic words are the semantics of words in the standard information semantics, the standard information semantic sentences are the semantics of clauses in the standard information semantics, the semantic matching vector includes a semantic matching word vector and a semantic matching sentence vector, the reference semantic matching vector includes a reference semantic matching word vector and a reference semantic matching sentence vector, based on which the semantic matching vector coefficients of the standard information to be compared are obtained, the semantic matching vector coefficients include a semantic matching word vector coefficient and a semantic matching sentence vector coefficient, the semantic matching word vector represents the vectorization of the standard information semantic words, and the semantic matching sentence vector represents the vectorization of the standard information semantic sentences; the semantic matching index of the standard information to be compared is obtained based on the obtained semantic matching vector coefficients.

[0072] In this embodiment, the semantic matching index is calculated using the following formula:

[0073]

[0074] Where f is the number of the standard information to be compared, f = 1, 2, ..., F, F is the total number of standard information to be compared, b is the number of the semantic words of the standard information, b = 1, 2, ..., B, B is the total number of semantic words of the standard information, d is the number of the semantic sentences of the standard information, d = 1, 2, ..., D, D is the total number of semantic sentences of the standard information, YY f represents the semantic matching index of the fth standard information to be compared, is the semantic matching word vector coefficient of the bth standard information semantic word of the fth standard information to be compared, is the semantic matching sentence vector coefficient of the dth standard information semantic sentence of the fth standard information to be compared.

[0075] Specifically, the process of semantic segmentation is: using the word segmentation tool in natural language processing technology to segment the standard information text into separate words, and then marking the part of speech of each semantic word, such as noun, verb, adjective, etc., which is the standard information semantic word, and using syntactic analysis technology to identify the subject, predicate, object and other components in the sentence, as well as the relationship between them, based on the results of syntactic analysis, the related semantic words are combined into a complete standard information semantic sentence; the semantic matching word vector coefficient is the cosine value of the semantic matching word vector, and its expression is: The semantic matching sentence vector coefficient is the cosine value of the semantic matching sentence vector, and its expression is: is the semantic matching word vector of the bth standard information semantic word of the fth standard information to be compared, is the reference semantic matching word vector of the bth standard information semantic word of the fth standard information to be compared, is the semantic matching sentence vector of the dth standard information semantic sentence of the fth standard information to be compared, It is the reference semantic matching sentence vector of the dth standard information semantic sentence of the fth standard information to be compared; it realizes the numeration of the semantic matching degree between the standard information semantics of the standard information to be compared and the standard information annotation label, and further improves the comparison accuracy of standard information in the field of engineering technology.

[0076] Furthermore, the semantic recognition index is calculated using the following formula:

[0077]

[0078] In the formula, SB f represents the semantic recognition index of the fth standard information to be compared, WB f is the phrase text matching index of the fth standard information to be compared, WB0 is the reference phrase text matching index, ΔWB is the phrase text matching index reference deviation, YY f is the semantic matching index of the fth standard information to be compared, YY0 is the reference semantic matching index, ΔYY is the reference deviation of the semantic matching index, and σ is the reference relative deviation of the phrase text matching index.

[0079] In this embodiment, the reference phrase text matching index, the phrase text matching index reference deviation, the reference semantic matching index, the semantic matching index reference deviation and the reference relative deviation of the phrase text matching index are all set by the preset personnel; Micro-semantic matching index WB f ′ is the text matching index of the micro-word group of the fth standard information to be compared, YY f ′ is the micro-semantic matching index of the fth standard information to be compared. When the micro-phrase text matching index is greater than its reference relative deviation, the semantic recognition index can be simplified as: For example, when taking 0.9, Figure 2As shown, it is a statistical graph of the changes in the semantic recognition index provided in the embodiment of the present application. It can be seen from the figure that the micro-word group text matching index and the micro-semantic matching index are positively correlated with the comprehensive detection anomaly index; when the micro-word group text matching index increases, the larger the semantic recognition index, the greater the degree of conformity between the standard information to be compared and the standard information annotation labels in the standard information knowledge graph; when the micro-semantic matching index increases, the larger the semantic recognition index, the greater the degree of conformity between the standard information to be compared and the standard information annotation labels in the standard information knowledge graph; the digitization of the degree of conformity between the standard information to be compared and the standard information annotation labels in the standard information knowledge graph is realized, which further improves the comparison accuracy of standard information in the field of engineering technology.

[0080] like Figure 3 As shown, it is a structural schematic diagram of the standard information comparison system based on semantic recognition provided in an embodiment of the present application. The standard information comparison system based on semantic recognition provided in an embodiment of the present application includes: a knowledge graph construction module, a feature extraction module and a semantic recognition comparison module; wherein the knowledge graph construction module is used to collect historical standard information and construct a standard information knowledge graph, and at the same time perform information annotation to obtain standard information annotation labels; the feature extraction module is used to obtain the standard information to be compared in real time, perform information segmentation on the standard information to be compared to obtain standard information blocks to be compared, and perform feature extraction on the standard information blocks to be compared to obtain standard information features to be compared; the semantic recognition comparison module is used to perform information semantic recognition on the features of the standard information to be compared until the acquired semantic recognition index meets the semantic comparison index, and divide the standard information to be compared into standard information annotation labels that meet the semantic comparison index, and the semantic recognition index is used to describe the degree of conformity between the standard information to be compared and the standard information annotation labels in the standard information knowledge graph.

[0081] In this embodiment, the historical standard information is a preset standard within a preset time interval, the standard information annotation tag is used to divide the historical standard information into different categories, and the standard information knowledge graph is used to store the historical standard information according to the standard information annotation tag; the standard information to be compared is the engineering technical standard acquired in real time, and the engineering technical standard includes technical standards, engineering specifications and design specifications. The features of the standard information to be compared are used to vectorize the standard information block to be compared; the bag-of-words model is used to extract the features of the standard information to be compared, and the preset feature representation method can be used to extract features of the standard information block to be compared. For example, the frequency of each word in the bag-of-words model in the text can be counted, and a vector representation of the text can be constructed; the semantic comparison index is set by the preset personnel, and the label feedback is used to divide the standard information to be compared that meets the semantic comparison index into the corresponding standard information annotation tag; the device-based comparison of standard information in the field of engineering technology is realized, and the comparison accuracy of standard information in the field of engineering technology is further improved.

[0082] To summarize, the embodiment of the present application collects historical standard information and constructs a standard information knowledge graph, then performs information annotation to obtain standard information annotation tags, then performs information segmentation on the standard information to be compared obtained in real time to obtain standard information blocks to be compared, and performs feature extraction to obtain standard information features to be compared, and finally performs information semantic recognition on the standard information features to be compared, thereby achieving accurate comparison of standard information in the field of engineering technology, thereby improving the comparison accuracy of standard information in the field of engineering technology, and effectively solving the problem of low accuracy of standard information data in the standard information comparison process in the prior art.

[0083] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0084] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0085] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0086] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0087] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0088] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. Standard information comparison method based on semantic recognition, It is characterized in that The following steps are involved: S1, collect historical standard information and build a standard information knowledge graph, and annotate the information to obtain standard information annotation labels; S2, obtaining the standard information to be compared in real time, performing information segmentation on the standard information to be compared to obtain standard information blocks to be compared, and performing feature extraction on the standard information blocks to obtain features of the standard information to be compared; S3, perform information semantic recognition on the features of the standard information to be compared until the obtained semantic recognition index meets the semantic comparison index, and divide the standard information to be compared into standard information annotation labels that meet the semantic comparison index. The semantic recognition index is used to describe the degree of conformity between the standard information to be compared and the standard information annotation labels in the standard information knowledge graph.

2. The standard information comparison method based on semantic recognition as claimed in claim 1, It is characterized in that The construction process of the standard information knowledge graph is as follows: extracting historical standard information entities from the historical standard information and assigning unique historical standard information identifiers to the historical standard information entities; Extracting entity relationships between entities of the historical standard information from the historical standard information and converting unstructured and semi-structured data in the entities and entity relationships of the historical standard information into structured data; The historical standard information entities and entity relationships are converted into nodes and edges of the standard information knowledge graph to obtain the standard information knowledge graph and store it in the graph database.

3. The standard information comparison method based on semantic recognition as claimed in claim 1, It is characterized in that The specific steps of information segmentation are: Normalizing the format of the standard information to be compared to obtain the first standard information to be compared; Identifying key information of the first standard information to be compared, and dividing the first standard information to be compared into information blocks; The segmented information blocks are automatically verified to obtain a segmentation accuracy index. If the segmentation accuracy index meets the reference verification index, the segmented information blocks are recorded as standard information blocks to be compared with the standard information to be compared. The segmentation accuracy index is used to describe the accuracy of information segmentation of the standard information to be compared.

4. The standard information comparison method based on semantic recognition as claimed in claim 3, It is characterized in that The method for obtaining the segmentation accuracy index is as follows: Obtaining automated verification indicators of the standard information to be compared from the automated verification process, and obtaining a reference total verification time for the automated verification, wherein the automated verification indicators include verification coverage, verification environment setup time, verification execution time, and verification failure rate; Obtaining a segmentation accuracy index of information segmentation of the standard information to be compared according to the obtained automated verification index and the reference verification total time; The segmentation accuracy index is calculated using the following formula: Where, e is a natural constant, f is the number of the standard information to be compared, f = 1, 2, ..., F, F is the total number of standard information to be compared, JY f A represents the segmentation accuracy index of the fth standard information to be compared, f is the verification coverage of the fth standard information to be compared, The time to set up the verification environment for the fth standard information to be compared. is the verification execution time of the fth standard information to be compared, is the total reference verification time for the fth standard information to be compared, L f is the verification failure rate of the fth standard information to be compared, and δ is the lowest point of the verification coverage.

5. The standard information comparison method based on semantic recognition as claimed in claim 1, It is characterized in that The specific method of feature extraction is as follows: Remove special characters and punctuation marks from the standard information block to be compared and unify the uppercase and lowercase formats to obtain a first standard information block to be compared; The features of the information block to be evaluated of the first standard information block to be compared are extracted and feature evaluation is performed to obtain a feature typical evaluation index, and the feature typical evaluation index is compared with the feature threshold range. If the corresponding feature threshold range is met, the corresponding information block feature to be evaluated is recorded as the standard information feature to be compared, and the feature typical evaluation index is used to describe the degree of matching between the feature of the information block to be evaluated and the standard information block to be compared.

6. The standard information comparison method based on semantic recognition as claimed in claim 5, It is characterized in that The method for obtaining the characteristic typical evaluation index is as follows: Obtain characteristic typical indicators of the characteristics of the information block to be evaluated, and obtain characteristic typical coefficients according to the characteristic typical indicators, wherein the characteristic typical indicators include characteristic information gain, characteristic information value, and characteristic variance expansion degree, and the characteristic typical coefficients include characteristic information gain coefficient, characteristic information value coefficient, and characteristic variance expansion degree coefficient; The characteristic typicality evaluation index of the characteristic of the information block to be evaluated is obtained according to the characteristic typicality coefficient.

7. The standard information comparison method based on semantic recognition as claimed in claim 1, It is characterized in that The specific steps of information semantic recognition are as follows: Step 1: decompose the standard information features to be compared into standard information phrase texts, and match them with the reference standard information phrase texts in the standard information annotation label according to a preset first correspondence to obtain a phrase text matching index, and compare the phrase text matching index with a phrase text matching threshold. If the phrase text matching threshold is met, proceed to step 2, wherein the phrase text matching index is used to describe the degree of text matching between the standard information phrase text and the standard information annotation label; Step 2: extract the standard information semantics in the standard information features to be compared, and match the standard information semantics in the standard information annotation label with the reference standard information semantics according to the preset second corresponding relationship to obtain a semantic matching index, and compare the semantic matching index with the semantic matching threshold. If the semantic matching threshold is met, proceed to step 3, wherein the semantic matching index is used to describe the degree of semantic matching between the standard information semantics of the standard information to be compared and the standard information annotation label; Step three, obtaining reference values ​​corresponding to the phrase text matching index and the semantic phrase matching index and their corresponding reference deviations, and based on these, obtaining the semantic recognition index of the standard information to be compared.

8. The standard information comparison method based on semantic recognition as claimed in claim 7, It is characterized in that The method for obtaining the semantic matching index is as follows: Semantically dividing the standard information semantics of the standard information to be compared to obtain standard information semantic words and standard information semantic sentences, and obtaining a semantic matching vector of the standard information semantics, and simultaneously obtaining a reference semantic matching vector, the semantic matching vector including a semantic matching word vector and a semantic matching sentence vector, the reference semantic matching vector including a reference semantic matching word vector and a reference semantic matching sentence vector, and thereby obtaining a semantic matching vector coefficient of the standard information to be compared, the semantic matching vector coefficient including a semantic matching word vector coefficient and a semantic matching sentence vector coefficient; The semantic matching index of the standard information to be compared is obtained according to the obtained semantic matching vector coefficient.

9. The standard information comparison method based on semantic recognition as claimed in claim 8, It is characterized in that The semantic recognition index is calculated using the following formula: In the formula, SB f represents the semantic recognition index of the fth standard information to be compared, WB f is the phrase text matching index of the fth standard information to be compared, WB 0 is the reference phrase text matching index, ΔWB is the phrase text matching index reference deviation, YY f is the semantic matching index of the fth standard information to be compared, YY 0 is the reference semantic matching index, ΔYY is the reference deviation of the semantic matching index, and σ is the reference relative deviation of the phrase text matching index.

10. Standard information comparison system based on semantic recognition, It is characterized in that include: Knowledge graph construction module, feature extraction module and semantic recognition and comparison module; The knowledge graph construction module is used to collect historical standard information and construct a standard information knowledge graph, and to annotate the information to obtain a standard information annotation label; The feature extraction module is used to obtain the standard information to be compared in real time, perform information segmentation on the standard information to be compared to obtain the standard information block to be compared, and perform feature extraction on the standard information block to obtain the standard information feature to be compared; The semantic recognition and comparison module is used to perform information semantic recognition on the features of the standard information to be compared until the obtained semantic recognition index meets the semantic comparison index, and divide the standard information to be compared into standard information annotation labels that meet the semantic comparison index. The semantic recognition index is used to describe the degree of conformity between the standard information to be compared and the standard information annotation labels in the standard information knowledge graph.

Citation Information

Cited By

  • Intelligent text labeling system, method and equipment oriented to standard semantic knowledge extraction

    CN120765857A