Standard system generation method based on standard optimization classification

Abstract

The invention discloses a standard system generation method based on standard optimization classification, and relates to the technical field of standard system digital management, and the method comprises the steps: extracting a subject term of each standard in an effective standard text set in the petroleum field, and constructing a standard subject term vector set in combination with an original standard text; based on the standard subject term vector set, the user demand and the number of levels given by the user, a solver is used for repeatedly executing the number of times corresponding to the number of levels to obtain a multi-level classification result, and the solver is used for determining subject term union set vectors of all standards under the category to which each standard belongs and generating a classification table; and determining a target subject term corresponding to each level in the multi-level classification result, and constructing a standard system corresponding to the original standard text through the multi-level classification result and the target subject term. According to the invention, rapid updating can be realized only by adjusting hierarchy parameters, and the maintenance cost is remarkably reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of standard system digital management, and particularly relates to a standard system generation method based on standard optimization classification. BACKGROUND

[0002] In the petroleum and natural gas, chemical engineering, automation control and other industries, the number of standards is growing explosively. Taking a natural gas sales company as an example, its business covers five major plates of LNG, LPG, branch pipeline, urban gas and gas-electricity new energy, and only in these fields, more than 2,000 national standards, industry standards, group standards and enterprise standards are involved. In the face of a large number of standards, enterprises need to quickly, scientifically and sustainably build and dynamically maintain a set of standards system covering all aspects, with clear levels and easy to search, to support compliance operation, quality control and technology upgrading. However, the traditional manual sorting method is time-consuming and difficult to adapt to the frequent updating needs brought by rapid technical iteration.

[0003] The current mainstream methods include the following categories: According to GB / T 13016-2018 "Standard System Construction Principles and Requirements", the standards are read by field experts one by one, and a multi-level framework is established by experience and manually classified. For example, related technology CN116737872A--policy text quantitative analysis method based on text mining and PMC index model, the steps are: collect policies -> TXT storage -> TF-IDF extract keywords -> co-occurrence matrix -> hierarchical clustering -> PMC index scoring, used for policy text commonality / personality research; or for example, related technology CN110232149A--"hot event detection method and system", the steps are: text preprocessing -> TF-IDF / Word2Vec vectorization -> AP or GSDMM clustering -> key word vector distance judgment -> merging similar clusters -> classifier output hot events; or for example, related technology CN119149736A--"quantitative scientific research project screening method based on patent trend analysis", the steps are: Chinese-BERT vectorization -> UMAP dimension reduction -> K-Means++ clustering -> TF-IDF+MMR extraction of technology / efficiency keywords -> construction of "technology-efficiency" matrix -> three-dimensional quantitative scoring of life cycle, innovation and market competitiveness; or for example, related technology CN117521638A--"text label determination method and device", the steps are: distributed representation model vectorization -> multi-level clustering -> TF-IDF extract candidate labels -> similar matrix merge synonyms -> attention weight determine final label.

[0004] The above prior art has many problems that are difficult to overcome, including: complete dependence on experts reading one by one, workload linearly expanding with the number of standards; expert cognitive differences leading to the same standard being classified into different categories, poor system consistency; manual adjustment is required after technical update, response cycle is measured in months. In related technology CN116737872A, only policy text is considered, the variable system is fixed as 9 main variables + a number of sub-variables, and the unique constraints of technical standard clauses, reference relationship and membership are not considered; when new technical dimensions are added, the variables need to be manually adjusted, and incremental updates are not possible. In related technology CN110232149A, although word vectors and clustering are used for hot event detection, the goal is to identify sudden events in social media short texts, and the "clustering-merging-classification" process does not solve the hard constraint that "standards can only belong to one category", resulting in overlapping or gaps in technical topics. In related technology CN119149736A, after patent clustering, "field experts exclude, replace, and manually summarize keywords" to generate technical efficacy expressions, the subjective step has not been eliminated; and the "maximum intra-class similarity and minimum inter-class similarity" is not mathematically modeled, making it difficult to directly output a hierarchical standard system. Related technology CN117521638A focuses on improving the semantic accuracy of labels, but lacks modeling of key constraints such as "unique classification of each document", "upper and lower limits of the number of categories", and "intra-class / inter-class difference quantification" for standard system construction, and the clustering results cannot be directly mapped to the standard framework.

[0005] In summary, existing technologies either remain in inefficient manual-dominated modes or are limited to specific scenarios such as policies, patents, and short texts, and generally lack a massive technology standard-oriented approach, therefore there is an urgent need for an automatic, scientific, and continuously evolving standard system generation method and system. SUMMARY

[0006] The technical problem to be solved by the present application is to overcome the deficiencies of the prior art, and specifically provides a standard system generation method based on standard optimization classification, as follows: 1) In a first aspect, the present application provides a standard system generation method based on standard optimization classification, and the specific technical solutions are as follows: Extract the subject headings of each standard in the effective standard text in the oil field, and construct a standard subject heading vector set in combination with the original standard text; Based on the standard subject heading vector set, user demand, and the number of levels given by the user, the solver is repeatedly executed for the corresponding number of times, and a multi-level classification result is obtained, the solver is used to determine the subject heading vector set of all standards in the category to which each standard belongs and generate a classification table; Determine the target subject headings corresponding to each level in the multi-level classification result, and construct a standard system corresponding to the original standard text through the multi-level classification result and the target subject headings.

[0007] The standard system generation method based on the standard optimization classification provided by the application has the following beneficial effects: By extracting the subject words of all effective standards in the oil field at one time and automatically constructing the subject word vector, combining the user demand and the given hierarchical quantity, repeatedly calculating by the solver until the multi-level classification result is output, the high time consumption and subjective deviation of manually interpreting standards one by one can be completely eliminated, and the standard system construction process is ensured to be objective, efficient and reproducible; the subject word union vector and the hierarchical constraint are repeatedly used to realize automatic clustering and dynamic expansion from coarse to fine, so that the system can be quickly updated by only adjusting the hierarchical parameters when the technology iterates or the business expands, and the maintenance cost is significantly reduced; the finally generated multi-level classification table directly corresponds to the original standard text of the target subject word, forming a visual standard system framework that can be searched and used immediately, and the precision and response speed of standard retrieval, compliance review and scientific research project are greatly improved.

[0008] On the basis of the above-mentioned scheme, the application can also be improved as follows.

[0009] Further, the process of constructing the standard subject word vector set is specifically: extracting the subject words of each standard by natural language processing technology, and forming a subject word set by the subject words of all standards; in the subject word set, representing each original standard as an n-dimensional zero vector, determining the target number k of the first subject word contained in the subject word set in each original standard, and assigning the kth dimension of the n-dimensional zero vector of each original standard to 1, to generate the corresponding optimization vector of each original standard; integrating the optimization vectors corresponding to all original standards to generate a standard subject word vector set.

[0010] The beneficial effects of the above-mentioned further scheme are: After extracting all the subject words, each standard is directly mapped to a binary n-dimensional vector, and only the positions corresponding to the k subject words appearing in the standard are set to 1, so that large-scale text vectorization can be completed within milliseconds, significantly reducing storage and calculation overhead; the mechanism of zero vector setting to 1 naturally retains the binary information of “appearance / non-appearance”, avoids the subjective disturbance introduced by the weight parameter, enables the subsequent solver to quickly converge in the same semantic space, and completely eliminates the deviation of manual scoring; the generated standard subject word vector set can be directly used as the input of any hierarchical clustering or mathematical programming model, realizing one-time construction and multiple reuse, and greatly improving the agility of standard system construction and dynamic expansion.

[0011] Further, the user demand includes: the category number demand range, the upper and lower limits of the number of standards of each category, and the upper and lower limits of the difference proximity.

[0012] Further, the process of determining the target subject word corresponding to each level in the multi-level classification result is specifically: According to the attribute information of all standard subject words corresponding to each level in the multi-level classification result, the target subject word corresponding to each level is determined. The attribute information includes the word length and the word frequency of the subject word.

[0013] Further, the process of determining the target subject word corresponding to each level in the multi-level classification result is specifically: Based on the standard system, the whole life cycle management of the oil management system is performed.

[0014] 2) In the second aspect, the application further provides a standard system generation system based on standard optimization classification, and the specific technical scheme is as follows: The extraction module is used for extracting the subject word of each standard in the effective standard text in the oil field, and constructing a standard subject word vector set in combination with the original standard text; The processing module is used for obtaining a multi-level classification result by repeatedly executing the corresponding number of times of the level number through a solver based on the standard subject word vector set, user demand and the level number given by the user, the solver being used for determining the union vector of the subject words of all standards in each category to which each standard belongs and generating a classification table; The generation module is used for determining the target subject word corresponding to each level in the multi-level classification result, and constructing a standard system corresponding to the original standard text through the multi-level classification result and the target subject word.

[0015] On the basis of the above-mentioned scheme, the application can be further improved as follows.

[0016] Further, the process of constructing the standard subject word vector set is specifically: The subject word of each standard is extracted through natural language processing technology, and the subject words of all standards are combined to form a subject word set; In the subject word set, each original standard is represented as an n-dimensional zero vector, the target number k of the first subject word contained in the subject word set in each original standard is determined, and the kth dimension of the n-dimensional zero vector of each original standard is assigned a value of 1 to generate an optimization vector corresponding to each original standard; All optimization vectors corresponding to the original standards are integrated to generate a standard subject word vector set.

[0017] Further, the user demand includes: The obtained category number demand range, the upper and lower limits of the number of standards of each category, and the upper and lower limits of the difference proximity.

[0018] Further, the process of determining the target subject word corresponding to each level in the multi-level classification result is specifically: According to attribute information of all the standard thesauri corresponding to each level in the multi-level classification result, determine the target thesaurus corresponding to each level; Wherein, the attribute information includes: word length and word frequency of the thesaurus.

[0019] Further, it further comprises: Based on the standard system, the oil management system is managed in the whole life cycle.

[0020] 3) In a third aspect, the present application also provides an electronic device, the electronic device comprising a processor, the processor being coupled with a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to enable the electronic device to implement any of the above methods.

[0021] 4) In a fourth aspect, the present application also provides a computer readable storage medium, the computer readable storage medium storing at least one computer program, the at least one computer program being loaded and executed by a processor to enable the computer to implement any of the above methods.

[0022] It should be noted that the technical solutions of the second to fourth aspects of the present application and the corresponding possible implementation manners have the beneficial effects as described above for the first aspect and the corresponding possible implementation manners, which will not be described here again. BRIEF DESCRIPTION OF DRAWINGS

[0023] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments with reference to the attached drawings: Figure 1 One of the flowcharts of a standard system generation method based on standard optimization classification according to an embodiment of the present application; Figure 2 The second flowchart of a standard system generation method based on standard optimization classification according to an embodiment of the present application; Figure 3 The standard thesaurus weight diagram of a standard system generation method based on standard optimization classification according to an embodiment of the present application; Figure 4 The tree structure diagram of a standard system generation method based on standard optimization classification according to an embodiment of the present application; Figure 5 The structural framework diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical scheme and advantages of the present application more clear, the embodiments of the present application will be further described in detail below with reference to the drawings.

[0025] As Figure 1and Figure 2 As shown in the figure, the standard system generation method based on standard optimization classification according to an embodiment of the application comprises the following steps: S1, extracting the subject words of each standard in the effective standard text set in the oil field, and constructing a standard subject word vector set in combination with the original standard text; S2, based on the standard subject word vector set, user demand and the number of levels given by the user, repeatedly executing the number of levels corresponding times through a solver to obtain a multi-level classification result, the solver being used to determine the subject word union vector of all standards in the category to which each standard belongs and generate a classification table; S3, determining the target subject words corresponding to each level in the multi-level classification result, and constructing the standard system corresponding to the original standard text through the multi-level classification result and the target subject words.

[0026] The standard system generation method based on standard optimization classification provided by the application has the following beneficial effects: By extracting the subject words of all effective standards in the oil field at one time and automatically constructing the subject word vector, combining the user demand with the given number of levels, repeatedly calculating by the solver until the multi-level classification result is output, the high time consumption and subjective bias of manually interpreting standards one by one can be completely eliminated, and the standard system construction process can be ensured to be objective, efficient and reproducible; the subject word union vector and the level constraint are repeatedly used to realize automatic clustering and dynamic expansion from coarse to fine, so that the system can be quickly updated by only adjusting the level parameter when the technology iterates or the business expands, thereby significantly reducing the maintenance cost; the finally generated multi-level classification table and target subject words directly correspond to the original standard text, forming a visual standard system framework that can be searched and used immediately, and the accuracy and response speed of standard retrieval, compliance review and scientific research project establishment are greatly improved.

[0027] It should be noted that the effective standard text set refers to the set of all standard texts that are currently effective, not abandoned or replaced at a specific time node (such as the current date). For example: national standards (GB, GB / T, etc.); industry standards (SY, HG, NB, SH, etc.); group standards (T / CAS, T / CECS, etc.); enterprise standards (Q / SY, Q / CNPC, etc.).

[0028] Each standard in the effective standard text set refers to a single standard text that is currently effective, i.e. an independent numbered and named standardized file. The specific features can be: Independent standard number (such as GB 17820-2018, SY / T 5922-2020, Q / SY 01005-2023). Complete text content (including cover, preface, scope, normative reference, technical provisions, appendix, etc.). Current effective version (the latest version that has not been abandoned or replaced).

[0029] In another embodiment of the present scheme, the specific implementation process of S1 can include: Through official channels, the existing standard catalogues of target fields such as oil and gas, chemical industry, and automation control are batched, and the GB, SY, HG, NB, T / CAS, T / CECS and enterprise standard Q / SY, Q / CNPC texts are retained only by using standard number rules, publication date and real-time cleaning of abolition announcements, and only the GB, SY, HG, NB, T / CAS, T / CECS and enterprise standard Q / SY, Q / CNPC texts in the state of "current effective" are retained; then, each standard PDF full text is downloaded to the local by calling Python requests or wget, and the pure text string is obtained by using pdfminer.six page by page analysis, and the cover, table of contents, appendix, reference, etc. Non-text content is deleted by using regular expressions, and the core chapters of "scope, normative reference, term definition, technical requirements, test method, test rule" are retained to form a single standard original text. Then, based on the oil industry self-defined dictionary and stop word table, jieba or pkuseg is used for accurate segmentation of each standard text, professional terms such as "wellhead device, LNG storage tank, SCADA system, integrity management" are loaded to prevent excessive segmentation, and the stop word table is used to remove "de, should, can" and other non-meaningful words; the word weight is calculated by using TF-IDF on the segmentation result, and the Top-N high weight words are taken as candidate subject headings, and the key phrases are extracted in the unsupervised mode by using TextRank or KeyBERT, and the union is reviewed by artificial de-duplication, and finally the subject heading list of the standard is obtained. The global subject heading dictionary is formed by merging and de-duplicating all the extracted subject headings of the standards, and is mapped to the index of dimension n in alphabetical order; then, the n-dimensional zero vector is constructed for each standard, if the subject heading hits the kth word in the dictionary, the kth dimension of the vector is 1, and the rest of the dimensions remain 0, and finally the standard subject heading vector set with the shape of [m x n] is obtained, wherein m is the total number of effective standards, and n is the number of global subject headings, the set not only retains the semantic features of the original text, but also can be directly input into the subsequent clustering or optimization model.

[0030] The method for automatically generating a standard system according to the present application comprises the following specific operation steps: (1) A natural language processing method is used to extract standard subject headings from all standards in a certain professional field related to a certain professional field, and a standard subject heading table is constructed.

[0031] All standard subject headings of all standards are extracted by comprehensively considering national, industry, enterprise and group standards related to a certain professional field, and the extracted standard subject headings are arranged in the order of Arabic numerals 1, 2, 3, …, n to form a set containing n standard subject headings. It can be expressed as: (1) Each standard is initialized as an n-dimensional zero vector. For each of the m standard subject headings extracted in the standard, all standard subject headings in the standard subject heading set are traversed in turn, and whether it is consistent with the standard subject heading in the standard. If consistent, the component at the position corresponding to the index of the standard subject heading set in the n-dimensional zero vector is assigned a value of 1. After the traversal of the m standard subject headings is completed, a standard subject heading vector containing m components with a value of 1 and n-m components with a value of 0 is formed. The standard subject heading vector is as follows: (2) The kth component of the standard subject heading vector v is denoted as vk, which has a value of 0 or 1.

[0032] In another embodiment of the present scheme, the specific implementation of S2 is as follows: After loading the standard subject heading vector set [m x n] in the memory, the hierarchical number L input by the user and the upper and lower limits of the expected number of categories per level [Kmin, Kmax] are read, and then the iterative solution loop is entered. In the first round, the original vector matrix is taken as the input, the "maximal sum of intra-class difference similarity absolute values and minimal sum of inter-class difference similarity absolute values" is set as the double objective, four types of hard constraints, i.e., "each standard belongs to only one class, the number of standards in each class ∈ [Nmin, Nmax], the total number of classes ∈ [Kmin, Kmax], and the difference similarity between any two standards ∈ [δmin, δmax]", are added, and the surrounding prey, bubble net attack, random search, and fluctuating foraging operators are iterated until the termination condition is met, and the primary classification table is output. Then, the subject heading vectors of all standards in each primary class are bitwise unioned to generate a new 1 x n union vector, which is taken as the input of the next round of clustering. The above optimization solution is repeated until L rounds are completed. The classification table is written back to SQLite immediately after each round is completed. The table structure includes the standard number, hierarchical number, class number, union vector, and the objective function value of the current round. The union vector is read in the next round, and the old table is emptied to only keep the latest classification result. Finally, the multi-level classification result table is exported after L rounds, in which each record gives the class affiliation of the standard in the 1…L hierarchical levels and the corresponding union vector, which can be directly used for subsequent hierarchical subject heading extraction and standard system visualization.

[0033] (2) Taking the membership of the standard (the standard mentioned in this step refers to the standard topic word vector in step (1)) to different standard categories (the standard category mentioned in this step refers to the category of standard topic word vector) as the decision variable, the objective is to maximize the sum of the absolute values ​​of the difference similarity between the standards within the standard category and minimize the sum of the absolute values ​​of the difference similarity between the standard categories. Considering the uniqueness of the standard membership to the standard category, the upper and lower limits of the difference similarity value calculated by the standard, the number of standards contained in the standard category and the number of standard categories as constraints, a mathematical model of the optimal standard classification is established.

[0034] The objective function of this mathematical model is: ① The sum of the absolute values ​​of the similarity between the standards within a standard classification is the largest. The expression is: (3) In the formula, S is the sum of the absolute values ​​of the total similarity within the standard categories; N is the number of standards to be classified; S is the number of standard categories; for The i-th standard in the n-th standard belongs to the i-th standard. The membership variable for each category takes a value of 1 if it belongs and a value of 0 if it does not belong. for The j-th standard in the n-th standard belongs to the j-th standard. The membership variable for each category takes a value of 1 if it belongs and a value of 0 if it does not belong. Let represent the similarity between standards i and j. The similarity between standards i and j is an index that measures the degree of similarity between the non-zero positions of the two standard topic word vectors, based on the standard topic word vectors representing the two standards. The formula for calculating the similarity between standards i and j can be described as the ratio of the sum of the two standard topic word vectors minus the absolute value of the error between the two standard topic word vectors to the sum of the two standard topic word vectors.

[0035] (4) In the formula, , Let be the k-th dimension component value of the standard topic word vectors i and j.

[0036] ② The sum of the absolute values ​​of the similarity between standard classifications is minimized, expressed as: (5) In the formula, This is the sum of the absolute values ​​of the total similarity between different standard classifications. for The i-th standard in the n-th standard belongs to the i-th standard. The membership variable for each category takes a value of 1 if it belongs and a value of 0 if it does not belong. For The jth standard belongs to the hth standard classification, and the membership variable is 1, and 0 otherwise.

[0037] The constraints of the mathematical model are as follows: ① Standard classification quantity constraint: the number of standard classifications should be within a certain range.

[0038] (7) Where, is the minimum value of the number of standard classifications; is the maximum value of the number of standard classifications.

[0039] ② Standard quantity constraint within standard classification: the number of standards contained in each standard classification should be within a certain range.

[0040] (8) Where, is the minimum value of the number of standards that the hth standard classification can contain; is the maximum value of the number of standards that the hth standard classification can contain.

[0041] ③ Standard total quantity constraint: the sum of the number of standards contained in the standard classification should be equal to the total number of standards.

[0042] (9) ④ Standard membership uniqueness constraint: each standard can only belong to one standard classification.

[0043] (10) ⑤ Difference proximity value constraint: the difference proximity between two standards should be within a certain range.

[0044] (11) Where, is the difference proximity between standards i and j; is the minimum value of the difference proximity; is the maximum value of the difference proximity.

[0045] (3) The above optimization model is solved by using the fluctuation foraging whale optimization algorithm to obtain the optimized classification result of the standards.

[0046] Based on the principles of simulation and standard whale optimization algorithm, according to the law of whales foraging with the fluctuation of sea water flow, a fluctuation foraging whale optimization algorithm is proposed, which is used to solve the standard classification optimization model. The main steps are as follows: ​1) Linear weighting method is used to convert the multi-objective optimization function of the optimization model into a comprehensive optimization objective function, as shown in formula (12). At the same time, the constraint condition is programmed, and the population of the whale optimization algorithm is initialized.

[0047] (12) In the formula, F is the value of the comprehensive objective function; , is the linear weighting coefficient.

[0048] 2) For each iteration calculation, the surrounding prey stage, bubble net attack stage and random search stage calculation formula in the standard whale algorithm are used to obtain the first population.

[0049] 3) Based on the first population, the fluctuation foraging operator is developed, and the fluctuation foraging operator is shown in formula (12). The population after executing the fluctuation foraging operator is the second population. The execution steps of the fluctuation foraging operator are as follows: ① Each individual represents a standard scheme belonging to different standard categories. According to formula (13), it is judged whether all individuals (individuals refer to each individual representing a standard scheme belonging to different standard categories) in the first population execute the fluctuation foraging operator. If formula (13) is satisfied, the following step ② is executed. Otherwise, the individual directly enters the second population.

[0050] (13) In the formula, is the maximum value of the comprehensive objective function value of all individuals in the population; is a random number between 0 and 1.

[0051] ② For the individual of the first population judged by formula (13), formula (14) is executed, and the individual after execution enters the second population.

[0052] (14) In the formula, is the individual obtained after executing formula (14); is the individual satisfying formula (13); is a random number satisfying 0 as mean and 1 as variance; , are the upper and lower control coefficients, respectively; is a constant; is the fluctuation period; t is the iteration number at this time; is the maximum iteration number.

[0053] 4) It is judged whether each individual in the second population satisfies the constraint condition. If yes, go to step 5); if not, adjust the individual and re-execute step 4).

[0054] 5) Determine whether the population after step 4) meets the termination condition, if yes, output the optimal solution; if not, repeat steps 2) to 4).

[0055] Based on the above optimization model and solving algorithm, the primary standard classification result can be obtained.

[0056] (4) Based on the above one level standard classification result, higher level standard classification is carried out.

[0057] The union of the standard subject heading vectors of the standard subject headings classified into one category after the last level classification is taken, and the values at the index positions corresponding to the original standard subject heading vectors are added to obtain the standard subject heading union vector as shown in (15).

[0058] (15) In the formula, is the standard subject heading union vector; is the subscript set corresponding to the standard set obtained after the last level classification, wherein represents the i-th standard of the standard set obtained after the last level classification.

[0059] The standard subject heading union vector is taken as a new standard subject heading vector, and steps (2) to (4) are repeated for classification until the level meeting the requirement is reached, and the classification is ended.

[0060] In another embodiment of the present scheme, the specific implementation process of S3 is as follows: ​For the generated multi-level classification results, the category records in SQLite are scanned layer by layer. First, for each category c in the first layer, all the original standard texts contained by the category c are extracted, all the subject words are summarized, and the word frequency and the word length are counted. The adaptive weighting formula weight = a norm_freq + b (1-norm_len) is used to calculate the comprehensive weight of each subject word, wherein norm_freq is the normalized word frequency of the word in the category c, norm_len is the normalized word length, and a and b are system preset coefficients. The subject word with the highest weight is selected as the target subject word of the category c and written into the field target_term. After all categories are completed, a tree index is constructed using the level and the category number. The root node is "petroleum field standard system", the child nodes of the root node are the first layer categories and their target_term in turn, and each first layer node is further connected to the corresponding category nodes and their target_term in the second layer until the Lth layer. At the same time, a visual JSON is generated. The node fields include level, class_id, target_term, std_count, and the child node list children. After being loaded by the front end, the JSON is automatically rendered into a collapsible tree graph. When a user clicks any node, the user can expand and view the list of all original standard texts under the category, thereby realizing a dynamic standard system directly mapped from the multi-level classification results and the target subject words.

[0061] (5) The adaptive weighting method is used to determine the level subject word of each standard class in each level.

[0062] For each standard subject word in the set of standard subject words, the word length of the standard subject word and the frequency value of the standard subject word appearing in all standard full texts belonging to the same category are obtained, thereby obtaining two attribute values corresponding to the standard subject word. The word length and the word frequency of the standard subject word are normalized, wherein the word length is normalized by using the negative index formula (16).

[0063] (16) In the formula, is the word length attribute value of the standard subject word; is the normalized value of the word length attribute of the standard subject word.

[0064] The absolute value of the difference between the two attribute values is calculated. All the absolute values of the calculated differences are compared with a threshold value. If all the differences are less than the threshold value, the maximum value of the two attribute values is selected as the weight of the standard subject word. Otherwise, the sum of the product of the weight coefficient and each attribute value of the standard subject word is used as the weight of the standard subject word. The adaptive weighting formula is as follows: (17) In the formula, is the weight of the standard subject word; , is a standard subject word weight coefficient; is an absolute value of the difference; is a comparison threshold value; is a maximum value function for obtaining the maximum value of the values in the parentheses; is a standard subject word length and word frequency normalization value, respectively.

[0065] (6) Based on the above 5 steps, a hierarchical classification map is drawn, a standard system framework is obtained, and the standards are automatically classified.

[0066] Further, the process of constructing the standard subject word vector set is specifically as follows: Extract the subject words of each standard through natural language processing technology, and form a subject word set by combining the subject words of all standards; In the subject word set, each original standard is represented as an n-dimensional zero vector, the target number k of the first subject word contained in the subject word set in each original standard is determined, and the kth dimension of the n-dimensional zero vector of each original standard is assigned a value of 1 to generate an optimized vector corresponding to each original standard; Integrate the optimized vectors corresponding to all original standards to generate a standard subject word vector set.

[0067] The beneficial effects of the above further scheme are: After extracting all the subject words, each standard is directly mapped to a binary n-dimensional vector, and only the positions corresponding to the k subject words appearing in the standard are set to 1, so that large-scale text vectorization can be completed within milliseconds, significantly reducing storage and computing overhead; The mechanism of zero vector setting to 1 naturally retains the binary information of "appearance / non-appearance", avoids the subjective disturbance introduced by weight parameters, enables the subsequent solver to quickly converge in the same semantic space, and completely eliminates the deviation caused by manual scoring; The generated standard subject word vector set can be directly used as the input of any hierarchical clustering or mathematical programming model, realizing one-time construction and multiple reuse, and greatly improving the agility of standard system construction and dynamic expansion.

[0068] Further, the user requirements include: The category number requirement range, the upper and lower limits of the number of standards in each category, and the upper and lower limits of the difference proximity.

[0069] It should be noted that the category number requirement range refers to the value interval of the total number of standard classifications expected to be generated by the user. It prevents the algorithm from outputting extreme results such as too fine (e.g., only one standard per category) or too coarse (e.g., all standards are classified into one category). For example, the user can set the category number range to [3, 8], indicating that the final standard system should contain 3-8 first-level categories.

[0070] The upper and lower bounds of the number of standards in each category refer to the minimum and maximum number of standards allowed to be included in each category. Avoiding too many standards in a certain category (such as 2000 standards in one category) or too few standards (such as only 1 standard in a certain category leading to redundancy). For example, set the number of standards in each category to [10, 500], then the number of standards in any category cannot be less than 10 or more than 500.

[0071] The upper and lower bounds of the difference proximity refer to the legal value range of the similarity between two standards (difference proximity), which is used to control the strictness of the classification. By setting the similarity range (such as [0.3, 0.8]), the user forces the standards in the same category to be sufficiently similar (≥0.3), while avoiding excessive similarity (such as ≥0.9 leading to duplicate standards), or too small differences between categories (such as ≤0.8 to prevent different categories from being merged).

[0072] Further, the process of determining the target subject heading corresponding to each level in the multi-level classification result is specifically: According to the attribute information of the subject headings of all standards corresponding to each level in the multi-level classification result, determine the target subject heading corresponding to each level; Wherein, the attribute information includes: the word length and the word frequency of the subject heading.

[0073] Further, it further comprises: Based on the standard system, the whole life cycle management of the oil management system is carried out.

[0074] Example 1, taking the generation of urban gas industry standard system as an example. Based on 20 urban gas industry standards (C1, C2, C3, C4, C5, C6, C7, C8, C9, C10, C11, C12, C13, C14, C15, C16, C17, C18, C19, C20), 35 subject headings are extracted from 20 standards by using natural language processing technology, and a standard subject heading set is constructed. The standard subject headings specifically include: urban gas, odorization system, engineering construction, facility replacement, process operation, facility operation, pipeline integrity management, oil and gas pipeline construction, oil and gas pipeline electrical equipment maintenance regulations, pipeline corrosion control engineering, safety emergency, pipeline detection and evaluation, oil and gas pipeline safety design and protection, pipeline high consequence area, engineering design, pipeline leak detection, production operation management requirements, planning, equipment management, oil and gas pipeline instrument detection and automation control, system operation safety, oil and gas pipeline monitoring and data acquisition system, supply system design, quality measurement, system anti-terrorism prevention, production storage and transportation, transportation system design, digitalization, oil and gas transportation pipeline engineering, gas lightning protection, construction quality acceptance criteria, process program, pre-commissioning safety inspection of transmission and distribution engineering, repair equipment and tool configuration specification, instrument engineering. The standard system needs to be constructed into a three-level system, that is, two levels of classification need to be completed.

[0075] The standard terms are mapped one-to-one with the standard topic word vectors. Based on the standard topic word vectors, a standard optimal classification mathematical model is established, and the wave-foraging whale optimization algorithm of step (3) of this invention is used to solve the standard optimal classification mathematical model. The preliminary classification results are shown in Table 1.

[0076] Table 1 Primary classification results The standard keywords in each category of the primary classification are numerically summed using formula (15) in step (5) to obtain a new standard topic word vector. Steps (2)-(4) are repeated to obtain the advanced classification results as shown in Table 2.

[0077] Table 2 Advanced Classification Results According to step (6) of the present invention, the attributes of the keywords are calculated, and the standard topic word weights are obtained using formulas (16) and (17). Figure 3 As shown.

[0078] like Figure 4 As shown, based on Figure 3 The weights of the obtained standard keywords were used to select "urban gas" as the high-level keyword and "engineering construction, process operation, pipeline integrity management and safety emergency" as the primary-level keywords.

[0079] In the above embodiments, although the steps are numbered S1, S2, etc., they are only specific embodiments given by the present invention. Those skilled in the art can adjust the execution order of S1, S2, etc. according to the actual situation, which is also within the protection scope of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0080] This invention also provides a standard system generation system based on standard optimization classification, the specific technical solution of which is as follows: The extraction module is used to: extract the subject terms of each standard in the effective standard collection in the petroleum field, and construct a set of standard subject term vectors by combining them with the original standard text; The processing module is used to: based on the standard topic word vector set, user requirements, and the number of levels given by the user, repeatedly execute the solver a number of times corresponding to the number of levels to obtain multi-level classification results. The solver is used to determine the union vector of topic words of all standards under the category to which each standard belongs and generate a classification table. The generation module is used to: determine the target topic words corresponding to each level in the multi-level classification results, and construct the standard system corresponding to the original standard text through the multi-level classification results and target topic words.

[0081] It should be noted that the beneficial effects of the system for generating a standard system based on standard optimization classification provided by the above embodiment are the same as those of the method for generating a standard system based on standard optimization classification, which will not be repeated here. In addition, when the system provided by the above embodiment implements its functions, only the division of the above functional modules is exemplified, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the system is divided into different functional modules according to actual conditions to complete all or part of the above described functions. In addition, the system and method embodiments provided by the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0082] As shown in Figure 5 The electronic device 300 provided by the embodiment of the present application, the electronic device 300 includes a processor 320, the processor 320 is coupled with a memory 310, and the memory 310 stores at least one computer program 330. The at least one computer program 330 is loaded and executed by the processor 320, so that the electronic device 300 implements any of the above methods, specifically: The electronic device 300 can have great differences due to different configurations or performances, and can include one or more processors 320 (Central Processing Units, CPU) and one or more memories 310, wherein the one or more memories 310 store at least one computer program 330, and the at least one computer program 330 is loaded and executed by the one or more processors 320, so that the electronic device 300 implements the method for generating a standard system based on standard optimization classification provided by the above embodiment. Of course, the electronic device 300 can also have a wired or wireless network interface, a keyboard, an input and output interface and other components for realizing the functions of the device, so as to perform input and output. The electronic device 300 can also include other components for realizing the functions of the device, which will not be repeated here.

[0083] The computer readable storage medium provided by the embodiment of the present application stores at least one computer program, and the at least one computer program is loaded and executed by the processor, so that the computer implements any of the above methods.

[0084] Optionally, the computer readable storage medium can be a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), a magnetic tape, a floppy disk and an optical data storage device, etc.

[0085] In an example embodiment, a computer program product or computer program including computer instructions stored in a computer readable storage medium is also provided. A processor of an electronic device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to cause the electronic device to perform any of the above methods.

[0086] It should be noted that the terms "first", "second" and "third" in the specification and claims of the present application are used to distinguish similar objects, and do not represent a specific order or a chronological order. The order of use of similar objects can be interchanged, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described.

[0087] Those skilled in the art understand that the present application can be implemented as a system, a method or a computer program product, and thus the present disclosure can be embodied in the form of a complete hardware, a complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, which is generally referred to as "circuitry", "module" or "system". In addition, in some embodiments, the present application can also be implemented in the form of a computer program product in one or more computer readable media, which contains computer readable program code.

[0088] Any combination of one or more computer readable medium can be employed. The computer readable medium can be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium can be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination of the above. More specific examples (a non-exhaustive list) of the computer readable storage medium include an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus or device.

[0089] Although the embodiments of the present application have been shown and described above, it should be understood that the above-described embodiments are exemplary, and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-described embodiments within the scope of the present application.

Claims

1. A method for generating a standard system based on standard optimization classification, characterized in that, include: Extract the subject terms of each standard from the effective standard collection in the petroleum field, and construct a set of standard subject term vectors by combining them with the original standard text; Based on the standard topic word vector set, user requirements, and the number of levels given by the user, the solver repeatedly executes the function the number of times corresponding to the number of levels to obtain multi-level classification results. The solver is used to determine the union vector of topic words of all standards under the category to which each standard belongs and to generate a classification table. Determine the target keywords corresponding to each level in the multi-level classification results, and construct a standard system corresponding to the original standard text based on the multi-level classification results and target keywords.

2. The method for generating a standard system based on standard optimization classification according to claim 1, characterized in that, The process of constructing a standard set of topic word vectors is as follows: The subject terms for each standard are extracted using natural language processing technology, and all the subject terms for the standards are combined into a subject term set. In the keyword set, each original standard is represented as an n-dimensional zero vector. The number of targets k of the first keyword in the keyword set contained in each original standard is determined. The k-th dimension of the n-dimensional zero vector of each original standard is assigned a value of 1, and the optimization vector corresponding to each original standard is generated. By integrating the optimized vectors corresponding to all original standards, a set of standard topic word vectors is generated.

3. The method for generating a standard system based on standard optimization classification according to claim 1, characterized in that, User requirements include: The required range of the number of categories, the upper and lower bounds of the standard number for each category, and the upper and lower bounds of the similarity difference.

4. The method for generating a standard system based on standard optimization classification according to claim 1, characterized in that, The process of determining the target topic words corresponding to each level in the multi-level classification results is as follows: Based on the attribute information of all standard topic words corresponding to each level in the multi-level classification results, determine the target topic words corresponding to each level; The attribute information includes the word length and frequency of the subject terms.

5. The method for generating a standard system based on standard optimization classification according to claim 1, characterized in that, Also includes: The entire lifecycle management of the petroleum management system is based on a standard system.

6. A standard system generation system based on standard optimization classification, characterized in that, include: The extraction module is used to: extract the subject terms of each standard in the effective standard collection in the petroleum field, and construct a set of standard subject term vectors by combining them with the original standard text; The processing module is used to: based on the standard topic word vector set, user requirements, and the number of levels given by the user, repeatedly execute the solver a number of times corresponding to the number of levels to obtain multi-level classification results. The solver is used to determine the union vector of topic words of all standards under the category to which each standard belongs and generate a classification table. The generation module is used to: determine the target topic words corresponding to each level in the multi-level classification results, and construct the standard system corresponding to the original standard text through the multi-level classification results and target topic words.

7. A standard system generation system based on standard optimization classification according to claim 6, characterized in that, The process of constructing a standard set of topic word vectors is as follows: The subject terms for each standard are extracted using natural language processing technology, and all the subject terms for the standards are combined into a subject term set. In the keyword set, each original standard is represented as an n-dimensional zero vector. The number of targets k of the first keyword in the keyword set contained in each original standard is determined. The k-th dimension of the n-dimensional zero vector of each original standard is assigned a value of 1, and the optimization vector corresponding to each original standard is generated. By integrating the optimized vectors corresponding to all original standards, a set of standard topic word vectors is generated.

8. A standard system generation system based on standard optimization classification according to claim 6, characterized in that, User requirements include: The required range of the number of categories, the upper and lower bounds of the standard number for each category, and the upper and lower bounds of the similarity difference.

9. An electronic device, characterized in that, The electronic device includes a processor coupled to a memory, the memory storing at least one computer program, which is loaded and executed by the processor to enable the electronic device to perform the method as claimed in any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one computer program, which is loaded and executed by a processor to enable the computer to implement the method as claimed in any one of claims 1 to 5.

Images