Scientific data set naming specification check automatic updating model training method and system

Abstract

The invention relates to the technical field of data processing, in particular to a scientific data set naming specification check automatic updating model training method and system.The method comprises the steps that naming rules are formulated according to naming requirements, and a naming specification check task tree is established; extracting auditing suggestions attached to each named instance of the scientific data set, corresponding matched information carriers and naming features, generating training samples based on the matched information carriers and naming features of each named instance, and classifying the normativity of the auditing suggestions to obtain the classified training samples; and analyzing the classified training samples based on the naming specification checking task tree to obtain a scientific data set naming specification checking model. According to the method, the naming of the scientific data set can be automatically checked, updated and optimized, so that the error probability caused by manual operation is reduced, the naming consistency and accuracy of the scientific data set are improved, and the scientific data set is ensured to have good readability and usability.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to a scientific data set naming specification checking and automatic updating model training method and system. Background Art

[0002] Scientific data mainly include data generated through basic research, applied research, experimental development, etc. in the fields of natural sciences, engineering sciences, etc., as well as original data and its derivative data obtained through observation, monitoring, investigation, inspection and testing and used in scientific research activities. Scientific data sets are collections of scientific data organized and managed based on certain rules and standards. With the increasing number of sources of scientific data submission and the increasingly diverse requirements for scientific data set management, the problem of non-standard data naming is becoming increasingly serious. Researchers often use custom naming methods when submitting and sharing scientific data. This inconsistent naming not only increases the difficulty of data understanding and retrieval, but also creates obstacles to data reuse and integration. In addition, non-standard data naming also brings challenges to the acceptance process of scientific data, which adds additional workload to data receiving institutions when checking data against research funding project books, task books and other documents.

[0003] Although existing data management tools and methods have made some progress in data processing and analysis, they are still insufficient in checking and automatically updating data naming specifications. Current technologies mainly rely on manual intervention and lack systematic and standardized solutions. The functions of automatic naming specification checking and updating are not perfect, resulting in inefficient data management and prone to errors. Summary of the invention

[0004] The present invention aims to overcome at least one defect (shortcoming) of the above-mentioned prior art, and provides a method and system for automatically updating a model training for checking scientific dataset naming specifications, so as to systematically handle the problem of scientific dataset naming specifications.

[0005] According to a first aspect of the present application, a method for automatically updating a model training method for checking naming conventions of scientific datasets is provided, comprising:

[0006] Formulate naming rules according to naming requirements and establish a naming specification checking task tree, wherein the naming requirements include structured naming requirements and unstructured naming requirements, and the naming rules include structured naming rules and unstructured naming rules;

[0007] Extracting the review opinions, corresponding supporting information carriers, and naming features attached to each naming instance of the scientific data set, generating training samples based on the review opinions, the supporting information carriers, and the naming features of each naming instance, and classifying them according to the standardization of the review opinions to obtain classified training samples;

[0008] The classified training samples are analyzed based on the naming convention checking task tree to obtain the scientific data set naming convention checking model.

[0009] It is understandable that by formulating naming rules, building a naming specification checking task tree based on the naming rules, and using training samples to train the model, the model can automatically check and optimize the naming of scientific data sets. This can reduce the probability of errors caused by human operations, improve the consistency and accuracy of scientific data set naming, and ensure that scientific data sets have good readability and ease of use.

[0010] Optionally, the naming rules are formulated according to the naming requirements to establish a naming specification checking task tree, wherein the naming requirements include structured naming requirements and unstructured naming requirements, and the naming rules include structured naming rules and unstructured naming rules, specifically including:

[0011] Establishing a general naming template based on the structured naming requirements;

[0012] Build domain knowledge models;

[0013] Generate a configurable naming template through the general naming template and the domain knowledge model;

[0014] Generate a structured naming rule based on the configured naming template, wherein the structured naming rule includes one or more of naming field characteristics, field order, naming instance hierarchy constraints, version control, and rule weight;

[0015] Extracting semantic features of the unstructured naming requirements and generating naming example reference rules;

[0016] Based on the naming instance reference rule, construct an unstructured naming rule, wherein the unstructured naming rule includes a semantic consistency check and / or a text information integrity check;

[0017] The rule checking items involved are extracted according to the content of the naming rule, the priorities of the rule checking items are extracted according to the rule weights of the naming rule, and a naming specification checking task tree is hierarchically constructed according to the rule checking items and the priorities.

[0018] It is understandable that by obtaining naming rules based on naming requirements and establishing a naming specification checking task tree through the naming rules, each field of each naming instance can be accurately checked to achieve the effect of checking without omissions and improve the accuracy of the model checking.

[0019] Optionally, extracting the review opinions, corresponding supporting information carriers and naming features attached to each naming instance of the scientific data set, generating training samples based on the review opinions, the supporting information carriers and the naming features of each naming instance, and classifying the standardization of the training samples according to the review opinions to obtain the classified training samples specifically includes:

[0020] The training samples are divided into positive samples and negative samples according to the audit opinions, and based on the relationship between the audit opinions and the naming rules, the standard type of each training sample is determined, and the training samples are normatively classified according to the standard types to obtain training samples of various standard types.

[0021] It is understandable that processing the training samples and distinguishing between positive and negative samples can make the model highly targeted when learning samples, learn standardized naming information for positive samples, and avoid non-standard problems for negative samples, which can improve the accuracy and generalization ability of the model inspection.

[0022] Optionally, analyzing the classified training samples based on the naming convention check task tree to obtain the scientific dataset naming convention check model specifically includes:

[0023] Forming a naming feature set based on the naming features in the various training samples, and forming a matching information feature set based on the matching information carriers in the various training samples;

[0024] According to the semantic features of the supporting information feature set, the supporting information feature set is sub-classified to generate multiple supporting information feature subsets, and a naming specification check item set is formed according to the supporting information feature subsets;

[0025] Applying an association rule mining algorithm to construct an association rule matrix of the naming feature set and the matching information feature set;

[0026] Optimizing the naming convention check task tree based on the naming convention check item set and the association rule matrix;

[0027] The training samples are analyzed based on the optimized naming convention checking task tree to obtain a scientific data set naming convention checking model.

[0028] It is understandable that by optimizing the naming specification inspection task tree, it is possible to reduce redundant inspection paths, place highly associated tree nodes in the same path, and simplify the task execution process, which can improve the overall inspection efficiency and ensure the logical consistency of the inspection process.

[0029] Optionally, the scientific dataset naming standard checking model further includes:

[0030] Analyzing the positive samples using the scientific dataset naming convention checking model, identifying misclassified positive samples that are mistaken for naming conventions in the positive samples, and extracting naming rules and association rule matrices associated with the misclassified positive samples;

[0031] The scientific dataset naming convention checking model is further optimized by optimizing the rule weights in the naming rules of the misclassified positive samples and eliminating irrelevant naming rules.

[0032] It is understandable that by identifying misclassified positive samples that are mistakenly identified as naming specifications in the positive samples and optimizing the model, the error rate of the inspection work can be reduced and the accuracy of the model inspection can be improved.

[0033] Optionally, the scientific dataset naming standard checking model further includes:

[0034] Comparing the naming feature distribution differences between the positive sample and the negative sample, and obtaining all naming rules associated with the naming feature distribution differences;

[0035] Sorting the feature frequencies of the naming feature distribution differences to obtain a frequency order, and constructing a negative sample naming feature set according to the naming specification check task tree nodes corresponding to the naming rules associated with the naming feature distribution differences and the frequency order;

[0036] Analyze the negative samples using the scientific dataset naming standard check model, identify misclassified negative samples that are mistakenly considered to be non-standardly named, and extract the naming rules and association rule matrix associated with the misclassified negative samples;

[0037] The naming rules and association rule matrix associated with the misclassified negative samples are compared with the negative sample naming feature set to generate negative sample weighted features, and the scientific data set naming specification checking model is further optimized according to the negative sample weighted features.

[0038] It is understandable that by comparing the features of the positive samples and the negative samples, identifying the misclassified negative samples in the negative samples that are mistakenly identified as having non-standard naming conventions, and optimizing the model, the misjudgment rate of the inspection work can be reduced and the accuracy of the model inspection can be improved.

[0039] Optionally, the method further comprises:

[0040] Receiving inputs of different types of specific naming requirements, identifying specific naming rules based on the specific naming requirements, and generating a specific naming rule checking task tree;

[0041] Collecting specific named instances from the scientific data set, extracting supporting information carriers and naming features of the specific named instances as specific training samples;

[0042] The specific training samples are analyzed and trained based on the specific naming rule checking task tree to obtain a specific naming requirement recognition model.

[0043] It is understandable that the specific naming requirement recognition model is constructed and trained to adapt to the specific naming requirements in actual situations and improve the usability of the model.

[0044] Optionally, the method further comprises:

[0045] Extracting the naming features and supporting information carriers of the scientific data set to be evaluated, and inputting the naming features and the supporting information carriers into a specific naming requirement recognition model to generate specific naming suggestions;

[0046] Eliminating the part of the scientific dataset to be evaluated that adopts the specific naming convention according to the specific naming suggestion, obtaining a changed naming feature and a supporting information carrier, and then inputting the changed naming feature and the supporting information carrier into the scientific dataset naming convention check model to obtain a naming convention check result;

[0047] According to the naming specification check result, a naming update is performed on the scientific data set to be evaluated and an update log is recorded.

[0048] It is understandable that the naming of the scientific dataset is checked using a dual model of a specific naming requirement identification model and a scientific dataset naming specification check model to enhance the accuracy of the check, and the naming is updated and logged so that the naming process has a certain traceability. When problems are found later, they can be traced back according to the logs and solved.

[0049] Optionally, the method further comprises:

[0050] Obtain user feedback on the naming convention check result and the naming update, readjust the naming rules in the scientific dataset naming convention check model according to the feedback, thereby adjusting the naming convention check task tree, reprocessing the misclassified positive samples or the misclassified negative samples, and re-analyzing the training samples according to the adjusted naming convention check task tree to complete the optimization of the scientific dataset naming convention check model.

[0051] It is understandable that optimizing the model based on user feedback can reduce system errors in inspection work due to system mislearning, and timely human intervention can improve system vulnerabilities.

[0052] According to a second aspect of the present application, a scientific dataset naming specification checking and automatic updating model training system is provided, comprising:

[0053] A construction module is used to formulate naming rules according to naming requirements and establish a naming specification checking task tree, wherein the naming requirements include structured naming requirements and unstructured naming requirements, and the naming rules include structured naming rules and unstructured naming rules;

[0054] a sample processing module, configured to extract the review opinions, the corresponding supporting information carriers and the naming features attached to each naming instance of the scientific data set, generate training samples based on the review opinions, the supporting information carriers and the naming features of each naming instance, and classify the standardization of the training samples according to the review opinions to obtain the classified training samples;

[0055] The training module analyzes the classified training samples based on the naming convention checking task tree to obtain a scientific data set naming convention checking model.

[0056] Based on any of the above aspects, the embodiment of the present application provides a scientific data set naming standard check automatic update model training method and system, obtains the naming rules through the naming requirements, and then establishes a naming standard check tree according to the naming rules, analyzes and learns the training samples according to the naming standard check tree, obtains the scientific data set naming standard check model, and optimizes the model. Compared with the prior art, the beneficial effects of the present invention are:

[0057] (1) The model can automatically check and optimize the naming of scientific datasets. Automated naming standard checking reduces the probability of errors caused by human operations, improves the consistency and accuracy of scientific dataset naming, and accelerates the naming and updating process, thereby ensuring that scientific datasets have good readability and ease of use.

[0058] (2) The present invention supports the generation of configurable naming templates according to different industries and application scenarios by presetting general naming templates and domain knowledge models. Through user feedback and misclassified sample analysis, the model can dynamically adjust the naming specifications and inspection rules to keep up to date with actual needs. This makes the model not only highly adaptable and extensible, but also able to meet the changes in different fields and specific naming needs.

[0059] (3) By optimizing the performance of the naming standard checking model, the present invention provides users with accurate naming standard checking services and improves user experience. At the same time, functions such as compliance checking and audit records enhance the transparency and traceability of data management and ensure that data processing complies with relevant laws and standards. The real-time update and feedback mechanism enables users to obtain the latest naming standard checking results and suggestions in a timely manner, thereby improving the standardization and security of data management. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0061] Figure 1 A flow chart of the method for automatically updating model training by checking scientific dataset naming conventions provided in this embodiment.

[0062] Figure 2 The present embodiment provides a process for formulating naming rules according to naming requirements and establishing a naming specification checking task tree flowchart.

[0063] Figure 3 A flow chart of a scientific data set naming standard checking model provided in this embodiment.

[0064] Figure 4 A flow chart for processing misclassified positive samples provided in this embodiment.

[0065] Figure 5 A flow chart for processing misclassified negative samples provided in this embodiment.

[0066] Figure 6 A flow chart of a specific naming requirement identification model provided for this embodiment.

[0067] Figure 7 A flow chart of the method for evaluating the naming of scientific datasets using a dual model provided in this embodiment.

[0068] Figure 8 Module diagram of the scientific dataset naming specification check and automatic update model training system provided in this embodiment. DETAILED DESCRIPTION

[0069] The drawings of this application are only used for illustrative purposes and should not be construed as limiting the present application. In order to better illustrate the following embodiments, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product; it is understandable to those skilled in the art that some well-known structures and their descriptions in the drawings may be omitted.

[0070] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0071] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0072] This embodiment provides a technical solution that can solve the above-mentioned problem. The specific implementation methods of this application are described in detail below in conjunction with the accompanying drawings.

[0073] For example, Figure 1 As shown, the flowchart of the scientific dataset naming standard check and automatic update model training method provided in this embodiment includes:

[0074] S110, formulating naming rules according to naming requirements and establishing a naming specification checking task tree, wherein the naming requirements include structured naming requirements and unstructured naming requirements, and the naming rules include structured naming rules and unstructured naming rules;

[0075] The step S110 can be subdivided into the following steps: Figure 2 The present embodiment provides a method for formulating naming rules according to naming requirements and establishing a naming standard checking task tree flow chart, including:

[0076] S111, establishing a general naming template based on the structured naming requirement;

[0077] S112. Build domain knowledge model;

[0078] S113, generating a configurable naming template through the general naming template and the domain knowledge model;

[0079] S114, generating a structured naming rule based on the configured naming template, wherein the structured naming rule includes one or more of naming field characteristics, field order, naming instance hierarchy constraints, version control, and rule weight;

[0080] Understandably, assuming that a scientific data receiving organization manages scientific data in a certain field with a unified naming standard, the organization requires that the name of each scientific data set must contain specific fields and conform to a specific format. In addition, the naming of scientific data sets must follow certain file structures and version control rules.

[0081] In this embodiment, the structured naming requirement may be in the following form:

[0082] (1) Data folder structure: Scientific data sets are stored in different folders. The folder names need to reflect the project name and data type. For example: project name / data type / data name / data version / …

[0083] (2) Dataset file structure: In the above folder, the name of the dataset file needs to include the project type, discipline code, project abbreviation, data type, date and version number.

[0084] For example: GDSF_F01_PRJ1_RAW_20240814_V1.CSV

[0085] (3) Naming hierarchy: Define the naming hierarchy to ensure that the order of the fields is not disrupted. For example, the naming hierarchy is:

[0086] First level field: Project abbreviation

[0087] Secondary Field: Data Type

[0088] Third level field: Date

[0089] Level 4 field: version number

[0090] (4) Field order: specifies the order in which the fields appear in the name. In the above example, the field order is: {project abbreviation}_{data type}_{date}_{version number}

[0091] (5) Version number rules: The dataset version number starts with "V" followed by a number. The number part indicates the iteration of the version. For example: V1, V2, V3...

[0092] (6) Version upgrade rules: After each update of the data set content, the version number and timestamp must be updated. For example: from "GDSF_F01_PRJ1_RAW_20240814_V1.csv"

[0093] Updated to "GDSF_F01_PRJ1_RAW_20240816_V2.csv"

[0094] Exemplarily, establishing a general naming template based on structured naming requirements includes:

[0095] {ProjectCode}_{DataType}_{YYYYMMDD}_{Version}

[0096] ProjectCode: The unique identifier of the project, such as "PRJ1", "PRJ2"

[0097] DataType: data type, such as "RAW", "PROCESSED"

[0098] YYYYMMDD: date the data was collected or processed

[0099] Version: version number, such as "V1", "V2"

[0100] Understandably, the domain knowledge model provides suggested naming schemes by learning historical data naming patterns and domain-specific requirements. For example, for a meteorological data project, the model may suggest:

[0101] {LocationCode}_{DataType}_{YYYYMMDD}_{HHMM}_{Version}

[0102] Among them, LocationCode: weather station code

[0103] DataType: data type, such as "temperature", "humidity"

[0104] HHMM: Data recording time

[0105] Version: version number, such as "V1", "V2"

[0106] Configurable naming templates generated based on common naming templates and domain knowledge models, for example:

[0107] {ProjectCode}_{LocationCode}_{DataType}_{YYYYMMDD}_{HHMM}_{Version}

[0108] Construct naming instance hierarchical constraints based on folder structured naming requirements, such as:

[0109] First-level folder: ProjectName; example: ClimateProject /

[0110] Second level folder: DataType; example: ClimateProject / TemperatureData /

[0111] The third folder level: date, including year, month and day;

[0112] Example: ClimateProject / TemperatureData / 2024 / 08 / 14 /

[0113] The fourth layer of files: version;

[0114] Example: ClimateProject / TemperatureData / 2024 / 08 / 14 / V1 /

[0115] Fifth-level file name: follow the field order;

[0116] Example: ClimateProject_TemperatureData_20240814_V1.csv

[0117] S115, extracting semantic features of the unstructured naming requirements and generating naming example reference rules;

[0118] S116. Constructing an unstructured naming rule based on the naming instance reference rule, wherein the unstructured naming rule includes a semantic consistency check and / or a text information integrity check;

[0119] In this embodiment, the unstructured rules are obtained based on unstructured naming requirements, which are generally proposed by business personnel, project managers, etc. who are not familiar with scientific data management specifications. Examples of unstructured naming requirement elements are as follows:

[0120] (1) Naming field content: It should include fields such as project number, data category, data description, date, etc., and may include detailed information such as specific project parameters or environmental factors. Data naming should correspond to the relevant content of the data quality report.

[0121] (2) Folder naming hierarchy: Naming should be based on the hierarchical structure and logical relationship of the supporting information carrier. For example, the folder naming hierarchy should be able to reflect

[0122] (3) Associated information carriers: Ensure the consistency between the data content contained in the naming rules and the corresponding content in the supporting information carriers (such as scientific and technological project task books, data management plans, etc.).

[0123] Extract the semantic features of unstructured naming requirements and generate naming example reference rules, including:

[0124] (1) Naming field content characteristics

[0125] Project number: Each project has a unique project number, which must be reflected in the name.

[0126] Data category: Clarify which category the data belongs to, such as observational data, experimental data, etc.

[0127] Data description: A brief description of the data content for easy identification.

[0128] Date: The date when the data is generated or collected, in the format of YYYY-MM-DD.

[0129] Project parameters / environmental factors: If specific parameters or conditions are involved, this information needs to be included in the name.

[0130] Further, a search prompt word is generated according to the content characteristics of the named field, and the corresponding corpus is obtained from the supporting information carrier.

[0131] (2) Folder naming hierarchy characteristics

[0132] Extract the naming hierarchical features in the supporting information carriers such as the project scientific and technological report, data management plan, data submission plan, data quality report, etc. in the folder. For example:

[0133] For project science and technology reports that are indirectly related to scientific data, extract the semantic features and structure of the chapters and paragraphs of the project science and technology reports, extract the names or descriptions of the data resources associated with each chapter and paragraph, and generate the data resource structure table of the scientific data set;

[0134] A data submission plan that is directly related to scientific data, retrieves chapters that are strongly related to the data resource description structure, extracts the chapter paragraph structure and paragraph semantics to obtain the number, type, and feature distribution of related data resources, and extracts the semantic information of each paragraph sentence based on dependency syntax to obtain detailed description features of data resources.

[0135] Specifically, paragraph structure feature extraction, for example, there is a chapter in the project's scientific and technological report titled "4. Analysis of Climate Monitoring Situation", which contains two secondary titles "4.1 Precipitation Statistics" and "4.2 Temperature Statistics". Through natural language processing, the paragraph summary under "4.1 Precipitation Statistics" is extracted: "This paragraph mainly describes the observation frequency, measurement method and data range of precipitation data". The key information of the paragraph is extracted based on the abstract keywords, and the results are: "Data is collected daily from January 2010 to December 2020", "Measured using automatic rain gauges with an accuracy of 0.1 mm", and "Data covers major cities across the country, with a focus on coastal areas". The paragraph structure feature extraction results are obtained based on the key information: Chapter number 4.1; Paragraph level; Level 2; Description level: Data resources; Data range: 2010-2020; Geographic coverage: nationwide, coastal areas.

[0136] In paragraph semantic feature extraction, semantic extraction prompt words should be preset according to the domain scientific dataset management specifications or sharing requirements. For example, for temperature data, the time range, data source, observation tool, spatial range, etc. should be extracted.

[0137] Based on the above features and the detailed description of the data resources extracted by dependency syntax, the redundant common features are removed and the difference features are retained to obtain the folder structure:

[0138] First level: project number, using the project number in the data management plan or scientific and technological project task book

[0139] Example: P001 (project number is P001)

[0140] The second level: data type, based on the extracted paragraph structure and semantic features, determines the data type

[0141] Examples: precipitation data, temperature data

[0142] The third level: data description, key description information extracted based on paragraph semantics and dependency syntax

[0143] Example:

[0144] Precipitation data: Daily observations_Automatic rain gauge_0.1 mm

[0145] Temperature data: daily maximum and minimum temperatures_standard thermometer

[0146] The fourth level: time range, using the time range in paragraph structure feature extraction

[0147] Example: 2010-2020

[0148] The fifth level: geographical coverage, geographical coverage information extracted based on paragraph structure characteristics

[0149] Example: Major cities in China_Coastal areas

[0150] Level 6: Version control, add version information for different versions of data

[0151] Example: V1.0

[0152] S117, extracting the rule check items involved according to the content of the naming rule, extracting the priorities of the rule check items according to the rule weights of the naming rule, and constructing a naming standard check task tree in layers according to the rule check items and the priorities.

[0153] For example, a structured naming rule includes rule checking items: checking whether the folder structure complies with the naming rules, checking whether the file name contains all required fields, checking whether the field order is correct, checking whether the field format and content comply with the rules, and checking whether the version control is correct.

[0154] Establish a naming convention check task tree based on the rule check items. Task tree structure root node 1: folder structure check, sub-node 1.1 check project name, sub-node 1.2 check data type, sub-node 1.3 check date folder, sub-node 1.3 check version folder; root node 2 file name field check, sub-node 2.1 check file name field integrity, sub-node 2.2 check field format and content; root node 3 field order and format check; root node 4 version control check.

[0155] S120, extracting the review opinions, corresponding supporting information carriers, and naming features attached to each naming instance of the scientific data set, generating training samples based on the review opinions, the supporting information carriers, and the naming features of each naming instance, and classifying the standardization of the training samples according to the review opinions to obtain classified training samples;

[0156] The step S120 specifically includes:

[0157] The training samples are divided into positive samples and negative samples according to the audit opinions, and based on the relationship between the audit opinions and the naming rules, the standard type of each training sample is determined, and the training samples are normatively classified according to the standard types to obtain training samples of various standard types.

[0158] It is understandable that the naming characteristics of the naming instance refer to the naming field characteristics of the naming instance, including field name, field type, and field order. The naming characteristics can be represented by a naming structure table;

[0159] The naming structure table describes the fields and levels of the naming instance. For example, if there is a naming instance "P001_Precipitation Data

[0160] _2010-2020_Daily Observation_Automatic Rain Gauge_V1.0.csv", and get the naming structure table according to the naming instance:

[0161]

[0162]

[0163] The supporting information carrier refers to the documents related to the named instance, such as the science and technology project task book, data management plan, data submission plan, data quality report, etc. For example, the science and technology project task book specifies the project number, data type, and time range for collecting data.

[0164] The audit opinion is an audit record of the named instance, including the reasons for passing or failing. For example, the audit opinion may include "missing time range information" or "version number does not comply with the naming specification".

[0165] For each named instance, natural language processing technology is used to identify naming features, including the hierarchical structure of the named fields, field content, and its semantic features. These fields include but are not limited to project number, data type, time range, data description, version control, etc. The hierarchical order, content format, and naming constraints of each field are obtained from the naming structure table. By analyzing the naming features in the named instances, the semantic information of each field and the semantic association between the named instances and the supporting information carriers are extracted. For each named instance, its related supporting information carriers (such as scientific and technological project task books, data management plans, data submission plans, data quality reports, etc.) are analyzed to extract key information related to the named fields. For example, information such as project number, data type, time range, data description, version number, etc. in the supporting information carrier is extracted, and this information is organized according to paragraph structure features and semantic features. According to the semantic similarity between the corpus in the supporting information carrier and each field in the named instance, semantic matching features related to the named fields are generated.

[0166] Based on the associated information features, feature vectors are generated to represent the feature combination of each named instance, and these feature vectors will be used to construct training samples. The associated information extraction is performed according to the naming rule features. For example, the project number corresponding to the named instance is extracted from the scientific and technological project task book, and the project number in the task book is P001; the data type and time range are extracted from the data management plan or data submission plan, and the data management plan stipulates that the data type is "precipitation data" and the time range is "2010-2020; the data description and version control information are extracted from the data quality report or other relevant documents, and the data description is "daily observation, using an automatic rain gauge" and the version control is "V1.0".

[0167] The extracted naming features, associated information features and audit opinions are integrated to form a complete training sample. For example, for a naming instance, if the audit opinion points out that the reason for the non-standard naming is "lack of time range information", the corresponding features will be marked as negative samples in the generated training samples; if the naming is standard and meets the requirements in the supporting information carrier, it will be marked as a positive sample.

[0168] The training samples are classified based on the semantic features of the audit opinions and the connection between the naming rules.

[0169] Identify key features in the naming convention. These features typically include project numbers, data types, time ranges, version control, etc. Each naming convention feature represents an important dimension in the naming convention that affects the consistency and accuracy of the naming. By analyzing these naming convention features, training samples can be more effectively classified into different groups. (The purpose of classification is to allow the model to focus on the naming inconsistencies that may be caused by each naming convention feature.) For example, one naming feature may be inconsistent in project numbers as missing or inconsistent numbers, while another naming feature may be inaccurate or incomplete in the representation of time ranges. Classifying samples according to these features helps the model identify and correct these specific types of inconsistencies.

[0170] For example, if the audit opinion is "observation data, but the observation time range is not clear", the extracted semantic features include "observation data features, missing time scale", and based on the naming rule mapping, the naming rule stipulates that "observation data should be marked with the observation time range YYMMDD-YYMMDD". The samples are classified according to the above-obtained naming features, standard problem features and associated naming rule semantics and feature types to obtain samples corresponding to different naming rules and standard problems.

[0171] The training samples are classified according to the type of naming rules. Specifically, by extracting the naming rules of the training samples, the system can automatically classify the samples. For example: if the naming rules of a sample include a complete project number, a clear time range, and version control, then the sample can be classified into the "complete project number" group or the "standard time range" group. Conversely, if the sample naming lacks clear time range or version control information, the sample can be classified into the "irregular time range" group or the "missing version control" group. The ultimate goal of sample classification is to generate a model that can distinguish between positive samples (samples with consistent naming rules) and negative samples (samples with irregular naming) under each naming rule feature.

[0172] Specifically, for the classification of training samples in the naming standard check of scientific datasets in a certain field, sample examples are as follows:

[0173]

[0174]

[0175] Item Number Naming Rules Feature Group:

[0176]

[0177] Version number naming rules:

[0178]

[0179] S130: Analyze the classified training samples based on the naming convention check task tree to obtain the scientific data set naming convention check model.

[0180] The step S130 can be subdivided into the following steps: Figure 3 The flowchart of the scientific dataset naming standard checking model provided in this embodiment includes:

[0181] S131, forming a naming feature set according to the naming features in the various training samples, and forming a matching information feature set according to the matching information carriers in the various training samples;

[0182] S132, according to the semantic features of the supporting information feature set, sub-classifying the supporting information feature set to generate multiple supporting information feature subsets, and forming a naming standard check item set according to the supporting information feature subsets;

[0183] It is understandable that semantic analysis is performed on each feature in the supporting information feature set to identify the core semantic content of these features. Semantic features may include time range, data type, version control, etc., which are the most common standards in naming specifications. Sample features can be subdivided according to semantic features. For example, the semantic features of time range may include multiple different time ranges, the semantic features of data types may cover different data categories, and the semantic features of version control may include different version numbers. Further, feature subsets such as time range feature subsets, data type feature subsets, and version control feature subsets can be generated. These feature subsets are multi-dimensional representations of sample naming rule features, ensuring that the model can check naming in a variety of situations.

[0184] S133, applying an association rule mining algorithm to construct an association rule matrix of the naming feature set and the matching information feature set;

[0185] It is understandable that there is an inherent correlation between the naming feature set and the supporting information feature set, and these relationships can be captured and represented by data mining techniques, such as using association rule mining algorithms. The association rule matrix can reveal the correlation between feature sets, provide feature vectors for model training, and help the model understand the complex logical relationships behind the naming specifications, thereby improving the accuracy and effectiveness of the model. The association rule matrix contains the association rules between the folder / file naming fields and the feature content / feature location / feature source of the supporting information carrier, as well as the association rules between the folder hierarchy and the semantic logic / semantic feature structure of the supporting information. For example, some data sets submitted by the "Pearl River Basin Hydrological Monitoring Project" include "Pearl River Basin Hydrological Monitoring / Dataset Summary / 2024 / Precipitation Data_v1.csv", "Pearl River Basin Hydrological Monitoring / Dataset Summary / 2024 / Water Level Data_v1.csv", and "Pearl River Basin Hydrological Monitoring / Dataset Summary / 2023 / Precipitation Data_v1.csv". The naming features are extracted based on the data set naming to obtain:

[0186] Project Category Data Types Data content years Version Pearl River Basin Hydrological Monitoring Observational data Precipitation data 2024 V1 Pearl River Basin Hydrological Monitoring Observational data Water level data 2024 V1 Pearl River Basin Hydrological Monitoring Observational data Precipitation data 2023 V1

[0187] Based on the naming features, the related paragraph sentences are obtained from the supporting information carrier, and then the supporting information feature set is obtained, such as:

[0188] Project Name years Data Category Data Format Feature Source Pearl River Basin Hydrological Monitoring 2024 Precipitation data CSV Data Management Plan Pearl River Basin Hydrological Monitoring 2024 Water level data CSV Data Management Plan Pearl River Basin Hydrological Monitoring 2023 Precipitation data CSV Data Management Plan Pearl River Basin Hydrological Monitoring 2023 Rainfall distribution JPG Project Technology Report Pearl River Basin Hydrological Monitoring 2023 Disaster Warning doc Project Technology Report

[0189] A high-dimensional supporting information feature set is established based on the semantic associations between supporting information features. For example, in the paragraph of the chapter "Statistical Analysis of Rainfall Distribution in the Pearl River Basin in 2024" in the project science and technology report, the associated contextual semantic features are obtained by extracting the data description feature semantics from the data management plan, and the keyword features of "rainfall distribution" and "disaster public opinion" and the corresponding project themes "rainfall forecast" and "flood warning" are obtained. Then a high-dimensional named feature set can be established, such as:

[0190]

[0191] By mining the association rules between the naming instance features and the supporting information features, we can obtain the association between the naming fields and the supporting information carrier types and the supporting information semantic features, which can be used to accurately obtain the information source of the naming fields, making data naming easier to understand.

[0192] On the other hand, the folder hierarchy can be configured based on the association rule matrix. For example, according to the semantic feature of the scientific dataset folder naming "Pearl River Basin Hydrological Monitoring / Dataset Summary / 2024 / Precipitation Data_v1.csv", the initial folder hierarchy structure is "{Project Name} / {Unknown} / {Year} / {Data Type Field})_Version Number.{Data Format}". According to the association rule mining, the folder hierarchy structure can be configured as "{Project Name} / {Project Theme} / {Year} / {Data Type Field 1} / {Data Type Field 2} / {Data File Name}". The updated folder naming examples are "Pearl River Basin Hydrological Monitoring / Basin Rainfall Forecast / 2023 / Precipitation Data / Site Collection / 20230101-20231031 Rainfall Time-sharing Statistics_v1.csv", "Pearl River Basin Hydrological Monitoring / Basin Rainfall Forecast / 2023 / Public Opinion Report / On-site Shooting / Hanjiang Site / ...". The updated hierarchical structure is easier to understand and share.

[0193] S134, optimizing the naming convention check task tree based on the naming convention check item set and the association rule matrix;

[0194] S135. Analyze the training samples based on the optimized naming convention check task tree to obtain a scientific data set naming convention check model.

[0195] It is understandable that the naming standard check task tree includes task levels, task dependencies and evaluation data foundations. Each layer of tasks corresponds to a naming rule, and each naming rule corresponds to a naming check item set. The preliminary structure of the task tree is directly derived from these check items. First, the association rule matrix is ​​analyzed and trained to obtain various types of naming field names and field level association information extraction models to obtain a naming structure that conforms to the supporting information carrier. Furthermore, the naming check item set is analyzed and trained to obtain the specification inspection model corresponding to each naming rule feature. Based on the information in the naming standard check item set, it is possible to identify which check items have strong correlations or logical dependencies in the naming specification. For example, if "project number format check" and "project number integrity check" are closely related in most naming instances, you can consider merging these two check items into one task node to reduce redundant tasks.

[0196] Based on the correlation between the naming feature set and the supporting information feature set in the correlation matrix, the naming specification check task tree can be optimized. Each element in the correlation matrix represents the correlation strength between the naming field and the supporting information feature, and this correlation information can be used to optimize the task tree. For example, some naming fields have a strong correlation with specific supporting information features. These strongly correlated check items can be merged or adjusted in order in the task tree to improve the inspection efficiency. If the correlation between the check items is high, these check items can be merged into a comprehensive task node. For example, if the "project number integrity check" and "project number format check" have a high correlation, they can be merged into one node to reduce the depth of the task tree. According to the analysis results of the correlation matrix, the execution order of the task nodes can also be adjusted to give priority to those tasks with high correlation and great impact on the naming specification, so that problems in the naming specification can be discovered more quickly. By reducing redundant inspection paths, highly correlated task nodes are placed in the same path to simplify the task execution process. This can improve the overall inspection efficiency and ensure the logical consistency of the inspection process.

[0197] In this embodiment, the training naming convention checking model is as follows: Figure 4 The flowchart for processing misclassified positive samples provided for this example includes:

[0198] S51, analyzing the positive samples using the scientific dataset naming standard checking model, identifying misclassified positive samples that are mistaken for naming standards in the positive samples, and extracting the naming rules and association rule matrix associated with the misclassified positive samples;

[0199] S52: Further optimizing the scientific dataset naming standard checking model by optimizing the rule weights in the naming rules of the misclassified positive samples and eliminating irrelevant naming rules.

[0200] It is understandable that the names of a batch of data files marked as positive samples are checked and verified. The model will analyze each positive sample based on the known association rule matrix to determine whether it complies with the naming specifications. During the model application or verification process, those positive samples that are misclassified as non-compliant are identified. For these misclassified positive samples, their naming feature sets and supporting information feature sets are extracted and compared with the association rule matrix. Through the analysis of the misclassified positive samples, the features that may cause misclassification are identified. For example, some features may be given too high weights in the naming rules, or the model may ignore some specific features. For the association rule matrix of these misclassified samples, the specific rules and features that cause misclassification are extracted.

[0201] For example, there is a sample file named "Pearl River Basin Hydrological Monitoring_Precipitation Data_2010-2020_Daily Observation_Automatic Rain Gauge_V1.0.csv". According to the naming convention, this naming is compliant because it contains key elements of the dataset, such as monitoring scope (Pearl River Basin), data type (precipitation data), time range (2010-2020), observation frequency (daily observation), tool (automatic rain gauge) and version information (V1.0). However, due to the rule weight setting of the model, the sample was mistakenly identified as non-compliant because of the lack of geographic location information in the naming. Based on the misidentified features (geographic location information), the associated features in the supporting information carriers of the same type of data / the same type of data source are obtained. According to cluster analysis and word frequency statistics, if the geographic information is not a high-frequency feature of the classified scientific dataset, the rule weight is adjusted. If the geographic information is a high-frequency feature, the user is prompted to supplement the relevant information.

[0202] After identifying the features that cause misclassification, analyze the frequency and importance of these features in the positive samples. If a feature (such as geographic location information) is not important in most positive samples, its weight can be appropriately reduced. Through cluster analysis, word frequency statistics and other methods, quantify the distribution of misclassified features in the sample, and adjust the weights of association rules accordingly to optimize the accuracy of the model. Clean up the association rule matrix and eliminate those rules that are not related to misclassification or those that have little impact on classification accuracy. Through model evaluation and repeated testing, confirm that the optimized rule matrix can more accurately reflect the naming specifications and reduce the occurrence of misclassification. After completing weight optimization and rule elimination, use the optimized rule matrix to retrain the model.

[0203] In this embodiment, the training naming convention checking model is as follows: Figure 5 The flowchart for processing misclassified negative samples provided for this example includes:

[0204] S61, comparing the naming feature distribution difference between the positive sample and the negative sample, and obtaining all naming rules associated with the naming feature distribution difference;

[0205] S62, sorting the feature frequencies of the naming feature distribution differences to obtain a frequency order, and constructing a negative sample naming feature set according to the naming specification check task tree nodes corresponding to the naming rules associated with the naming feature distribution differences and the frequency order;

[0206] It is understandable that after analyzing the difference in the distribution of naming features between positive and negative samples, the system will identify features with large differences, such as the lack of key fields in some negative samples. These difference features are usually associated with specific requirements in the naming rules. For example, if negative samples generally lack a project number that meets the specifications, this difference feature is associated with the rule in the naming rules that "the complete project number of the project task book must be included in the naming". The previously established naming specification check task tree provides a structured reference for these naming rules. For each difference feature, we will find the corresponding hierarchical node in the task tree according to its associated naming rules. For example, "Project Number Check" may be a node in the task tree, and there may be sub-nodes such as "Integrity Check" and "Format Check" under it. Next, determine which hierarchical node of the task tree these difference features belong to, and confirm their relative positions in the task tree. Then, count the frequency of each difference feature in the negative samples. For example, the frequency of negative samples lacking project numbers may be 60%, while the frequency of missing timestamps is 40%. The difference features are sorted according to these frequencies, and those with higher frequencies are prioritized to ensure that the most common naming problems can be effectively targeted when optimizing the model. Finally, the difference features are added to the negative sample naming feature set in order of their level and frequency in the task tree. This feature set mainly contains high-frequency difference features, and each feature corresponds to a specific level node in the task tree.

[0207] S63, using the scientific dataset naming standard checking model to analyze the negative samples, identify misclassified negative samples that are mistakenly considered to have irregular naming, and extract the naming rules and association rule matrix associated with the misclassified negative samples;

[0208] S64. Compare the naming rules and association rule matrix associated with the misclassified negative samples with the negative sample naming feature set to generate negative sample weighted features, and further optimize the scientific data set naming specification check model according to the negative sample weighted features.

[0209] Understandably, the naming convention check model is used to analyze negative samples and identify those samples that are misclassified. For these misclassified samples, the association rule matrix between their naming features and the supporting information features is extracted. This matrix shows the strength and pattern of associations between features, such as whether samples with "missing timestamps" also lack item numbers. The extracted association rule matrix of misclassified samples is compared with the constructed negative sample naming feature set. Specifically, the system looks at which misclassified samples have association rules that match the features in the negative sample naming feature set more highly and which do not. This comparison helps determine which features are more likely to cause misclassification and whether these features are well represented in the negative sample naming feature set. Based on the comparison results, the features in the negative sample naming feature set are weighted. The weighting process takes into account that features with higher association strength in misclassified samples may require higher weights because these features are more likely to cause naming convention issues. The parameters of the naming convention check model are adjusted using the negative sample weighted features. For example, the part of the model that handles the "missing timestamp" feature may receive a higher weight to more strictly check for such issues. By optimizing the weights, the model can better identify common problems in negative samples and reduce the occurrence of misclassification. After the initial optimization, the model may need to be run again to test whether the adjusted model can effectively reduce misclassification. If the new negative samples still have problems, it may be necessary to adjust the weights again or further analyze the newly generated misclassified samples. This process is iterative, and the accuracy of the naming convention check is gradually improved by continuously adjusting the model weights and feature processing methods.

[0210] In this embodiment, if Figure 6 The flowchart of the specific naming requirement identification model provided in this example includes:

[0211] S71, receiving different types of specific naming requirement inputs, identifying specific naming rules based on the specific naming requirements, and generating a specific naming rule checking task tree;

[0212] S72, collecting specific naming instances from the scientific data set, extracting supporting information carriers and naming features of the specific naming instances as specific training samples;

[0213] S73. Analyze and train the specific training samples based on the specific naming rule checking task tree to obtain a specific naming requirement recognition model.

[0214] It is understandable that in scientific data management, sensitive information is often involved, such as confidential research topics, confidential project names, names of people, special geographical locations, etc., which require naming according to specific confidentiality rules. These confidentiality rules cannot be directly trained through traditional dataset naming models, and need to be manually defined according to specific needs and regulations. The project team entered these specific naming requirements, including that the name must start with the letter "C", followed by the project code, data type, confidentiality level, and current date. The system receives these different types of specific naming requirements input, identifies specific naming rules based on the associated naming rule features, such as "C_[project code][data type][confidentiality level]_[date].csv", and generates a corresponding specific naming rule checking task tree. The system collects naming instances that need to follow this specific naming specification, such as "C_PRJ1_GenomeData_High_20240101.csv", extracts its naming features (such as project code, data type, confidentiality level, etc.), and uses these instances as specific training samples. In this process, the system not only simply checks the naming format, but also ensures that each part of the name is consistent with the supporting information carrier (such as data management plan or legal and regulatory documents). By checking the task tree with specific naming rules, the system conducts in-depth analysis and training on these training samples to obtain a recognition model specifically tailored for these specific confidential naming needs.

[0215] This specific naming requirement recognition model can not only check whether the naming of the newly input data set meets the requirements, but also automatically update and optimize itself based on the new input to more accurately reflect the new specific requirements. This process solves the management needs of naming rules in scientific research projects that cannot be obtained through general model training due to regulations or special provisions, ensuring that the naming of the data set is not only standardized, but also complies with specific regulations or confidentiality requirements.

[0216] In this embodiment, if Figure 7 The flowchart of the method for evaluating the naming of scientific data sets using the dual model provided in this embodiment includes:

[0217] S81, extracting the naming features and supporting information carriers of the scientific data set to be evaluated, and inputting the naming features and the supporting information carriers into a specific naming requirement recognition model to generate specific naming suggestions;

[0218] S82, according to the specific naming suggestion, removing the part of the scientific dataset to be evaluated that adopts the specific naming standard to obtain a changed naming feature and a supporting information carrier, and then inputting the changed naming feature and the supporting information carrier into the scientific dataset naming standard checking model to obtain a naming standard checking result;

[0219] S83: perform a naming update on the scientific dataset to be evaluated according to the naming specification check result, and record an update log.

[0220] Understandably, all update operations will be recorded in the update log, including the comparison of the old name with the new name, the updated timestamp, the operator, etc. This process ensures that each update has a detailed record for easy traceability and auditing.

[0221] Obtain the naming convention check results, update logs, and update feedback feature association matrix to establish a naming update verification knowledge base. The system aggregates the naming convention check results, update logs, and update feedback feature association matrix into a naming update verification knowledge base. The knowledge base contains detailed information on all updates and their results for future naming convention improvements and optimizations. This process includes mining association rules to analyze which naming features (such as field order and structure) have the greatest impact on convention compliance, thereby guiding subsequent data naming strategies. For example, the knowledge base may store the following information: comparison of naming structures before and after updates, statistical analysis of naming convention compliance, and the impact of various naming suggestions on data management efficiency.

[0222] In this embodiment, a model optimization method further provided in this embodiment specifically includes:

[0223] Obtain user feedback on the naming convention check result and the naming update, readjust the naming rules in the scientific dataset naming convention check model according to the feedback, thereby adjusting the naming convention check task tree, reprocessing the misclassified positive samples or the misclassified negative samples, and re-analyzing the training samples according to the adjusted naming convention check task tree to complete the optimization of the scientific dataset naming convention check model.

[0224] It is understandable that during the process of checking and updating the naming conventions of scientific datasets, the system will provide the final naming convention check results and naming update results to users. Users may include data managers, domain experts, and project leaders, who will evaluate and provide feedback on these results based on actual application scenarios. The system will design a user feedback interface to allow users to evaluate each updated naming structure, including whether it meets expectations and whether there are any missing contents. For example, in the climate monitoring data project, the system updated a dataset named "Project A_Precipitation Data_2023_Beijing", and user feedback may include: "The naming lacks specific data collection site information". This feedback will be collected and stored by the system.

[0225] The system extracts the collected user feedback information and performs semantic feature analysis to identify the main problems in the feedback. For example, the "lack of data collection site information" mentioned by the user is a problem of missing content in the naming field. The system will match the feedback information with the naming rule features to determine the specific problem type of the feedback.

[0226] Specifically, when the user provides feedback on the changed naming, the system will map it to the misclassified samples of the model based on the specific information of the user feedback, and identify the misclassified features contained in the user feedback. User feedback is actually a further calibration of the naming check model. If the naming result of the user feedback is still misclassified, the system will convert the feedback information into re-identification and correction of the misclassified samples. For example: Suppose user feedback indicates that the system misrecognized a date format (such as "2021.5.8") in the naming. By analyzing the feedback, the model automatically locates the misclassified samples and rules, which enables the model to re-identify misclassified samples with similar features.

[0227] It is understandable that user feedback triggers the optimization of naming rules and the adjustment of the naming standard checking task tree within the model through the feedback mechanism. The information on the feedback page will trigger the system to re-analyze the relevant misclassified features, and then update the rules in the model and adjust the weights. For example:

[0228] 1. User feedback: The user pointed out in the feedback page that the system omitted the variety number in the file name when naming, and required that each test data file name must include the variety number. The specific problem lies in the missing variety number field. The user pointed out that the naming specification does not meet the requirement of the project task book that "the test data file must include the variety number".

[0229] 2. After receiving user feedback, the model identified that the problem was related to the variety number field and located the "Variety Number Check" node in the naming convention check task tree. When analyzing the naming convention check task tree, the system found that the task node only checked the "Variety Name" and "Test Data" fields, but did not further refine to include the "Variety Number". The system adjusted the naming convention check task tree and added a node to specifically check the integrity of the variety number to ensure that this field would not be missed in subsequent naming.

[0230] 3. The system optimized the naming standard check task tree based on feedback, and added a completeness check of the variety number in the check of the "Test Data" field. The new rule clearly requires that each test data file name must contain a detailed variety number, such as "ZMB001".

[0231] 4. The system weights the feedback features of the variety number to ensure that the variety number field will be given special attention in future naming checks. This weighted optimization will ensure that in future naming checks, the model can more strictly perform checks on the variety number field to avoid similar problems from happening again.

[0232] It is understandable that user feedback will guide the system to re-analyze the training samples. The misclassified features contained in the feedback information will be extracted by the system and matched with the features in the training samples, the misclassified samples will be re-labeled, and the model's inspection rules will be optimized. For example: user feedback indicates that the naming of "climate data" is inappropriate and should include the geographical location, such as the "experimental site name" specified in the project task book, and the variety number "ZMB001" in the "seed germination test data" should also be placed after the data type field, not at the beginning of the file name. Therefore, based on the key information of the feedback, such as: "The experimental site is missing in the naming of climate data (such as 'Beijing')"; "The variety number 'ZMB001' should be placed after the data type field of the file name"; extract the features of the misclassified samples, such as: "The experimental site is missing in the climate data file name", "The naming order of the variety number does not meet the user's requirements (the field order in the file name is wrong)"; the model compares the misclassified sample features (such as "missing experimental site" and "wrong variety number position") with the original training samples. In the original training samples, the model found that the similar "climate data" naming did not require the experimental site field, so the naming rules obtained did not cover this feature.

[0233] Therefore, based on the deviation information from the above comparison, the model relabeled the misclassified training samples. The "climate data" samples containing similar naming features were relabeled as negative samples because they lacked experimental site information; the samples with incorrect variety number positions were reclassified to ensure that the field order was no longer incorrect; and the naming rules were optimized based on the feedback.

[0234] Based on the same inventive concept, the embodiment of the present application also provides a scientific data set naming standard checking and automatic updating model training system, such as Figure 3 As shown, the scientific dataset naming specification checking and automatic updating model training system may include:

[0235] Construction module 311, sample processing module 312, training module 313. Among them:

[0236] The construction module 311 is used to formulate naming rules according to naming requirements and establish a naming specification checking task tree, wherein the naming requirements include structured naming requirements and unstructured naming requirements, and the naming rules include structured naming rules and unstructured naming rules;

[0237] In this embodiment, the construction module 311 can be used to execute Figure 1 As shown in step S110, for the detailed description of the building module 311, reference may be made to the description of the step S110.

[0238] The sample processing module 312 is used to extract the review opinions, corresponding supporting information carriers and naming features attached to each naming instance of the scientific data set, generate training samples based on the review opinions, the supporting information carriers and the naming features of each naming instance, and classify the standardization of the training samples according to the review opinions to obtain classified training samples;

[0239] In this embodiment, the sample processing module 312 can be used to perform Figure 1 As shown in step S120 , for a detailed description of the sample processing module 312 , reference may be made to the description of step S120 .

[0240] The training module 313 is used to analyze the classified training samples based on the naming standard checking task tree to obtain the scientific data set naming standard checking model;

[0241] In this embodiment, the training module 313 can be used to perform Figure 1 As shown in step S130, for the specific description of the training module 313, reference may be made to the description of step S130.

[0242] Obviously, the above embodiments of the present application are merely examples for clearly illustrating the technical solution of the present application, and are not intended to limit the specific implementation methods of the present application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the claims of the present application shall be included in the protection scope of the claims of the present application.

Claims

1. A scientific dataset naming standard checking and automatic updating model training method, characterized in that: include: Formulate naming rules according to naming requirements and establish a naming specification checking task tree, wherein the naming requirements include structured naming requirements and unstructured naming requirements, and the naming rules include structured naming rules and unstructured naming rules; Extracting the review opinions, corresponding supporting information carriers and naming features attached to each naming instance of the scientific data set, generating training samples based on the review opinions, the supporting information carriers and the naming features of each naming instance, and classifying the standardization of the training samples according to the review opinions to obtain classified training samples; The classified training samples are analyzed based on the naming convention checking task tree to obtain the scientific data set naming convention checking model.

2. A scientific data set naming standard checking and automatic updating model training method according to claim 1, characterized in that: The naming rules are formulated according to the naming requirements, and a naming specification checking task tree is established, wherein the naming requirements include structured naming requirements and unstructured naming requirements, and the naming rules include structured naming rules and unstructured naming rules, specifically including: Establishing a general naming template based on the structured naming requirements; Build domain knowledge models; Generate a configurable naming template through the general naming template and the domain knowledge model; Generate a structured naming rule based on the configured naming template, wherein the structured naming rule includes one or more of naming field characteristics, field order, naming instance hierarchy constraints, version control, and rule weight; Extracting semantic features of the unstructured naming requirements and generating naming example reference rules; Based on the naming instance reference rule, construct an unstructured naming rule, wherein the unstructured naming rule includes a semantic consistency check and / or a text information integrity check; The rule checking items involved are extracted according to the content of the naming rule, the priorities of the rule checking items are extracted according to the rule weights of the naming rule, and a naming specification checking task tree is hierarchically constructed according to the rule checking items and the priorities.

3. The method for automatically updating a model for scientific dataset naming standard checking according to claim 1, characterized in that: The extracting of the audit opinions, the corresponding supporting information carriers and the naming features attached to each naming instance of the scientific data set, generating training samples based on the audit opinions, the supporting information carriers and the naming features of each naming instance, and classifying the standardization of the training samples according to the audit opinions to obtain the classified training samples specifically includes: The training samples are divided into positive samples and negative samples according to the audit opinions, and based on the relationship between the audit opinions and the naming rules, the standard type of each training sample is determined, and the training samples are normatively classified according to the standard types to obtain training samples of various standard types.

4. The method for automatically updating a model for checking scientific dataset naming standards according to claim 1, characterized in that: The analyzing the classified training samples based on the naming standard checking task tree to obtain the scientific data set naming standard checking model specifically includes: Forming a naming feature set based on the naming features in the various training samples, and forming a matching information feature set based on the matching information carriers in the various training samples; According to the semantic features of the supporting information feature set, the supporting information feature set is sub-classified to generate multiple supporting information feature subsets, and a naming specification check item set is formed according to the supporting information feature subsets; Applying an association rule mining algorithm to construct an association rule matrix of the naming feature set and the matching information feature set; Optimizing the naming convention check task tree based on the naming convention check item set and the association rule matrix; The training samples are analyzed based on the optimized naming convention checking task tree to obtain a scientific data set naming convention checking model.

5. A scientific dataset naming standard checking and automatic updating model training method according to any one of claims 3 and 4, characterized in that: The scientific dataset naming standard checking model further includes: Analyzing the positive samples using the scientific dataset naming convention checking model, identifying misclassified positive samples that are mistaken for naming conventions in the positive samples, and extracting naming rules and association rule matrices associated with the misclassified positive samples; The scientific dataset naming convention checking model is further optimized by optimizing the rule weights in the naming rules of the misclassified positive samples and eliminating irrelevant naming rules.

6. A scientific data set naming standard checking and automatic updating model training method according to any one of claims 3 and 4, characterized in that: The scientific dataset naming standard checking model further includes: Comparing the naming feature distribution differences between the positive sample and the negative sample, and obtaining all naming rules associated with the naming feature distribution differences; Sorting the feature frequencies of the naming feature distribution differences to obtain a frequency order, and constructing a negative sample naming feature set according to the naming specification check task tree nodes corresponding to the naming rules associated with the naming feature distribution differences and the frequency order; Analyze the negative samples using the scientific dataset naming standard check model, identify misclassified negative samples that are mistakenly considered to be non-standardly named, and extract the naming rules and association rule matrix associated with the misclassified negative samples; The naming rules and association rule matrix associated with the misclassified negative samples are compared with the negative sample naming feature set to generate negative sample weighted features, and the scientific data set naming specification checking model is further optimized according to the negative sample weighted features.

7. A scientific dataset naming standard checking and automatic updating model training method according to any one of claims 1 to 4, characterized in that: Also includes: Receiving inputs of different types of specific naming requirements, identifying specific naming rules based on the specific naming requirements, and generating a specific naming rule checking task tree; Collecting specific named instances from the scientific data set, extracting supporting information carriers and naming features of the specific named instances as specific training samples; The specific training samples are analyzed and trained based on the specific naming rule checking task tree to obtain a specific naming requirement recognition model.

8. A scientific data set naming standard checking and automatic updating model training method according to claim 7, characterized in that: Also includes: Extracting the naming features and supporting information carriers of the scientific data set to be evaluated, and inputting the naming features and the supporting information carriers into a specific naming requirement recognition model to generate specific naming suggestions; Eliminating the part of the scientific dataset to be evaluated that adopts the specific naming convention according to the specific naming suggestion, obtaining a changed naming feature and a supporting information carrier, and then inputting the changed naming feature and the supporting information carrier into the scientific dataset naming convention check model to obtain a naming convention check result; The naming of the scientific data set to be evaluated is updated according to the naming specification check result, and an update log is recorded.

9. A scientific data set naming standard checking and automatic updating model training method according to claim 8, characterized in that: Also includes: Obtaining user feedback on the naming convention check result and the naming update, and readjusting the naming rules in the scientific dataset naming convention check model according to the feedback, thereby adjusting the naming convention check task tree; The misclassified positive samples or the misclassified negative samples are reprocessed, and the training samples are re-analyzed according to the adjusted naming convention check task tree to complete the optimization of the scientific data set naming convention check model.

10. A scientific data set naming standard checking and automatic updating model training system, characterized in that: include: A construction module is used to formulate naming rules according to naming requirements and establish a naming specification checking task tree, wherein the naming requirements include structured naming requirements and unstructured naming requirements, and the naming rules include structured naming rules and unstructured naming rules; a sample processing module, configured to extract the review opinions, the corresponding supporting information carriers and the naming features attached to each naming instance of the scientific data set, generate training samples based on the review opinions, the supporting information carriers and the naming features of each naming instance, and classify the standardization of the training samples according to the review opinions to obtain the classified training samples; The training module analyzes the classified training samples based on the naming standard checking task tree to obtain the scientific data set naming standard checking model.

Images