Cross-enterprise product data standardization classification method and management software system thereof

Abstract

The invention provides a cross-enterprise product data standardization classification method and a management software system thereof, and belongs to the technical field of product data management, standardization management, artificial intelligence and natural language process.The classification method comprises the steps that a product database and a front-end interface are designed; inputting product data, and verifying and checking the product data; determining a data set for training, and constructing a text classification model of the long-short term memory network; classifying the product data based on a trained long-short term memory network text classification model; the management software system comprises a project information module, a contract information module, a data filling module and a standard downloading module. According to the method, the problem that classification texts processed by a traditional classification algorithm are mostly discrete and are poor in mutual relevance is considered, the global memory result of each layer is supplemented by means of a long-short-term memory network architecture, a training data set is expanded in combination with random recombination, product data management services with high universality can be provided, and the product data quality is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical fields of product data management, standardization management, artificial intelligence and natural language processing, and in particular to a cross-enterprise product data standardization classification method and a management software system thereof. Background Art

[0002] Product Data Management (PDM) is a key enterprise management strategy and technology. It focuses on the comprehensive, systematic and centralized management of all product-related information, data and processes within the enterprise. Through an integrated information system, it comprehensively manages various product-related data (such as part information, configuration, documents, CAD files, structure, permission information, etc.) and processes (such as design, manufacturing, sales, etc.). By effectively managing the product data as described above, the PDM system can ensure the accuracy, consistency and security of product data, improve the collaborative efficiency of product development, and support enterprises to respond quickly to market changes.

[0003] Standardized management of product data is an indispensable part of modern enterprise management. First, standardization helps ensure the consistency and accuracy of data and reduce misunderstandings and errors; second, unified data formats and definitions can improve the efficiency of collaboration between different departments and teams; third, standardized data management can reduce maintenance costs while ensuring data security; fourth, good data standardization management helps companies better comply with industry standards and regulatory requirements and avoid potential legal risks. However, on the one hand, existing product data management software is developed based on corporate standards and even business needs, with a low degree of standardization, which is not conducive to team collaboration and secondary development of products, and the reusability of PDM software is poor when facing new business scenarios; on the other hand, because PDM software pays too much attention to internal information management of enterprises, the standards for data processing such as the presentation form, storage method, and verification rules of product data between different enterprises are inconsistent, which reduces the circulation of products and the efficiency of price comparison, procurement, acceptance, and archiving. Natural Language Processing (NLP) is an interdisciplinary subject in the fields of computer science, artificial intelligence and linguistics. It is a technology that converts natural language (such as the language used by humans in daily life) into a computer-processable form. It aims to enable computers to understand, interpret and generate human language, so as to perform various tasks such as text classification, sentiment analysis, machine translation, etc., providing a new direction for product data standardization. However, there are conflicts and differences between the classification rules of existing text classification tools and the classification rules in this business scenario, which leads to the inability of existing classification algorithms to effectively complete the classification task and the problem of high error rate. In addition, due to the limitations of business scenarios, the amount of training sample data is small, and the effect of training using traditional text classification models is poor, and there are problems such as underfitting and unsuccessful feature extraction. Therefore, in order to solve the problems of small samples and high error rates of traditional text classification models, it is urgent and necessary to seek a cross-enterprise product data standardization classification method and its management software system to achieve versatile product data management, ensure product data quality and improve work efficiency. Summary of the invention

[0004] In view of the defects in the above-mentioned prior art, the present invention proposes a cross-enterprise product data standardization classification method. The method includes designing a product database and a front-end interface; inputting product data and verifying the product data; determining a training data set and constructing a long short-term memory network text classification model; classifying product data based on the trained long short-term memory network text classification model, and based on this, providing a management software system, which includes a project information module, a contract information module, a data reporting module and a standard download module. Taking into account the problem that the classified texts processed by traditional classification algorithms are mostly discrete and have poor correlation with each other, the present invention uses the long short-term memory network architecture to supplement the global memory results of each layer, and combines random reorganization to expand the training data set, so as to provide a highly versatile product data management service and ensure the quality of product data.

[0005] The present invention provides a cross-enterprise product data standardization classification method, which comprises the following steps:

[0006] S1. Design product database and front-end interface;

[0007] S11. Design product database: extract the reporting requirements for each product attribute from the standard, design the database field attributes and field length; extract the relationship between projects, contracts, products and between products from the standard, and design the database table structure;

[0008] S12. Designing the front-end interface: extracting the reporting requirements for each product attribute from the standard, and displaying the standard requirements in the form of input box prompts and help boxes; the front-end interface is provided with required product data items;

[0009] S2. Input product data and verify the product data;

[0010] S21. Determine product data based on the reporting requirements of the front-end interface;

[0011] S22, performing a general verification check on the determined product data;

[0012] S23, performing digital verification check on the determined product data;

[0013] S24, performing a character verification check on the determined product data;

[0014] S3, determine the training data set and build a long short-term memory network text classification model;

[0015] S31, determine a training data set based on the product name S1, product summary S2, product keywords S3, product application scenarios and description S4;

[0016] S32. Expand the training data set: randomly reorganize the product name S1, product summary S2, product keyword S3, product application scenario and description S4 text sequences of different products in the same category to generate a new input sequence;

[0017] S33. Based on the long short-term memory network architecture, the global memory results of each layer are recorded and saved for a long time to build a long short-term memory network text classification model;

[0018] S4. Classify product data based on the trained long short-term memory network text classification model;

[0019] S41. Initialize contract information and perform a classification operation to obtain contract information processing results Among them, m represents the mth layer of the long short-term memory network text classification model; last represents the last time step of the text sequence:

[0020] S42, based on the trained parameters, perform n classification operations, calculate the probability that the product data belongs to the preset classification, and return the product classification result;

[0021] S421, during the nth classification operation, the input sequence X of the long short-term memory network text classification model n It is expressed as:

[0022] X n =S 1n +S 2n +S 3n +S 4n , n = 1, 2, ... (1)

[0023] Among them, S 1n , S 2n , S 3n and S 4n Respectively represent the product name, product summary, product keywords, product application scenarios and description entered during the nth classification operation;

[0024] S422, during the nth classification operation, the initial input of the mth layer of the long short-term memory network text classification model It is expressed as:

[0025]

[0026] in, represents the global memory result of the mth layer of the long short-term memory network text classification model during the n-1th classification operation; μ represents the first influencing factor; λ represents the second influencing factor.

[0027] Further, step S22 specifically includes the following steps:

[0028] S221, check whether the product data content is complete and whether there are any missing necessary items; if the content is incomplete or there are any missing necessary items, the inspection fails and step S21 needs to be executed to re-determine the product data;

[0029] S222, check whether the length of each field item of the product data meets the requirements and whether there is an overflow risk; if it does not meet the requirements or there is an overflow, the inspection fails and step S21 needs to be executed to re-determine the product data;

[0030] Step S23 specifically includes the following steps:

[0031] S231, check whether the value in the product data is a reasonable value and whether it is within a reasonable range; if the value is unreasonable or not within a reasonable range, the inspection fails and step S21 needs to be executed to re-determine the product data;

[0032] S232, check whether the digital format in the product data is valid; if not, the inspection fails, and step S21 needs to be executed to re-determine the product data;

[0033] Step S24 specifically includes the following steps:

[0034] S241, check whether the field items in the product data are safe input; if the field items have risks, the inspection fails, and step S21 needs to be executed to re-determine the product data;

[0035] S242. Based on natural language processing technology, check whether the product data is valid input; if invalid, the inspection fails and step S21 needs to be executed to redetermine the product data.

[0036] Preferably, in step S41, the contract information in the product data is preprocessed and encoded to obtain the initialization input digital sequence X0 of the long short-term memory network text classification model, and a classification operation is performed on the initialization input digital sequence X0, and the global memory result of the mth layer of the long short-term memory network text classification model is recorded. and will As a result of contract information processing.

[0037] Preferably, the amount of data in the same category in the training data set in step S32 is expanded to 4 q times, where q represents the amount of original data under the same category.

[0038] Preferably, the initial input of the mth layer of the long short-term memory network text classification model in step S42 is Set to null value 0.

[0039] Preferably, the training data set in step S31 includes data set product data, software product data, algorithm product data, knowledge product data, and model and hardware product data.

[0040] Another aspect of the present invention provides a management software system that utilizes the aforementioned cross-enterprise product data standardization classification method, which includes a project information module, a contract information module, a data reporting module and a standard download module. The project information module is used to manage project information, project documents, contract information and product data, and provides a project-level product data interface; the contract information module is used to manage contract information, contract documents and product data, and provides a contract-level product data interface; the data reporting module is used to classify and manage product data, perform standardized product data reporting work and automatic verification of product data; the standard download module is used to provide downloading of standard files; the data reporting module includes a front-end interface, a product database, a back-end verification function and a product classification intelligent auxiliary tool; the back-end verification function includes a general verification module, a digital verification module and a character verification module.

[0041] Preferably, the front-end interface designs prompt words and guide words according to standard requirements; the product database designs database field items and library table structure according to standard requirements; the back-end verification function designs detection rules according to input content type and standard requirements; the product classification intelligent auxiliary tool uses artificial intelligence and natural language processing technology to make product classification recommendations based on the basic product information filled in by the user.

[0042] Preferably, the general verification module includes verifying the integrity of product data, checking whether the required items of the reported product data are filled in completely, verifying the standardization of product data, and checking whether the length of the reported product data complies with standard specifications; the digital verification module includes verifying the rationality of numerical product data, checking whether the reported product data is within a valid range, verifying the standardization of digital product data, and checking whether the reported product data complies with standard specifications and format specifications; the character verification module includes verifying the security of character product data, checking whether there is an injection risk in the reported product data, verifying the validity of character product data, and checking whether the reported product data is meaningless garbled code.

[0043] Compared with the prior art, the technical effects of the present invention are:

[0044] 1. The present invention designs a cross-enterprise product data standardization classification method. Taking into account the problem that the classification texts that traditional classification algorithms need to process are mostly discrete and have poor correlation with each other, while considering product information as time series input, different product information under the same contract is also regarded as time series input. The long short-term memory network (LSTM) architecture is used to design a text classification model, and the correlation between products is used to improve the accuracy of the classification results.

[0045] 2. The present invention designs a cross-enterprise product data standardization classification method. In order to solve the problem of small training data scale, the product names, product summaries, product keywords, product application scenarios and description text sequences of different products under the same category are randomly reorganized, which greatly enriches the number of samples. At the same time, in order to avoid feature confusion that may be caused by sample expansion, secondary classification is introduced as the basis for sample expansion to refine the combination requirements of text data.

[0046] 3. The present invention designs a cross-enterprise product data standardization management software system, which provides versatile product data management services for small and medium-sized enterprises or multi-enterprise joint projects with the help of project information, contract information, data reporting and standard downloading, ensures product data quality, facilitates communication and cooperation between enterprises, and improves the efficiency of communication, exchange, acceptance, etc. between enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Other features, objects and advantages of the present application will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings.

[0048] Figure 1 is a flow chart of the cross-enterprise product data standardization classification method of the present invention;

[0049] Figure 2 It is a schematic diagram of the process of verifying and checking product data of the present invention;

[0050] Figure 3 is a schematic diagram of expanding the training data set of the present invention;

[0051] Figure 4 is a schematic diagram of the secondary classification of the training data set of the present invention;

[0052] Figure 5 It is a schematic diagram of the structure of the long short-term memory network text classification model of the present invention;

[0053] Figure 6 It is a schematic diagram of the use process of the product classification intelligent auxiliary tool of the present invention;

[0054] Figure 7 It is a schematic diagram of the composition of the management software system of the present invention. DETAILED DESCRIPTION

[0055] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It should also be noted that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0056] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0057] Figure 1 The cross-enterprise product data standardization classification method of the present invention is shown, and the method includes the following steps:

[0058] S1. Design product database and front-end interface;

[0059] S11. Design product database: extract the reporting requirements for each product attribute from the standard, design the database field attributes and field length; extract the relationship between projects, contracts, products and between products from the standard, design the database table structure, implement product association function, and facilitate user retrieval.

[0060] S12. Design the front-end interface: extract the reporting requirements for each product attribute from the standard, and display the standard requirements in the form of input box prompts and help boxes; set up the required product data items in the front-end interface.

[0061] S2. Input product data and verify the product data. Figure 2 shown.

[0062] S21. Determine product data based on the reporting requirements of the front-end interface.

[0063] S22. Perform a general verification check on the determined product data.

[0064] S221. Check whether the product data content is complete and whether there are any missing necessary items; if the content is incomplete or there are any missing necessary items, the inspection fails and step S21 needs to be executed to re-determine the product data.

[0065] S222. Check whether the length of each field item of the product data meets the requirements and whether there is an overflow risk; if it does not meet the requirements or there is an overflow, the inspection fails and step S21 needs to be executed to re-determine the product data.

[0066] In a specific embodiment, the general verification check is specifically performed as follows:

[0067] After the user completes the project information form, check whether the required items, such as project number, project name, project confidentiality level, project research unit, project leader, and project start and end time, are filled in completely. Mark the required items that are not filled in in red, and prompt the user to complete the project information through a pop-up window. Check the input information of all fields, mark the fields that exceed the length limit in red, and prompt the user to modify the field content through a pop-up window.

[0068] After the user completes the contract information form, check whether the required items, such as contract number, contract name, contract confidentiality level, contract price, contract party A unit, contract party B unit, contract party B person in charge, contract party B contact number, and contract start and end time, are filled in completely. Mark the required items that are not filled in in red, and remind the user to complete the contract information through a pop-up window. Check the input information of all fields, mark the fields that exceed the length limit in red, and remind the user to modify the field content through a pop-up window.

[0069] After the user completes the model product information form, check whether the required items, such as model number, model name, model confidentiality level, data product level, model completion time, model keywords, model summary, model application scenario and description, model format, associated data product number and relationship description, model operating environment (software environment), model operating environment (hardware environment), development unit, development unit contact number, model type, model version number, model application and verification status, are filled in completely. Mark the required items that are not filled in in red, and remind the user to complete the model product information through a pop-up window. Check the input information of all fields, mark the fields that exceed the length limit in red, and remind the user to modify the field content through a pop-up window.

[0070] After the user completes the data set product information form, check whether the required items, such as data set number, data set name, data set confidentiality level, data product level, data set completion time, data set keywords, data set summary, data set application scenario and description, data set format, related data product number and relationship description, data set operating environment (software environment), data set operating environment (hardware environment), development unit, development unit contact number, data set type, data set version number, potential application scenarios and value, data quality self-assessment, data reuse self-recommendation, and total data set, are filled in completely. Mark the unfilled required items in red, and remind the user to complete the data set product information through a pop-up window. Check the input information of all fields, mark the fields that exceed the length limit in red, and remind the user to modify the field content through a pop-up window.

[0071] After the user completes the hardware product information form, check whether the required items, such as hardware number, hardware name, hardware confidentiality level, data product level, hardware completion time, hardware keywords, hardware summary, hardware application scenario and description, hardware format, related data product number and relationship description, hardware operating environment (software environment), hardware operating environment (hardware environment), development unit, development unit contact number, hardware type, main functional indicators, main performance indicators, hardware version number, hardware location, hardware storage conditions, and hardware maintenance requirements, are filled in completely. Mark the required items that are not filled in in red, and remind the user to complete the hardware product information through a pop-up window. Check the input information of all fields, mark the fields that exceed the length limit in red, and remind the user to modify the field content through a pop-up window.

[0072] After the user completes the knowledge product information form, check whether the required items, such as knowledge number, knowledge name, knowledge confidentiality, data product level, knowledge completion time, knowledge keywords, knowledge summary, knowledge application scenario and description, knowledge format, related data product number and relationship description, knowledge operating environment (software environment), knowledge operating environment (hardware environment), development unit, development unit contact number, knowledge type, knowledge version number, key technology / innovation point, number of knowledge items, assessment and verification status, current application status, potential application scenarios and value, data quality self-assessment, and data reuse self-recommendation, are filled out completely. Mark the required items that are not filled in in red, and remind the user to complete the knowledge product information through a pop-up window. Check the input information of all fields, mark the fields that exceed the length limit in red, and remind the user to modify the field content through a pop-up window.

[0073] After the user completes the software product information form, check whether the required items, including software number, software name, software confidentiality level, data product level, software completion time, software keywords, software summary, software application scenario and description, software format, related data product number and relationship description, software operating environment (software environment), software operating environment (hardware environment), development unit, development unit contact number, software type, software function description, software performance description, software version number, software dependency, software development environment (hardware configuration), software development environment (software configuration), software application and verification status, are filled in completely. Mark the required items that are not filled in in red, and remind the user to complete the software product information through a pop-up window. Check the input information of all fields, mark the fields that exceed the length limit in red, and remind the user to modify the field content through a pop-up window.

[0074] After the user completes the algorithm product information form, check whether the required items, including algorithm number, algorithm name, algorithm confidentiality level, data product level, algorithm completion time, algorithm keywords, algorithm summary, algorithm application scenario and description, algorithm format, associated data product number and relationship description, algorithm operating environment (software environment), algorithm operating environment (hardware environment), development unit, development unit contact number, algorithm input format and description, algorithm output format and description, data set format, intelligent task, algorithm technology, algorithm version number, and key technology / innovation points, are filled in completely. Mark the required items that are not filled in in red, and remind the user to complete the algorithm product information through a pop-up window. Check the input information of all fields, mark the fields that exceed the length limit in red, and remind the user to modify the field content through a pop-up window.

[0075] S23. Perform digital verification check on the determined product data.

[0076] S231. Check whether the value in the product data is a reasonable value and whether it is within a reasonable range; if the value is unreasonable or not within a reasonable range, the inspection fails and step S21 needs to be executed to re-determine the product data.

[0077] S232, check whether the digital format in the product data is valid; if invalid, the inspection fails and step S21 needs to be executed to re-determine the product data.

[0078] In a specific embodiment, the digital verification check is specifically performed as follows:

[0079] For field items that should be filled with numerical values, such as "quantity", "total amount", "number of items", and "size", check whether the values ​​meet the integer requirements (if necessary) and whether the values ​​are within a reasonable range.

[0080] For fields with "contact phone number" in the description, check whether the number format meets the length and format requirements of a valid mobile phone or landline number. Fields with incorrect formats will be marked in red and a pop-up window will be used to remind users to modify them.

[0081] S24: Perform a character verification check on the determined product data.

[0082] S241. Check whether the field items in the product data are safe inputs; if there is a risk in the field items, the inspection fails and step S21 needs to be executed to re-determine the product data.

[0083] S242. Based on natural language processing technology, check whether the product data is valid input; if invalid, the inspection fails and step S21 needs to be executed to redetermine the product data.

[0084] In a specific embodiment, the character verification check is specifically performed as follows:

[0085] Check the input content to determine whether it contains SQL statements, regular expressions, and other potentially risky content. Mark risky fields in red and prompt users to modify them through pop-up windows.

[0086] Check the input content, use NLP (natural language processing) technology to automatically segment the input content, and check the validity of the sentence based on the segmentation results. Mark the field items where the input content may be invalid in red, and remind the user to check through a pop-up window.

[0087] S3. Determine the training data set and build a long short-term memory network text classification model.

[0088] S31. Determine the training data set based on the product name S1, product summary S2, product keywords S3, product application scenarios and description S4. The training data set includes data set product data, software product data, algorithm product data, knowledge product data, and model and hardware product data.

[0089] S32. Expand the training data set: Randomly reorganize the product name S1, product summary S2, product keyword S3, product application scenario and description S4 text sequences of different products in the same category to generate new input sequences, such as Figure 3 As shown. The amount of data in the same category in the training dataset is expanded to 4 q times, where q represents the amount of original data under the same category.

[0090] The quality of the training data set largely determines the effectiveness of the model, but in actual scenarios, there is often not enough diverse data, which greatly affects the accuracy of the model. The present invention is limited to a specific business field, and the number of training data samples is relatively limited. This problem is particularly serious when it comes to product categories under the secondary classification with a low frequency of occurrence, because these products account for too small a proportion of the data set.

[0091] The present invention proposes a new solution for this application scenario. By further strengthening the existing random exchange method and expanding the scope of exchangeable content, it aims to reduce the risk of model overfitting. At the same time, by optimizing the splitting method and enhancement strategy of the data set, the feature extraction effect of the model is enhanced, and the model's recognition ability for low-frequency product categories is improved.

[0092] Considering that data samples under the same secondary classification have more uniform text features, in order to ensure the validity of the expansion results, the present invention splits the data set more finely according to the secondary classification system specified in the standard. When performing random combination, the text is reorganized based on the secondary classification results rather than the primary classification.

[0093] In a specific embodiment, the secondary classification diagram of the training data set is as follows: Figure 4 As shown. Take software data products as an example. Software products account for about 21.17% of all samples, but this type of product is mainly concentrated in the secondary classification "application software", and the data samples under the secondary classification "system software" account for only 0.93%. In order to ensure the model's recognition ability for "system software" products and avoid misjudging "system software" as algorithm data products, it is necessary to expand the "system software" samples. The algorithm randomly selects product A from the data samples marked as "software data products-system software" and records the product name (S1), then repeats the operation and records the product summary (S2) of product B, the product keywords (S3) of product C, and the product application scenarios and descriptions (S4). The above data are merged and marked as "software data products-system software" and put into the new data set.

[0094] S33. Based on the long short-term memory network architecture, the global memory results of each layer are recorded and stored for a long time to build a long short-term memory network text classification model.

[0095] As a special recurrent neural network (RNN), the long short-term memory network (LSTM) effectively solves the gradient vanishing or gradient exploding problems faced by traditional RNN when processing long sequence data by introducing the "gate" mechanism (forget gate, input gate, output gate). This enables LSTM to capture and remember long-term dependencies in the data, so that it performs well when processing data with time series and correlation, and has applications in text classification, time series prediction, anomaly detection, pattern recognition and other scenarios.

[0096] The LSTM network as a whole consists of multiple layers, each layer consists of multiple LSTM units, each unit is forward propagated, each unit will get the output of the previous layer as its own input, and its output will be used as the input of the next layer. Except for the last layer, the output of all intermediate layers will be passed to the following layers, thus establishing a process of processing input data in chronological order. Each LSTM unit in each layer consists of four parts: input gate, forget gate, output gate and candidate memory cell. The input gate controls which information can enter the LSTM network, the forget gate controls which memories need to be forgotten, the output gate controls which information is output to the hidden state, and the candidate memory cell is used to store long-term information.

[0097] Considering that in the same project or contract, product R&D, application and other scenarios are the same or complementary, the adjacent classification data are more correlated than in traditional classification scenarios. According to the actual business scenario, while considering product information as time series input, different product information under the same contract is also considered as time series input. The long short-term memory network architecture is used to design the text classification model, and the relationship between products is used to improve the accuracy of the classification results.

[0098] In order to extract and process the correlation between text sequences, based on the long short-term memory network architecture, the global memory results of each layer are recorded and saved for a long time, and a long short-term memory network text classification model is constructed, such as Figure 5 shown.

[0099] Usually, LSTM networks are used to extract dependencies within a sequence, rather than dependencies between sequences. Since contract information and product information are closely related, contract information can be used to more accurately classify product information, so the model parameters need to be adjusted so that it can identify and extract dependencies between sequences.

[0100] S4. Classify product data based on the trained long short-term memory network text classification model.

[0101] S41. Initialize contract information and perform a classification operation to obtain contract information processing results

[0102] The contract information in the product data is preprocessed and encoded to obtain the initialization input digital sequence X0 of the long short-term memory network text classification model, and a classification operation is performed on the initialization input digital sequence X0, and the global memory result of the mth layer of the long short-term memory network text classification model is recorded. and will As the result of contract information processing, last represents the last time step of the text sequence.

[0103] S42. Based on the trained parameters, perform n classification operations, calculate the probability that the product data belongs to the preset classification, and return the product classification result.

[0104] Initial input of the mth layer of the LSTM text classification model Set to null value 0.

[0105] S421, during the nth classification operation, the input sequence X of the long short-term memory network text classification model n It is expressed as:

[0106] X n =S 1n +S 2n +S 3n+S 4n , n = 1, 2, ... (1)

[0107] Among them, S 1n , S 2n , S 3n and S 4n They respectively represent the product name, product summary, product keywords, product application scenarios and descriptions entered during the nth classification operation.

[0108] S422, initial input of the mth layer of the long short-term memory network text classification model during the nth classification operation It is expressed as:

[0109]

[0110] in, It represents the global memory result of the mth layer of the short-term memory network text classification model during the n-1th classification operation; μ represents the first influencing factor; λ represents the second influencing factor.

[0111] Another aspect of the present invention provides a management software system using the aforementioned cross-enterprise product data standardization classification method, which includes a project information module, a contract information module, a data reporting module and a standard download module, such as Figure 7 shown.

[0112] The project information module is used to manage project information, project documents, contract information and product data, and provides a project-level product data interface; the contract information module is used to manage contract information, contract documents and product data, and provides a contract-level product data interface; the data reporting module is used to classify and manage product data, perform standardized product data reporting work and automatic verification of product data; the standard download module is used to provide downloads of standard files.

[0113] The data reporting module includes the front-end interface, product database, back-end verification function and product classification intelligent auxiliary tools. The front-end interface prompts and guide words are designed according to the standard requirements; the product database is designed according to the standard requirements. The database field items and library table structure are designed according to the standard requirements; the back-end verification function is designed according to the input content type and standard requirements. Detection rules.

[0114] In the front-end interface design, the product data collection form is designed according to all the submission requirements stipulated in the "Data Product Submission Requirements" series of standards, and the mandatory items are reminded by "*". In the front-end interface design, Html5 and CSS technology are used to beautify the form interface to improve the usability and aesthetics of the system interface. In the front-end interface design, the relevant requirements and descriptions of product attributes in the standard are displayed as input requirements in the input box and "Help", saving users time to consult the standard requirements.

[0115] In database design, the length of database fields is designed and security redundancy is reserved to meet user input in accordance with the relevant requirements of the "Data Product Submission Requirements" series of standards and the "Data Product Core Metadata" series of standards for product attributes. In database design, the database table structure is designed to establish the relationship between projects, contracts, products, and between products, so that users can easily access and manage products.

[0116] The back-end verification function includes a general verification module, a digital verification module, and a character verification module. The general verification module includes verifying the integrity of product data, checking whether the mandatory items of the reported product data are filled in completely, verifying the standardization of product data, and checking whether the length of the reported product data meets the standard specifications; the digital verification module includes verifying the rationality of numerical product data, checking whether the reported product data is in the valid range, verifying the standardization of digital product data, and checking whether the reported product data meets the standard specifications and format specifications; the character verification module includes verifying the security of character product data, checking whether there is an injection risk in the reported product data, verifying the validity of character product data, and checking whether the reported product data is meaningless garbled, to ensure the quality of product data. The back-end verification function is designed to verify the product information entered by the user, including the length and content, and marking the fields that have not passed the verification in red to remind the user to modify them, to ensure data security and standards, and to improve data quality.

[0117] The intelligent product classification assistance tool uses artificial intelligence and natural language processing technology to make product classification recommendations based on all or part of the basic product information filled in by the user.

[0118] Product classification intelligent auxiliary tool usage flow chart Figure 6 As shown, after the user enters the data filling module, open the auxiliary tool and enter at least one of the product name, product summary, product keywords, product application scenarios and description. The more complete and comprehensive the product information is, the more accurate the classification result will be. The product classification intelligent auxiliary tool uses artificial intelligence technology and natural language processing technology to implement the classification algorithm and give classification suggestions based on the input content. Users can choose whether to accept the classification suggestions, and the system will perform subsequent processing based on the user's selection results, including opening a new page for the corresponding category product, reclassifying, and resetting the input content.

[0119] In a specific embodiment, after the user enters at least one of the product name, product summary, product keywords, product application scenarios and description, he clicks the "Start Classification" button. The intelligent auxiliary tool for product classification automatically copies part of the input content to complete the input dimension, implements classification recommendation through the trained long short-term memory network text classification model, and returns the classification result to the user, who makes a decision. If the user chooses "Accept", the system automatically opens a new page for this type of product based on the classification result; if the user chooses "Reclassify", the system will replace the automatically filled input content (if there are unentered field items) and re-classify the recommendation; if the user chooses "Reset", the system clears all input content and waits for user input.

[0120] In a specific embodiment, take Project A "XXX Digital Construction Project", Project A Subsidiary Contract A "XXX Platform Management Software R&D Contract", and Project A Subsidiary Contract B "XXX Platform Digital Capability Enhancement R&D Contract" as examples. Among them, Contract A includes Data Product A "XXX Platform Management Software" and Data Product B "XXX Platform Product Recommendation Algorithm".

[0121] After entering the data filling module through Contract A, the product classification intelligent auxiliary tool is opened. The product classification intelligent auxiliary tool automatically reads the contract information of Contract A and possible project information, completes the initialization operation, records the final output results of each layer of the model, and waits for user input.

[0122] After entering at least one basic product information of data product A, click "OK". After the program completes the automatic completion, it will classify and calculate the input text content in combination with the recorded contract information processing results, return the classification result with the most suitable features, and temporarily store the final processing results of each layer of the model. The result returned here is "Software Data Product".

[0123] After entering at least one basic product information of data product B, click "Confirm". The product classification intelligent auxiliary tool calculates the classification result of data product B based on the user input, the contract information processing results and the processing results of data product A. The result returned here is "algorithm data product".

[0124] After completing the data entry of the product under Contract A and switching to Contract B, the product classification intelligent auxiliary tool clears all previous records, automatically reads the contract information of Contract B and the project information of Project A, automatically completes a new round of initialization operations, and waits for user input.

[0125] The invention designs a cross-enterprise product data standardization classification method and its management software system. Considering the problem that most of the classification texts that need to be processed by traditional classification algorithms are discrete and have poor correlation with each other, while considering product information as time series input, different product information under the same contract is also regarded as time series input, and a long short-term memory network (LSTM) architecture is used to design a text classification model, and the correlation between products is used to improve the accuracy of the classification result; to solve the problem of small training data scale, the product names, product summaries, product keywords, product application scenarios and description text sequences of different products under the same category are randomly reorganized, which greatly enriches the number of samples; at the same time, in order to avoid feature confusion that may be caused by sample expansion, secondary classification is introduced as the basis for sample expansion, and the combination requirements of text data are refined; with the help of project information, contract information, data reporting and standard downloading, a highly versatile product data management service is provided for small and medium-sized enterprises or for multi-enterprise joint projects, product data quality is guaranteed, communication and cooperation between enterprises are facilitated, and the work efficiency of communication, exchange, acceptance, etc. between enterprises is improved.

[0126] Finally, it should be noted that the above embodiments are only intended to illustrate rather than limit the technical solutions of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the present invention can still be modified or replaced by equivalents. Any modification or partial replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A cross-enterprise product data standardization classification method, characterized in that: It includes the following steps: S1. Design product database and front-end interface; S11. Design product database: extract the reporting requirements for each product attribute from the standard, design the database field attributes and field length; extract the relationship between projects, contracts, products and between products from the standard, and design the database table structure; S12. Designing the front-end interface: extracting the reporting requirements for each product attribute from the standard, and displaying the standard requirements in the form of input box prompts and help boxes; the front-end interface is provided with required product data items; S2. Input product data and verify the product data; S21. Determine product data based on the reporting requirements of the front-end interface; S22, performing a general verification check on the determined product data; S23, performing digital verification check on the determined product data; S24, performing a character verification check on the determined product data; S3, determine the training data set and build a long short-term memory network text classification model; S31, determine a training data set based on the product name S1, product summary S2, product keywords S3, product application scenarios and description S4; S32. Expand the training data set: randomly reorganize the product name S1, product summary S2, product keyword S3, product application scenario and description S4 text sequences of different products in the same category to generate a new input sequence; S33. Based on the long short-term memory network architecture, the global memory results of each layer are recorded and saved for a long time to build a long short-term memory network text classification model; S4. Classify product data based on the trained long short-term memory network text classification model; S41. Initialize contract information and perform a classification operation to obtain contract information processing results Among them, m represents the mth layer of the long short-term memory network text classification model; last represents the last time step of the text sequence: S42, based on the trained parameters, perform n classification operations, calculate the probability that the product data belongs to the preset classification, and return the product classification result; S421, during the nth classification operation, the input sequence X of the long short-term memory network text classification model n It is expressed as: X n =S 1n +S 2n +S 3n +S 4n ,n=1,2,…… (1) Among them, S 1n , S 2n , S 3n and S 4n Respectively represent the product name, product summary, product keywords, product application scenarios and description entered during the nth classification operation; S422, during the nth classification operation, the initial input of the mth layer of the long short-term memory network text classification model It is expressed as: in, represents the global memory result of the mth layer of the long short-term memory network text classification model during the n-1th classification operation; μ represents the first influencing factor; λ represents the second influencing factor.

2. The cross-enterprise product data standardization classification method according to claim 1, characterized in that: Step S22 specifically includes the following steps: S221, check whether the product data content is complete and whether there are any missing necessary items; if the content is incomplete or there are any missing necessary items, the inspection fails and step S21 needs to be executed to re-determine the product data; S222, check whether the length of each field item of the product data meets the requirements and whether there is an overflow risk; if it does not meet the requirements or there is an overflow, the inspection fails and step S21 needs to be executed to re-determine the product data; Step S23 specifically includes the following steps: S231, check whether the value in the product data is a reasonable value and whether it is within a reasonable range; if the value is unreasonable or not within a reasonable range, the inspection fails and step S21 needs to be executed to re-determine the product data; S232, check whether the digital format in the product data is valid; if not, the inspection fails, and step S21 needs to be executed to re-determine the product data; Step S24 specifically includes the following steps: S241, check whether the field items in the product data are safe input; if the field items have risks, the inspection fails, and step S21 needs to be executed to re-determine the product data; S242. Based on natural language processing technology, check whether the product data is valid input; if invalid, the inspection fails and step S21 needs to be executed to redetermine the product data.

3. The cross-enterprise product data standardization classification method according to claim 1, characterized in that: In step S41, the contract information in the product data is preprocessed and encoded to obtain the initialization input digital sequence X0 of the long short-term memory network text classification model, and a classification operation is performed on the initialization input digital sequence X0 to record the global memory result of the mth layer of the long short-term memory network text classification model. and will As a result of contract information processing.

4. The cross-enterprise product data standardization classification method according to claim 1, characterized in that: The amount of data in the same category in the training data set in step S32 is expanded to 4 q times, where q represents the amount of original data under the same category.

5. The cross-enterprise product data standardization classification method according to claim 1, characterized in that: The initial input of the mth layer of the long short-term memory network text classification model in step S42 Set to null value 0.

6. The cross-enterprise product data standardization classification method according to claim 1, characterized in that: The training data set in step S31 includes data set product data, software product data, algorithm product data, knowledge product data, and model and hardware product data.

7. A management software system for the cross-enterprise product data standardization classification method according to any one of claims 1 to 6, characterized in that: It includes a project information module, a contract information module, a data reporting module and a standard download module. The project information module is used to manage project information, project documents, contract information and product data, and provide a project-level product data interface; the contract information module is used to manage contract information, contract documents and product data, and provide a contract-level product data interface; the data reporting module is used to classify and manage product data, perform standardized product data reporting work and automatic verification of product data; the standard download module is used to provide downloads of standard files; the data reporting module includes a front-end interface, a product database, a back-end verification function and a product classification intelligent auxiliary tool; the back-end verification function includes a general verification module, a digital verification module and a character verification module.

8. The management software system for cross-enterprise product data standardization classification method according to claim 7, characterized in that: The front-end interface is designed with prompt words and guide words according to standard requirements; The product database designs database field items and library table structure according to standard requirements; the back-end verification function designs detection rules according to input content type and standard requirements; the product classification intelligent auxiliary tool uses artificial intelligence and natural language processing technology to make product classification recommendations based on the basic product information filled in by the user.

9. The management software system for cross-enterprise product data standardization classification method according to claim 7, characterized in that: The general verification module includes verifying the integrity of product data, checking whether the required items of the reported product data are fully filled in, verifying the standardization of product data, and checking whether the length of the reported product data complies with the standard specifications; the digital verification module includes verifying the rationality of numerical product data, checking whether the reported product data is within the valid range, verifying the standardization of digital product data, and checking whether the reported product data complies with the standard specifications and format specifications; the character verification module includes verifying the security of character product data, checking whether there is an injection risk in the reported product data, verifying the validity of character product data, and checking whether the reported product data is meaningless garbled.

Images