A data management system for software development

Abstract

The present invention discloses a data management system, which relates to the technical field of software development. The system includes a data acquisition unit, a data analysis unit, a requirements document acquisition unit, a design data processing unit, and a data association unit. Through the data analysis unit, the present invention can obtain the requirements analysis documents of multiple existing projects and extract the keywords in the documents, enabling users to quickly understand the content of each requirements analysis document, facilitating more detailed initial requirement descriptions provided by users to the collector role model subsequently, reducing the number of missing requirements to a certain extent. At the same time, the data association unit calculates the similarity values between the keyword information of the function description-related documents and the text information of the program flow-related picture data, establishing an association between the function description documents with high similarity values and the program flow pictures. When the documents are modified, the related process pictures will also be automatically output for unified modification, ensuring the unity of the data.

Description

Technical Field

[0001] The present invention relates to the technical field of data management, and specifically to a data management system for software development. Background Art

[0002] Software development is a process of building a software system or the software part in the system according to user requirements, mainly including requirements analysis, design, implementation, testing, and maintenance. Corresponding documents and data need to be generated in each stage. Especially in the two stages of requirements analysis and design, continuous modifications are required at the beginning of the project. The purpose of data management is to effectively collect, store, process, and apply these data using software technology. In the invention patent with the application number 202310545415.6, "A data management method and system for software development based on cloud-edge collaboration" is disclosed, which relates to the technical field of data management. Step 1: Obtain the requirements of the development software and give the software development mode, display the advantages and disadvantages of the corresponding software development mode according to the stability coefficient of the historical software development mode, and then obtain the determined software development mode; Step 2: After determining the software development mode, determine the software developers; Step 3: After determining the software development mode and software developers; Step 4: Set the software development test mode and conduct software test intensity grading; Step 5: Monitor the whole process during software testing; Step 6: Obtain the historical management data of software testing. The present invention stores the stability of the software development mode as a reference for subsequent software development mode selection, greatly improving the efficiency of the preparatory work for information reference in the early stage of software development and reducing the probability of instability of the developed software.

[0003] The above-mentioned prior art solves the problems of difficult monitoring of the software development process and progress and time prediction, etc. However, during the system operation, for the generation of the requirements analysis document of the project in the software development process, continuous manual modification and improvement are required, so that the quality of the requirements analysis document cannot be guaranteed, and the time for determining the document is relatively long, affecting the progress of the project development. In the design process, there is usually a certain connection between the document and the picture content. However, when modifying the document in this system, the relevant pictures will not be actively output, and the unity of the data cannot be guaranteed. Summary of the Invention

[0004] The purpose of the present invention is to provide a data management system for software development to solve the problems raised in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: A data management system for software development, including a data analysis unit;

[0006] A data acquisition unit. After the data acquisition unit obtains the current project name and the field involved, it uses a search engine to query in the database according to the project name and the field involved, retrieves relevant project data that meets the conditions to a storage device, sets team tasks in the ChatModeler modeling framework, and uses a large language model to construct three role models according to the relevant definitions in the team tasks. The three role models are a collector role model, a modeler role model, and an inspector role model respectively;

[0007] A requirements document acquisition unit. The requirements document acquisition unit determines the initial requirements description based on the project requirements analysis document and keyword-related data in the data analysis unit, and transmits the initial requirements description to the collector role model. The collector role model sorts out all conversation records to generate requirements fragments, and transmits them to the modeler role model. The modeler role model extracts modeling elements. After the inspector role model obtains the modeling elements, it detects the modeling elements according to the detection rules. If the inspector role model detects that there is no information missing, it feedbacks it to the collector role model. The collector role model determines the requirements document according to the requirements analysis template and the complete requirements description, and stores the relevant data in the database;

[0008] A design data processing unit. After the design data processing unit obtains the function description-related document data and relevant data, it transmits these data to a storage device. After determining the keywords of these documents through a keyword extraction algorithm, it traverses the program flow-related picture data through a file detection algorithm to obtain the text information in the pictures;

[0009] A data association unit. The data association unit uses a similarity checker to calculate the similarity values between the text information and each document keyword, determines the document with the highest similarity value as being associated with the program flow-related picture data, and once it is determined that the document data has been marked, retrieves the picture text information with the same association mark as the document data. If the similarity value of the current text information is higher than the picture text information with the association mark, the current association mark is deleted, and the similarity checker is used to calculate the similarity values between the picture text information and other documents except the current document, and the document without the association mark is re-marked. When the document information is modified during the design process, the processor traverses all the program flow-related picture data to obtain the picture data with the same association mark as the document, and outputs it through a visual interface.

[0010] Preferably, the data acquisition unit includes a project information acquisition module, a task definition module, and a role model construction module. After the project information acquisition module obtains the current project name and the involved field, it transmits them to a search engine, and uses the search engine to query in the database according to the project name and the involved field, and retrieves relevant project data that meets the conditions to a storage device. The task definition module sets team tasks in the ChatModeler modeling framework, where the team tasks include team definition, function definition, behavior definition, and interaction definition. In the team definition, the numbers of collectors, modelers, and inspectors are clearly specified. In the function definition, the functions possessed by the collectors, modelers, and inspectors are clearly specified. In the behavior definition, the execution steps of the three are clearly specified. In the interaction definition, the feedback content among the three is clearly specified. The role model construction module uses a large language model to construct three role models according to the relevant definitions in the team tasks, where the three role models are the collector role model, the modeler role model, and the inspector role model. The large language model is specifically:

[0011]

[0012] Among them, A represents the attention matrix, D v represents the vector dimension of V, D k represents the vector dimension of K, N represents the length of Q, M represents the length of K, Q, K, V represent input vectors, K T represents the transpose matrix of K.

[0013] Preferably, the data analysis unit includes a vector generation module and a keyword output module. After the vector generation module extracts relevant project data that meets the conditions in the data acquisition unit, it retrieves the requirement analysis document in the project, performs sentence splitting on the document through natural language processing tools, and then performs word segmentation operation on a sentence-by-sentence basis to construct a document vector and a vocabulary vector. The semantic similarity algorithm is used to calculate the similarity between each vocabulary and the document content. The keyword output module sets corresponding weights for each vocabulary according to its similarity, uses the keyword extraction algorithm to calculate the scores according to the weights of each vocabulary, sorts the vocabulary according to the scores and deletes redundant vocabulary, and then selects multiple vocabulary with high scores as keywords for output. The semantic similarity algorithm is specifically:

[0014]

[0015] Among them, κ γ represents the semantic information discrimination authority measure based on the metric value γ, κ ιDenote the semantic information discrimination permission quantity based on the metric ι, γ, ι denote the semantic information sensitivity metric values of two randomly selected document segments, and γ≠ι, c γ Denote the semantic information similarity marking parameter based on the metric γ, c ι Denote the semantic information similarity marking parameter based on the metric ι, j denotes the preprocessing expression definition of text data, n, N denote the numbers.

[0016] Preferably, the requirement document acquisition unit includes a requirement segment generation module and an element detection module. The requirement segment generation module determines the initial requirement description according to the project requirement analysis document and keyword-related data in the data analysis unit, and transmits the initial requirement description to the collector role model in the form of a conversation. The collector role model sorts out all conversation records to generate requirement segments, and transmits them to the modeler role model. The element detection module uses the modeler role model to extract modeling elements from the requirement segments, and transmits the modeling elements to the inspector role model. After obtaining the modeling elements by the inspector role model, the modeling elements are detected according to the detection rules.

[0017] Preferably, the requirement document acquisition unit further includes a missing requirement supplement module and a document determination module. If the missing requirement supplement module detects information missing, it feeds back the missing information to the collector role model. The collector role model asks about the missing requirements in the form of a conversation, generates requirement segments again after getting the feedback, uses the modeler role model to extract new modeling elements, and the inspector role model conducts a re-detection. If the inspector role model detects no information missing, the document determination module feeds it back to the collector role model. The collector role model determines the requirement document according to the requirement analysis template and the complete requirement description, and at the same time, the modeler role model determines the requirement analysis model through the modeling elements, and stores the relevant data of the requirement document and the requirement analysis model in the database.

[0018] Preferably, the design data processing unit includes a data receiving module and a text information extraction module. After the data receiving module obtains the function description-related document data, project-related table data, program flow-related picture data, and function implementation algorithm data, it transmits these data to the storage device. The text information extraction module uses natural language processing tools to perform word segmentation on the function description-related document data in the data receiving module, determines the keywords of these documents through the keyword extraction algorithm, and traverses the program flow-related picture data through the file detection algorithm to obtain the text information in the pictures.

[0019] Preferably, the data association unit includes a numerical calculation module and an association mark analysis module. The numerical calculation module calculates the similarity value between the text information and each document keyword using a similarity checker, and determines the document with the highest similarity value as being associated with the picture data related to the program flow. If the document data is not marked, the association mark analysis module separately performs association marking on the document data and the picture data. If the document data is already marked, after retrieving the picture text information with the same association mark as the document data, the similarity checker is used to recalculate the similarity value between the picture text information with the association mark and the document keyword, and compare it with the similarity value of the current text information.

[0020] Preferably, the data association unit further includes a numerical comparison module and a data output module. If the similarity value of the current text information is higher than that of the picture text information with an association mark, the numerical comparison module deletes the current association mark, and uses the similarity checker to calculate the similarity value between the picture text information and other documents except the current document, and re-marks the document without an association mark. If the similarity value of the current text information is lower than that of the picture text information with an association mark, the similarity checker is used to calculate the similarity value between the current picture text information and other documents except the current document, and the document without an association mark is marked. When the document information is modified during the design process, the data output module causes the processor to traverse all the picture data related to the program flow, thereby obtaining the picture data with the same association mark as the document, and outputting it through a visual interface.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0022] 1. The present invention constructs role models of collectors, modelers, and inspectors through the data acquisition unit and the requirements document acquisition unit. After multiple interactions with the user, the requirements document can be determined in a short time. In the traditional requirements analysis process, the work of collectors, modelers, and inspectors is usually completed manually. Due to the different technical levels of collectors, modelers, and inspectors, the quality of the requirements analysis document often cannot be guaranteed. By analyzing it using large language model-related technologies, the user is enabled to actively provide relevant requirements, and the model quickly determines the user's requirements in the form of inquiries, which can not only save a large amount of time, but also make the relevant data of the requirements analysis document more complete.

[0023] 2. Through the data analysis unit, the present invention can obtain the requirement analysis documents of multiple existing projects and extract the keywords in the documents, enabling users to quickly understand the content of each requirement analysis document, determine the specific requirements of this project based on these documents, facilitating more detailed initial requirement descriptions provided by users to the collector role model later, reducing the number of missing requirements to a certain extent. At the same time, the data association unit calculates the similarity values between the keyword information of the function description-related documents and the text information of the program flow-related picture data, establishing associations between the function description documents with high similarity values and the program flow pictures. When the documents are modified, the related flow pictures will also be automatically output and modified uniformly, ensuring the unity of the data. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the overall system flow provided by an embodiment of the present invention;

[0025] Figure 2 It is a block diagram of the internal modules of the data acquisition unit provided by an embodiment of the present invention;

[0026] Figure 3 It is a block diagram of the internal modules of the requirement document acquisition unit provided by an embodiment of the present invention;

[0027] Figure 4 It is a block diagram of the internal modules of the design data processing unit provided by an embodiment of the present invention.

[0028] In the figure: 1. Data acquisition unit; 101. Project information acquisition module; 102. Task definition module; 103. Role model construction module; 2. Data analysis unit; 201. Vector generation module; 202. Keyword output module; 3. Requirement document acquisition unit; 301. Requirement segment generation module; 302. Element detection module; 303. Missing requirement supplement module; 304. Document determination module; 4. Design data processing unit; 401. Data reception module; 402. Text information extraction module; 5. Data association unit; 501. Numerical calculation module; 502. Association mark analysis module; 503. Numerical comparison module; 504. Data output module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0030] Please refer to Figures 1-4, the present invention provides a technical solution: a data management system for software development, including a data analysis unit 2;

[0031] A data acquisition unit 1. After the data acquisition unit 1 obtains the current project name and the involved field, it uses a search engine to query in the database according to the project name and the involved field, retrieves the relevant project data that meets the conditions to the storage device, sets the team tasks in the ChatModeler modeling framework, and uses a large language model to construct three role models according to the relevant definitions in the team tasks. Among them, the three role models are a collector role model, a modeler role model, and an inspector role model;

[0032] A requirements document acquisition unit 3. The requirements document acquisition unit 3 determines the initial requirements description according to the project requirements analysis document and the keyword-related data in the data analysis unit 2, transmits the initial requirements description to the collector role model. The collector role model sorts out all the conversation records to generate requirement fragments and transmits them to the modeler role model. The modeler role model extracts modeling elements. After the inspector role model obtains the modeling elements, it detects the modeling elements according to the detection rules. If the inspector role model detects that there is no information missing, it feeds back to the collector role model. The collector role model determines the requirements document according to the requirements analysis template and the complete requirements description, and stores the relevant data in the database;

[0033] A design data processing unit 4. After the design data processing unit 4 obtains the function description-related document data and the relevant data, it transmits these data to the storage device. After determining the keywords of these documents through the keyword extraction algorithm, it traverses the program flow-related picture data through the file detection algorithm to obtain the text information in the pictures;

[0034] A data association unit 5. The data association unit 5 uses a similarity checker to calculate the similarity value between the text information and each document keyword, determines the document with the highest similarity value as being associated with the program flow-related picture data. Once it is determined that the document data has been marked, it retrieves the picture text information with the same association mark as the document data. If the similarity value of the current text information is higher than the picture text information with the association mark, it deletes the current association mark and uses the similarity checker to calculate the similarity value between the picture text information and other documents except the current document, and re-marks the document without the association mark. When the document information is modified during the design process, the processor will traverse all the program flow-related picture data to obtain the picture data with the same association mark as the document and output it through the visualization interface.

[0035] The data acquisition unit 1 includes a project information acquisition module 101, a task definition module 102, and a role model construction module 103. After the project information acquisition module 101 obtains the current project name and the involved field, it transmits them to the search engine. The search engine queries the database according to the project name and the involved field, and retrieves the relevant project data that meets the conditions to the storage device. The task definition module 102 sets the team tasks in the ChatModeler modeling framework. The team tasks include team definition, function definition, behavior definition, and interaction definition. In the team definition, the numbers of collectors, modelers, and inspectors are clearly specified. In the function definition, the functions possessed by collectors, modelers, and inspectors are clearly specified. In the behavior definition, the execution steps of the three are clearly specified. In the interaction definition, the feedback content among the three is clearly specified. The role model construction module 103 uses a large language model to construct three role models according to the relevant definitions in the team tasks. The three role models are the collector role model, the modeler role model, and the inspector role model. The large language model is specifically:

[0036]

[0037] Among them, A represents the attention matrix, D v represents the vector dimension of V, D k represents the vector dimension of K, N represents the length of Q, M represents the length of K, Q, K, V represent input vectors, K T represents the transpose matrix of K;

[0038] The data analysis unit 2 includes a vector generation module 201 and a keyword output module 202. After the vector generation module 201 extracts the relevant project data that meets the conditions in the data acquisition unit 1, it retrieves the requirements analysis document in the project, performs sentence splitting on the document through natural language processing tools, and then performs word segmentation operation for each sentence to construct document vectors and vocabulary vectors. The semantic similarity algorithm is used to calculate the similarity between each vocabulary and the document content. The keyword output module 202 sets corresponding weights for each vocabulary according to its similarity, uses the keyword extraction algorithm to calculate the scores according to the weights of each vocabulary, sorts the vocabulary according to the scores and deletes redundant vocabulary, and then selects multiple vocabulary with high scores as keywords for output. The semantic similarity algorithm is specifically:

[0039]

[0040] Among them, κ γ represents the semantic information discrimination permission quantity based on the metric value γ, κ ιDenote the semantic information discrimination permission quantity based on the metric value ι, γ, ι denote the semantic information sensitivity metric values of two randomly selected document fragments, and γ ≠ ι, c γ Denote the semantic information similarity marking parameter based on the metric value γ, c ι Denote the semantic information similarity marking parameter based on the metric value ι, j denotes the preprocessing expression definition of text data, n, N denote numbers;

[0041] The keyword extraction algorithm is specifically as follows:

[0042]

[0043] Among them, b 1 , b 2 …, b n Denote n different semantic information classification indicators of similarity document fragments, Denote the vector residual value of the semantic information of the document fragment, θ 0 Denote the initial assignment of the vector residual, Denote the initial assignment of the similarity index, M denotes the similarity metric value, V denotes the keyword weight, Denote the general assignment condition of the alternative parameter;

[0044] The requirement document acquisition unit 3 includes a requirement fragment generation module 301 and an element detection module 302. The requirement fragment generation module 301 determines the initial requirement description according to the project requirement analysis document and keyword-related data in the data analysis unit 2, and transmits the initial requirement description to the collector role model in the form of a session. The collector role model organizes and generates requirement fragments based on all session records and transmits them to the modeler role model. The element detection module 302 extracts modeling elements from the requirement fragments using the modeler role model and transmits the modeling elements to the inspector role model. After obtaining the modeling elements using the inspector role model, the modeling elements are detected according to the detection rules;

[0045] The requirement document acquisition unit 3 further includes a missing requirement supplement module 303 and a document determination module 304. If the missing requirement supplement module 303 detects information missing, it feeds back the missing information to the collector role model. The collector role model asks about the missing requirements in the form of a conversation, generates a new requirement fragment after getting the feedback, extracts new modeling elements using the modeler role model, and the inspector role model conducts a re-inspection. If the inspector role model in the document determination module 304 detects no missing information, it feeds it back to the collector role model. The collector role model determines the requirement document according to the requirement analysis template and the complete requirement description. At the same time, the modeler role model determines the requirement analysis model based on the modeling elements, and stores the relevant data of the requirement document and the requirement analysis model in the database;

[0046] The design data processing unit 4 includes a data receiving module 401 and a text information extraction module 402. After the data receiving module 401 obtains the document data related to function descriptions, the table data related to the project, the picture data related to the program flow, and the function implementation algorithm data, it transmits this data to the storage device. The text information extraction module 402 uses natural language processing tools to perform word segmentation on the document data related to function descriptions in the data receiving module 401. After determining the keywords of these documents through the keyword extraction algorithm, it traverses the picture data related to the program flow through the file detection algorithm to obtain the text information in the pictures;

[0047] The data association unit 5 includes a numerical calculation module 501 and an association mark analysis module 502. The numerical calculation module 501 calculates the similarity value between the text information and each document keyword using a similarity checker, and determines the document with the highest similarity value as being associated with the picture data related to the program flow. If the document data is not marked, the association mark analysis module 502 respectively performs association marking on it and the picture data. If the document data is already marked, after retrieving the picture text information with the same association mark as the document data, it uses the similarity checker to recalculate the similarity value between the picture text information with the association mark and the document keyword, and compares it with the similarity value of the current text information;

[0048] The data association unit 5 further includes a numerical comparison module 503 and a data output module 504. If the similarity value of the current text information is higher than that of the picture text information with an association mark, the numerical comparison module 503 deletes the current association mark and uses the similarity checker to calculate the similarity value of the picture text information with other documents except the current document, and re-marks the document without an association mark. If the similarity value of the current text information is lower than that of the picture text information with an association mark, the similarity checker is used to calculate the similarity value of the current picture text information with other documents except the current document, and the document without an association mark is marked. When the document information is modified during the design process, the data output module 504 enables the processor to traverse all the picture data related to the program flow, so as to obtain the picture data with the same association mark as the document, and output it through the visualization interface.

[0049] Working principle: In the present invention, the project information acquisition module 101 in the data acquisition unit 1 uses a search engine to query in a database according to the project name and the involved field, retrieves relevant project data that meets the conditions into a storage device, uses the task definition module 102 to set team tasks, and the role model building module 103 uses a large language model to construct a collector role model, a modeler role model, and an inspector role model according to the relevant definitions in the team tasks. The vector generation module 201 in the data analysis unit 2 uses a semantic similarity algorithm to calculate the similarity between words and document content, and the keyword output module 202 selects multiple words as keywords for output. The requirement segment generation module 301 in the requirement document acquisition unit 3 transmits the initial requirement description to the collector role model, generates requirement segments through the collector role model, and transmits them to the modeler role model. The element detection module 302 uses the modeler role model to extract modeling elements according to the requirement segments and transmits the modeling elements to the inspector role model for detection. When the missing requirement supplement module 303 detects information missing, it feedbacks it to the collector role model, generates requirement segments again, uses the modeler role model to extract new modeling elements, and the inspector role model conducts re-detection. When the inspector role model detects no information missing, the document determination module 304 feedbacks it to the collector role model to determine the requirement document and the requirement analysis model. After the data receiving module 401 in the design data processing unit 4 obtains function description related document data and other data, it transmits these data to the storage device. The text information extraction module 402 determines the keywords of the relevant documents and the text information in the program flow pictures. The numerical calculation module 501 in the data association unit 5 uses a similarity checker to calculate the similarity value between the text information and each document keyword. Through the associated marking analysis module 502, when the document data is not marked, it is respectively associated and marked with the picture data. If the document data has been marked, after retrieving the picture text information with the same associated mark as the document data, it uses the similarity checker to recalculate the similarity value between the picture text information with the associated mark and the document keyword, and compares it with the similarity value of the current text information. The numerical comparison module 503, if the similarity value of the current text information is higher than the picture text information with the associated mark, deletes the current associated mark, and uses the similarity checker to calculate the similarity value between the picture text information and other documents except the current document, and selects the unmarked document to be marked again. If the similarity value of the current text information is lower than the picture text information with the associated mark, it uses the similarity checker to calculate the similarity value between the current picture text information and other documents except the current document, and selects the unmarked document for marking. The data output module 504 obtains the picture data with the same associated mark as the document and outputs it through a visualization interface.

[0050] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0051] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A data management system for software development, characterized in that: The data acquisition unit (1) acquires the name of the current project and the field involved, uses a search engine to query the database according to the project name and the field involved, retrieves relevant project data that meets the conditions to a storage device, sets the team task in the ChatModeler modeling framework, and uses a large language model to construct three role models according to the relevant definitions in the team task, wherein the three role models are a collector role model, a modeler role model, and a tester role model, and the large language model is specifically: in, A represents the attention matrix, D v Denotes the vector dimension of V, D k represents the vector dimension of K, N represents the length of Q, M represents the length of K, Q, K, V represents the input vector, K T represents the transposed matrix of K; The data analysis unit (2) comprises a vector generation module (201) and a keyword output module (202). After extracting relevant project data that meets the conditions in the data acquisition unit (1), the vector generation module (201) retrieves the demand analysis document in the project, performs sentence segmentation processing on the document using a natural language processing tool, and then performs word segmentation operation on the sentence as a unit, thereby constructing a document vector and a vocabulary vector, and calculating the similarity between each vocabulary and the document content using a semantic similarity algorithm. The keyword output module (202) sets a corresponding weight for each vocabulary according to its similarity, calculates a score according to the weight of each vocabulary using a keyword extraction algorithm, sorts the vocabulary according to the score and deletes redundant vocabulary, and then selects a plurality of vocabulary with high scores as keywords for output. The semantic similarity algorithm is specifically as follows: Among them, κ γ represents the semantic information discriminant authority based on the metric value γ, κ ι represents the semantic information discrimination authority based on the metric value ι, γ,ι represents the semantic information sensitivity metric of two randomly selected document fragments, and γ≠ι, c γ represents the semantic information similarity labeling parameter based on the metric value γ, c ι represents the semantic information similarity labeling parameter based on the metric value ι, j represents the preprocessing expression definition of the text data, and n, N represent the number; A requirement document acquisition unit (3), wherein the requirement document acquisition unit (3) determines an initial requirement description based on the project requirement analysis document and keyword-related data in the data analysis unit (2), and transmits the initial requirement description to the collector role model. The collector role model generates requirement fragments based on all session records, and transmits them to the modeler role model. The modeler role model is used to extract modeling elements. After the modeling elements are obtained by the inspector role model, the modeling elements are inspected according to the inspection rules. If the inspector role model detects that there is no missing information, it is fed back to the collector role model. The collector role model determines the requirement document based on the requirement analysis template and the complete requirement description, and stores the relevant data in the database. A design data processing unit (4) is provided, wherein the design data processing unit (4) obtains document data related to the function description and related data, transfers the data to a storage device, determines keywords of the documents through a keyword extraction algorithm, and then traverses program flow related image data through a file detection algorithm to obtain text information in the image; A data association unit (5), the data association unit (5) comprising a numerical calculation module (501), an association mark analysis module (502), a numerical comparison module (503) and a data output module (504), the numerical calculation module (501) using a similarity checker to calculate the similarity value between the text information and each document keyword, and to determine the document with the highest similarity value as being associated with the image data related to the program flow, the association mark analysis module (502) if the document data is not marked, then respectively associates it with the image data, if the document data is marked, after retrieving the image text information with the same association mark as the document data, and then using the similarity checker to recalculate the similarity value between the image text information with the association mark and the document keyword, and compares the similarity value with the similarity value of the current text information. The numerical comparison module (503) deletes the current associated mark if the similarity value of the current text information is higher than that of the picture text information with the associated mark, and uses a similarity checker to calculate the similarity value of the picture text information with other documents other than the current document, and takes the document without the associated mark to re-mark it; if the similarity value of the current text information is lower than that of the picture text information with the associated mark, the similarity checker is used to calculate the similarity value of the current picture text information with other documents other than the current document, and takes the document without the associated mark to mark it; when the document information is modified during the design process, the processor of the data output module (504) traverses all the picture data related to the program flow, thereby obtaining the picture data with the same associated mark as the document, and outputs it through a visual interface.

2. A data management system for software development according to claim 1, characterized in that: The data acquisition unit (1) includes a project information acquisition module (101), a task definition module (102) and a role model building module (103). After the project information acquisition module (101) acquires the current project name and the field involved, it transmits them to a search engine, uses the search engine to query the database according to the project name and the field involved, and retrieves the relevant project data that meets the conditions into a storage device. The task definition module (102) sets the team task in the ChatModeler modeling framework, wherein the team task includes a team definition, a function definition, a behavior definition and an interaction definition, wherein the team definition clearly indicates the number of collectors, modelers and inspectors, the function definition clearly indicates the functions of the collectors, modelers and inspectors, the behavior definition clearly indicates the execution steps of the three, and the interaction definition clearly indicates the feedback content between the three. The role model building module (103) uses a large language model to construct three role models according to the relevant definitions in the team task, wherein the three role models are a collector role model, a modeler role model and a inspector role model.

3. A data management system for software development according to claim 1, characterized in that: The requirement document acquisition unit (3) includes a requirement fragment generation module (301) and an element detection module (302). The requirement fragment generation module (301) determines the initial requirement description based on the project requirement analysis document and keyword-related data in the data analysis unit (2), and transmits the initial requirement description to the collector role model in the form of a conversation. The collector role model organizes and generates requirement fragments based on all conversation records, and transmits them to the modeler role model. The element detection module (302) uses the modeler role model to extract modeling elements based on the requirement fragments, and transmits the modeling elements to the tester role model. After obtaining the modeling elements using the tester role model, the modeling elements are detected according to the detection rules.

4. A data management system for software development according to claim 3, characterized in that: The requirement document acquisition unit (3) also includes a missing requirement supplement module (303) and a document determination module (304). If the missing requirement supplement module (303) detects that there is missing information, the missing information will be fed back to the collector role model. The collector role model will inquire about the missing requirements in the form of a conversation, and after receiving the feedback, the requirement fragment will be generated again. The modeler role model will be used to extract new modeling elements, and the inspector role model will be re-tested. If the inspector role model detects that there is no missing information, the document determination module (304) will feed it back to the collector role model. The collector role model will determine the requirement document according to the requirement analysis template and the complete requirement description. At the same time, the modeler role model will determine the requirement analysis model through the modeling elements, and the requirement document and the data related to the requirement analysis model will be stored in the database.

5. A data management system for software development according to claim 1, characterized in that: The design data processing unit (4) comprises a data receiving module (401) and a text information extraction module (402). The data receiving module (401) obtains document data related to function description, table data related to project, image data related to program flow and function realization algorithm data, and then transmits these data to a storage device. The text information extraction module (402) uses a natural language processing tool to perform word segmentation processing on the document data related to function description in the data receiving module (401), determines the keywords of these documents through a keyword extraction algorithm, and then traverses the image data related to program flow through a file detection algorithm to obtain text information in the image.

Images