Student management method and system based on large language model

Abstract

The invention provides a student management method and system based on a large language model, and belongs to the technical field of large model and teaching management. The method is applied to a management terminal and comprises the steps that a voice instruction is sent to the management terminal, a large language model recognizes indication information contained in the voice instruction, and a corresponding data set is called based on the indication information; after attribute analysis is carried out on the corresponding data set, a plurality of selectable items corresponding to an attribute analysis result are provided on a human-computer interaction interface of the management terminal; and in response to the selection of at least one of the plurality of selectable items, displaying a visual student data analysis report on the human-computer interaction interface. The system comprises a plurality of management terminals and a cloud database communicating with the management terminals, the cloud database comprises a plurality of subclass databases, and the large language model provides a plurality of data query API; each data query AP I corresponds to at least one subclass database. According to the invention, related teaching management instructions can be accurately responded, and meanwhile, the data transmission overhead of a large model is reduced.

Description

Technical Field

[0001] The present invention belongs to the field of large models and teaching management technology, and in particular relates to a student management method and system based on a large language model. Background Art

[0002] In the modern education system, traditional student management methods play a fundamental role in implementing attendance, behavior records, score records, grade records, and report applications. However, they have also exposed many limitations under the wave of informatization. Traditional data recording methods rely on teachers' subjective judgments, and it is difficult to unify standards. Different teachers record different details and focuses. For example, some teachers focus on recording students' disciplinary violations and record less positive performance; some teachers record briefly, which makes it difficult to fully reflect the overall picture of students' behavior. At the same time, paper records are inconvenient to retrieve. When it is necessary to analyze students' long-term behavioral performance, it takes a lot of time to review historical records, making it difficult to conduct systematic analysis and intervention of student behavior.

[0003] On this basis, when teachers or administrators apply for student-related reports, the traditional process is complicated. Data needs to be collected from multiple paper files and different spreadsheets, and then manually organized into a report format. If you want to generate reports that include comprehensive information such as attendance, behavior, and grades, it involves the aggregation of data from multiple departments and multiple teachers. Coordination is difficult and data consistency is difficult to ensure. Moreover, manual report preparation is time-consuming and cannot meet the school's needs for real-time understanding of student conditions and rapid decision-making. When responding to emergencies (such as statistics on student health status during public emergencies), traditional report application and production methods are inefficient and difficult to play an effective role.

[0004] With the development of artificial intelligence technology, especially large-scale model technology, it has become possible to introduce large-scale models into teaching management. Among related technologies, Chinese invention publication CN120181654A proposes a multimodal classroom evaluation system based on large-scale models; CN120086242A proposes a method and system for querying course grades using text2sql, a large language model based on the LLM.

[0005] However, large models are usually accompanied by large data. When processing data, large models often show a pursuit of comprehensiveness, and their default strategy tends to retrieve as much data as possible from all databases accessed. This feature aims to provide the most detailed analysis results for users, whether facing complex academic research problems or daily work information queries, large models try to filter out the most relevant and valuable content from massive data; even if the user limits the database range or analysis range or analysis target, the data that the large model needs to load will still exceed the user's expectations, resulting in heavy data transmission burden. This excessive data loading causes a large amount of unnecessary data to be transmitted in the network, not only occupying valuable network bandwidth resources, but also significantly increasing data transmission delay; too heavy data transmission burden can also cause network congestion, especially when multiple users use large language model services at the same time, further reducing the overall system efficiency. SUMMARY

[0006] To solve the above technical problems, the present application provides a student management method and system based on a large language model.

[0007] In a first aspect of the present application, a student management method based on a large language model is provided, which is applied to at least one management terminal.

[0008] The method comprises:

[0009] issuing a voice instruction to the management terminal, the large language model identifying the indication information contained in the voice instruction, and retrieving a corresponding data set based on the indication information;

[0010] After attribute analysis of the corresponding data set, a plurality of selectable items corresponding to the attribute analysis result are provided on the human-computer interaction interface of the management terminal;

[0011] In response to at least one of the plurality of selectable items being selected, a visual student data analysis report is displayed on the human-computer interaction interface;

[0012] Wherein, the indication information is used to indicate at least one of the target student, the target time period, and the target management element, and the target management element includes at least one of the attendance management, the score management, the archive management, and the classroom discipline management.

[0013] The management terminal is connected with a cloud database, and the cloud database is used to save pre-collected student management data;

[0014] The pre-collected student management data includes:

[0015] Through the campus card system, classroom smart access control devices or mobile terminal clock-in applications, students' arrival and departure times and class attendance can be obtained in real time;

[0016] Connect data with the school's academic management system to regularly synchronize students' test scores, homework scores, and regular test scores;

[0017] Connect to the school's reward and punishment points system to automatically obtain the reward and punishment points data earned by students.

[0018] The cloud database includes multiple sub-category databases, and the large language model provides multiple data query APIs; each data query API corresponds to at least one sub-category database;

[0019] The large language model recognizes the instruction information contained in the voice command, and retrieves the corresponding data set based on the instruction information, specifically including:

[0020] The large language model determines at least one target data query API based on the indication information, connects to at least one subcategory database based on the at least one target data query API, and retrieves a corresponding data set from the at least one subcategory database.

[0021] The performing attribute analysis on the corresponding data set specifically includes:

[0022] Analyze the time range, update frequency, and visibility range of the corresponding data set;

[0023] A plurality of optional items corresponding to the attribute analysis result are provided on the human-computer interaction interface of the management terminal and are associated with the time range, update frequency and visible range.

[0024] The large language model provides multiple data query APIs, at least a first portion of the multiple data query APIs are dependent on each other, and at least a second portion of the data query APIs can be executed in parallel.

[0025] The large language model recognizes the instruction information contained in the voice command, and retrieves the corresponding data set based on the instruction information, specifically including:

[0026] When the large language model recognizes that the instruction information includes a dependency instruction relationship, determining a plurality of first target data query APIs from the at least first portion of the data query APIs; and sequentially executing the plurality of first target data query APIs based on the dependency relationship;

[0027] When the large language model recognizes that the instruction information does not include a dependency instruction relationship, a plurality of second target data query APIs are determined from the at least second portion of the data query APIs; and the second target data query APIs are executed in parallel.

[0028] During the parallel execution of the second target data query API, the API communication link between the large language model and the cloud database is kept alive; and during the sequential execution of the multiple first target data query APIs based on the dependency relationship, the API communication links between the multiple first target data query APIs and the cloud database are sequentially disconnected.

[0029] In a second aspect of the present invention, a student management system based on a large language model is proposed, the system comprising a plurality of management terminals and a cloud database communicating with the management terminals;

[0030] The cloud database is used to store pre-collected student management data;

[0031] Each of the management terminals can call the large language model;

[0032] The large language model provides multiple data query APIs; there is a dependency relationship between at least a first portion of the multiple data query APIs;

[0033] The management terminal receives a voice instruction, and the large language model recognizes instruction information contained in the voice instruction;

[0034] When the large language model recognizes that the instruction information includes a dependency instruction relationship, determining a plurality of first target data query APIs from the at least first portion of the data query APIs; and sequentially executing the plurality of first target data query APIs based on the dependency relationship to retrieve corresponding data sets.

[0035] After performing attribute analysis on the corresponding data set, a plurality of optional items corresponding to the attribute analysis results are provided on the human-computer interaction interface of the management terminal;

[0036] In response to at least one of the multiple optional items being selected, a visual student data analysis report is displayed on the human-computer interaction interface.

[0037] At least a second part of the data query APIs among the multiple data query APIs can be executed in parallel;

[0038] When the large language model recognizes that the indication information does not include a dependency instruction relationship, multiple second target data query APIs are determined from the at least second part of the data query APIs; and the second target data query APIs are executed in parallel to retrieve the corresponding data sets.

[0039] The cloud database includes multiple sub-category databases, and each data query API corresponds to at least one sub-category database;

[0040] The large language model queries the API based on the at least one target data and connects to at least one subcategory database, and retrieves a corresponding data set from the at least one subcategory database.

[0041] In the system architecture of this invention, a large language model works in conjunction with multiple data query APIs, each corresponding to a specific sub-category database, to precisely retrieve data sets and significantly reduce data transmission. Based on command analysis results, the system of this invention, based on a large language model, precisely locates the required information only through the APIs in the corresponding databases, reducing unnecessary data transmission, effectively alleviating network pressure, and improving system response speed. This advantage is particularly evident in campus network environments with limited network bandwidth and moderate management terminal hardware and software performance.

[0042] Further specific advantages and implementation principles of the present invention will be further embodied in detail in the specific embodiments section in conjunction with the drawings of the specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1 This is a schematic diagram of the main flow of a student management method based on a large language model according to an embodiment of the present invention;

[0045] Figure 2 Is implemented Figure 1 The method is a schematic diagram of a remote device for collecting relevant data;

[0046] Figure 3 yes Figure 1 Schematic diagram of the principle of the large model in the method to retrieve the corresponding data set;

[0047] Figure 4 The figure is a schematic diagram showing the hardware / software functional unit composition of a student management system based on a large language model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In the specific implementation of this application, if the embodiments of the relevant technical solutions involve user-related data, when the embodiments of this application are applied to specific products or technologies, user permission or consent must be obtained, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0049] See also Figure 1 , Figure 1 This is a schematic diagram of the main flow of a student management method based on a large language model according to an embodiment of the present invention.

[0050] Figure 1 The method can be applied to at least one management terminal. Figure 1 The method can be applied or implemented on a plurality of different management terminals with different data permissions.

[0051] The multiple different management terminals with different data permissions can be various intelligent electronic devices such as desktop terminals, mobile terminals, etc. An application program (APP) for teaching management can be installed on the electronic device.

[0052] Teaching management personnel of different levels / categories / identities have different data permissions. For example, the principal may have data access permissions for all school staff, the head teacher of a class may have data access permissions for all students in the class, and a subject teacher may have relevant subject data permissions for all students in the corresponding subject, etc.

[0053] Therefore, in the present invention, when describing a user's "identity", "level", "grade" and other qualifiers, it means that the user has the corresponding "authority". Therefore, "identity", "level", "grade" and "authority" have the same meaning in certain contexts.

[0054] For details, see Figure 1 , the management terminal is connected to a cloud database, and the cloud database is used to store pre-collected student management data; the management terminal can call a large language model;

[0055] The management terminal can call the large language model, including local call and cloud call;

[0056] Preferably, when the management terminal is a centralized control terminal at the server location, the management terminal can call the large language model based on a local call method; when the management terminal is an ordinary portable mobile terminal, the management terminal can call the large language model based on a cloud call method;

[0057] On this basis, Figure 1The method comprises the following steps (for the convenience of subsequent description, each step is numbered here, but the relevant step numbers are omitted in the accompanying drawings):

[0058] S1: issuing a voice command to the management terminal, the large language model identifying instruction information contained in the voice command, and retrieving a corresponding data set based on the instruction information;

[0059] S2: After performing attribute analysis on the corresponding data set, a plurality of options corresponding to the attribute analysis results are provided on the human-computer interaction interface of the management terminal;

[0060] S3: In response to at least one of the multiple optional items being selected, a visual student data analysis report is displayed on the human-computer interaction interface.

[0061] Next, steps S1-S3 are described in detail.

[0062] First, in step S1, the management terminal receives a voice command issued by the current user, and then uses the large language model to recognize the instruction information contained in the voice command;

[0063] The indication information is used to indicate at least one of a target student, a target time period, and a target management element, and the target management element includes at least one of attendance management, grade management, file management, and classroom discipline management.

[0064] The above-mentioned instruction information is generally the content indicated by the voice instruction;

[0065] As an example, the voice command may be "Query the student performance report for this semester";

[0066] In this example, the target time period is indicated as "this semester" and the target management element is indicated as "grade management";

[0067] However, as a further improvement of the present invention, the large language model also recognizes user identity information contained in the voice instruction.

[0068] The large language model is trained based on pre-collected user group data. Different users have different speech features. Based on the speech features, the identity of the user who makes the current speech can be identified, and then the data acquisition authority corresponding to the user can be determined.

[0069] Continuing with the above example, the large language model further identifies the user identity information contained in the voice command, for example, identifying that the current user is a math teacher of Class 3 (1) of Senior High School, and the user only has the authority to obtain data on the math scores of Class 3 (1);

[0070] Therefore, the voice instruction further indicates that the target students are "Senior Three (1) Class" and the target management element is "(Mathematics) Score Management".

[0071] On this basis, the large language model retrieves the corresponding data set based on the indication information, that is, it only collects the data set limited by key indication information such as "this semester" + "Senior Three (1) Class" + "Mathematics scores", instead of generally querying all student scores in this semester.

[0072] After step S1, step S2 is executed: after performing attribute analysis on the corresponding data set, a plurality of optional items corresponding to the attribute analysis results are provided on the human-computer interaction interface of the management terminal;

[0073] Specifically, the attribute analysis of the corresponding data set includes:

[0074] Analyze the time range, update frequency, and visibility range of the corresponding data set;

[0075] A plurality of optional items corresponding to the attribute analysis result are provided on the human-computer interaction interface of the management terminal and are associated with the time range, update frequency and visible range.

[0076] Continuing with the above example, after obtaining the data set corresponding to "Mathematics scores of students in Class 1 of Senior 3 this semester", it is analyzed that the time range of the corresponding data set is from September to December, and the update frequency is once a month (monthly test scores), so the visible scope is all students and teachers in this class;

[0077] At this time, multiple options corresponding to the attribute analysis results are provided on the human-computer interaction interface of the management terminal, including:

[0078] "Show a list of students ranked by grades and output the month with the best grades"

[0079] "View the details of students with large fluctuations in grades and output the monthly distribution"

[0080] "Analyze the distribution time of the highest average score"

[0081] …

[0082] Next, step S3 is executed: in response to at least one of the multiple options being selected, a visual student data analysis report is displayed on the human-computer interaction interface.

[0083] Continuing with the above example, when the user selects at least one of the multiple options, the management terminal human-computer interaction interface displays a visual student data analysis report.

[0084] For example, if the user selects "Display the list of students ranked by grades and output the month with the best grades", the interface will intuitively present the math grade ranking of the senior high school (1) class in a table, including the student's name, student ID, grade and ranking, the corresponding best month and other information; if the user selects "View the details of students with large grade fluctuations and output the monthly distribution", the grade curve of students with fluctuating grades will be displayed as a line graph, with detailed grade data attached, so that users can see the students' learning situation at a glance.

[0085] As you can see, Figure 1 The implementation scheme realizes accurate response to teaching management instructions, reduces the data transmission overhead of large models, improves teaching management efficiency and data processing accuracy, provides educators with convenient and efficient student management tools, and helps to make teaching decisions scientific and precise.

[0086] Figure 2 Is implemented Figure 1 The method is a schematic diagram of a remote device for collecting relevant data.

[0087] exist Figure 2 In the example, the cloud database is used to store pre-collected student management data;

[0088] In a specific teaching management scenario, the following data collection and transmission systems usually exist:

[0089] Campus Card System: Using the campus card held by students, the system records students' arrival and departure times in real time, including when they enter and exit the campus, consume in the cafeteria, and borrow books from the library. This data is then transmitted to a cloud database, providing basic data support for student attendance and other management tasks.

[0090] Classroom smart access control device: Installed at the classroom door, when students enter or leave the classroom, the device automatically identifies the student's identity, records class attendance, and transmits relevant data to the cloud database in real time, making it easier for the school to accurately manage student class attendance.

[0091] Mobile terminal clocking-in application: Students can use the clocking-in application on mobile devices such as mobile phones to clock in at the specified time and report their location and clocking-in time. This data will also be transmitted to the cloud database to supplement student attendance data, which is especially suitable for attendance records in some special scenarios.

[0092] Academic Management System: Regularly connects data with the cloud database to synchronize students' test scores, homework scores, regular test scores and other academic-related data to the cloud database, providing a data basis for comprehensive assessment of students' learning status.

[0093] Reward and Punishment Points System: Automatically transmits the reward points students receive for participating in activities and performing well in school, as well as the penalty points they receive for violations, to the cloud database, making it easier for schools to quantify students' overall performance.

[0094] Therefore, see Figure 2 , the pre-collected student management data includes:

[0095] Through the campus card system, classroom smart access control devices or mobile terminal clock-in applications, students' arrival and departure times and class attendance can be obtained in real time;

[0096] Connect data with the school's academic management system to regularly synchronize students' test scores, homework scores, and regular test scores;

[0097] Connect to the school's reward and punishment points system to automatically obtain the reward and punishment points data earned by students.

[0098] Next, further preferred embodiments of the present invention are described.

[0099] Figure 1 The cloud database includes multiple sub-category databases, and the large language model provides multiple data query APIs; each data query API corresponds to at least one sub-category database;

[0100] The large language model recognizes the instruction information contained in the voice command, and retrieves the corresponding data set based on the instruction information, specifically including:

[0101] The large language model determines at least one target data query API based on the indication information, connects to at least one subcategory database based on the at least one target data query API, and retrieves a corresponding data set from the at least one subcategory database.

[0102] In the preferred embodiment described above, one of the core advantages of applying a large language model to management terminals is the accurate and efficient retrieval of required data sets. The cloud database structure in the system is complex and large, consisting of multiple sub-databases. These sub-databases are divided based on data type and purpose, such as student basic information databases, course performance databases, and classroom behavior databases. These sub-databases store different types of data that are crucial for teaching management.

[0103] To accurately retrieve data from this complex database system, the large language model provides multiple data query APIs. Each data query API acts as a dedicated channel to a specific subcategory database, carefully designed to correspond to at least one subcategory database. This correspondence is not randomly assigned but is based on rigorous database architecture design and a deep understanding of the logic behind the invocation of various types of teaching management data. For example, an API for querying student basic information specifically connects to a subcategory database storing basic information such as name, age, gender, and student ID number. An API for querying course grades establishes a robust connection to a subcategory database storing data such as test scores, homework grades, and attendance records.

[0104] When the management terminal receives and transmits the voice command to the large language model, the model starts the key process of identifying the instruction information and retrieving the corresponding data set. In the process of identifying the instruction information contained in the voice command, the large language model uses advanced natural language processing technology to conduct a comprehensive analysis of the voice command, and can accurately capture the key elements in the command, such as time limit ("this semester", "last semester"), specific object ("Senior Three (1) Class", "Student with Student ID [X]"), data type ("Mathematics score", "Classroom violation record"), etc.

[0105] Based on the accurate recognition of the instruction information, the next step of the large language model is to determine at least one target data query API. This process is like finding the right path to the destination in a complex map. The model will perform intelligent matching based on the key elements in the instruction information and the pre-set API and sub-category database corresponding rules. For example, when the instruction is "query the math scores of students in Class 1 of Senior 3 this semester", the model recognizes key information such as "this semester", "Class 1 of Senior 3", and "mathematics scores", and quickly determines that the instruction involves a query of subject scores for a specific semester and a specific class, and then determines the corresponding score query API. This API is associated with the sub-category database that stores the math score data for the corresponding class and semester.

[0106] After determining the target data query API, the large language model establishes a connection with the corresponding sub-category database based on these APIs to ensure that data can be transmitted safely and efficiently between the model and the database. Through the established connection, the large language model retrieves the corresponding data set from at least one sub-category database. For example, in the above example, the model uses the connected score query API to accurately extract the required score data from the sub-category database that stores the math score data of Class 3 (1) of Senior 3 this semester. The entire process is fast and accurate, avoiding the loading of irrelevant data and greatly reducing the data transmission overhead. It lays a solid data foundation for subsequent attribute analysis of the data set, providing options, and displaying visual reports.

[0107] In actual application scenarios, voice commands may include multiple continuous query actions (whether explicit or implicit). To handle such situations, as a further preferred embodiment, the large language model provides multiple data query APIs, wherein at least a first portion of the multiple data query APIs have dependencies, and at least a second portion of the data query APIs can be executed in parallel.

[0108] See also Figure 3 , Figure 3 yes Figure 1 Schematic diagram of the principle of the large model in the method to retrieve the corresponding data set.

[0109] exist Figure 3 , the method comprising:

[0110] S310: Sending a voice command to the management terminal, and the large language model recognizing instruction information contained in the voice command;

[0111] S320: When the large language model recognizes that the instruction information includes a dependency instruction relationship, determining a plurality of first target data query APIs from the at least first portion of data query APIs;

[0112] executing the plurality of first target data query APIs in sequence based on the dependency relationship;

[0113] When the large language model identifies that the instruction information does not include a dependency instruction relationship, determining a plurality of second target data query APIs from the at least second portion of the data query APIs; and executing the second target data query APIs in parallel;

[0114] S330: Based on the first or second target data query API, connect to at least one subcategory database and retrieve a corresponding data set from the at least one subcategory database.

[0115] Specifically, in a preferred embodiment of the present invention, the internal architecture of the multiple data query APIs provided by the large language model is highly flexible and intelligent. These APIs are cleverly divided into different parts, with at least the first part of the data query APIs being closely interdependent. This dependency is not constructed arbitrarily, but rather based on the inherent logical connections between teaching management data. For example, when querying a student's comprehensive grade report for the current semester, it may involve first obtaining information about the student's selected courses from the course schedule database API. Based on this course information, the grade entry database API is then called to obtain the grade data corresponding to each course. Here, the course schedule query API and the grade query API constitute the first part of the interdependent API.

[0116] Correspondingly, at least the second set of data query APIs are designed to execute in parallel. The data retrieval tasks associated with these APIs are independent of each other, without any prioritization constraints. For example, when querying data related to student classroom performance, the API for querying student speech counts and the API for querying student attendance status correspond to different sub-databases (such as the classroom interaction records database and the attendance database). These two APIs have no dependencies and can execute simultaneously, significantly reducing data acquisition time.

[0117] When the management terminal receives a user voice command, if it recognizes that the instruction information contains a dependent instruction relationship, the model will quickly and accurately determine multiple first target data query APIs from at least the first part of the data query API. For example, when the voice command is "query the detailed score analysis of the mathematics course of Class 1, Senior 3 this semester, including the scores of each chapter", the model parses the command to find that the command involves first obtaining the class schedule for this semester to determine the relevant information of the mathematics course, and then obtaining the corresponding mathematics grades based on the course information, and then determining the course schedule query API, mathematics grade detail query API, etc. as the first target data query API.

[0118] After determining the first target data query API, the large language model executes it in an orderly manner based on the established dependencies between these APIs. In the above example, the model first calls the course schedule query API and obtains the detailed schedule information of the mathematics course of the senior high school class (1) this semester from the corresponding course schedule subcategory database, including the lecturer, class time, course chapter planning, etc. After obtaining this information, the model inputs it as a parameter into the mathematics score detail query API and retrieves the detailed score data of each chapter of the mathematics course of this semester for this class from the score subcategory database to ensure the accuracy and logic of the data retrieval.

[0119] When the large language model identifies that the instruction information does not contain a dependency instruction relationship, the processing flow is different. The model will determine multiple second target data query APIs from at least the second part of the data query API. Assuming that the voice instruction is "simultaneously query the class attendance record and class performance score of Zhang San, a student in Class 1 of Senior 3", the model determines that the attendance record query and the class performance score query are independent of each other and have no dependency relationship, and thus determines the attendance query API and the class performance score query API as the second target data query API.

[0120] The large language model then fully leverages the system's parallel computing resources to execute these secondary data query APIs in parallel. During execution, the attendance query API quickly connects to the attendance subcategory database to retrieve Zhang San's class attendance records, including attendance days, number of tardiness, and number of early departures. Simultaneously, the class performance score query API connects to the class performance database to retrieve Zhang San's class performance score data, including participation scores and speech quality scores. This parallel execution significantly shortens overall data acquisition time and improves system response efficiency.

[0121] Whether using the primary or secondary data query API, the large language model ultimately uses these APIs to establish a connection with at least one subcategory database and accurately retrieve the corresponding dataset from that subcategory. These datasets serve as the core data foundation for subsequent attribute analysis, providing options in the human-computer interaction interface, and displaying visual student data analysis reports. This provides strong data support for teaching management and enables efficient and intelligent teaching management processes.

[0122] To further save or fully utilize existing system resources while ensuring data security, during the parallel execution of the second target data query API, the API communication link between the large language model and the cloud database is kept active; and during the sequential execution of the multiple first target data query APIs based on the dependency relationship, the API communication links between the multiple first target data query APIs and the cloud database are sequentially disconnected.

[0123] Specifically, in the process of large language models working together with cloud databases, in order to further optimize the utilization of system resources and improve operational efficiency, different strategies are adopted to manage the API communication link with the cloud database when executing data query APIs.

[0124] When the large language model identifies that the instruction information does not contain a dependency relationship and determines multiple second target data query APIs from at least the second portion of the data query APIs for parallel execution, maintaining the communication link between the large language model and the cloud database API offers significant advantages. During this process, the communication link remains open, like multiple lanes on a busy highway, ensuring efficient and rapid data transmission.

[0125] From a resource utilization perspective, keepalive communication links avoid the overhead of frequent connection establishment and disconnection. Establishing a new API communication link requires a complex series of operations, such as network handshakes, authentication, and resource allocation. For example, establishing a new database connection can involve multiple network round trips, consuming significant time and computing resources. Furthermore, when executing multiple unrelated data query tasks in parallel, reestablishing a new connection each time is a significant waste of system resources.

[0126] From the perspective of data transmission efficiency, a keep-alive communication link helps maintain a stable data transmission rate. During parallel execution, multiple data query tasks can interact with the cloud database through this stable link at the same time, reducing data transmission delays caused by unstable connections or frequent connection switching. For example, when querying independent data such as students' classroom attendance records, homework completion status, and test scores, these query tasks can quickly obtain data from the cloud database at the same time through a keep-alive communication link, greatly shortening the overall data acquisition time, improving the system response speed, and providing strong support for subsequent data analysis and report presentation.

[0127] When the large language model identifies that the instruction information contains a dependency relationship, multiple first target data query APIs are determined from at least the first data query API, and these are executed sequentially based on the dependency relationship, then it is more sensible to sequentially disconnect these APIs from the cloud database API communication links. This is because the first data query APIs are dependent on each other; subsequent APIs can only continue to execute based on this data after the previous API completes and obtains the required data. For example, when querying a student's comprehensive grade report for this semester, it may be necessary to first obtain the student's selected courses from the course schedule database API. Based on this course information, the grade entry database API is then called to obtain the grade data corresponding to each course.

[0128] In this case, when the first API completes execution, its communication link with the cloud database is no longer needed for subsequent API execution. Disconnecting this link promptly frees up related system resources, such as network ports and memory space. Taking network ports as an example, each API communication link occupies a network port. If completed API links are not disconnected promptly, excessive port usage can lead to system resource constraints and even affect network communications for other applications. Disconnecting communication links sequentially avoids idle and wasted resources.

[0129] At the same time, sequentially disconnecting communication links helps improve the security and stability of the system. During data communication, maintaining unnecessary links for extended periods of time poses certain security risks, such as potential cyberattacks or data leaks. Promptly disconnecting links can mitigate these risks and ensure the security of data transmission between the cloud database and the large language model. Furthermore, reducing the number of unnecessary links can reduce system complexity, reduce potential points of failure, and improve overall system stability. This ensures reliable system operation when executing complex data query tasks with dependencies, providing accurate and stable data support for teaching management.

[0130] The above has detailed the implementation principles and improved advantages of multiple method embodiments. Figure 4 , Figure 4 A schematic diagram illustrating the composition of hardware / software functional units of a student management system based on a large language model according to an embodiment of the present invention.

[0131] It should be understood that after the principles and advantages of the method embodiment are described in detail, since the system embodiment basically corresponds to the method embodiment, the advantages and principles of the system embodiment do not need to be repeated.

[0132] Figure 4 A student management system based on a large language model is shown, the system comprising a plurality of management terminals and a cloud database communicating with the management terminals;

[0133] The cloud database is used to store pre-collected student management data;

[0134] Each of the management terminals can call the large language model;

[0135] The large language model provides multiple data query APIs; there is a dependency relationship between at least a first portion of the multiple data query APIs;

[0136] The management terminal receives a voice instruction, and the large language model recognizes instruction information contained in the voice instruction;

[0137] When the large language model recognizes that the instruction information includes a dependency instruction relationship, determining a plurality of first target data query APIs from the at least first portion of the data query APIs; and sequentially executing the plurality of first target data query APIs based on the dependency relationship to retrieve corresponding data sets.

[0138] After performing attribute analysis on the corresponding data set, a plurality of optional items corresponding to the attribute analysis results are provided on the human-computer interaction interface of the management terminal;

[0139] In response to at least one of the multiple optional items being selected, a visual student data analysis report is displayed on the human-computer interaction interface.

[0140] At least a second part of the data query APIs among the multiple data query APIs can be executed in parallel;

[0141] When the large language model recognizes that the indication information does not include a dependency instruction relationship, multiple second target data query APIs are determined from the at least second part of the data query APIs; and the second target data query APIs are executed in parallel to retrieve the corresponding data sets.

[0142] The cloud database includes multiple sub-category databases, and each data query API corresponds to at least one sub-category database;

[0143] The large language model queries the API based on the at least one target data and connects to at least one subcategory database, and retrieves a corresponding data set from the at least one subcategory database.

[0144] In the system architecture of this invention, a large language model works in conjunction with multiple data query APIs, each corresponding to a specific sub-category database, to precisely retrieve data sets and significantly reduce data transmission. Based on command analysis results, the system of this invention, based on a large language model, precisely locates the required information only through the APIs in the corresponding databases, reducing unnecessary data transmission, effectively alleviating network pressure, and improving system response speed. This advantage is particularly evident in campus network environments with limited network bandwidth and moderate management terminal hardware and software performance.

[0145] Although not shown in the drawings, preferably, further product embodiments may also be an electronic device comprising: a memory and one or more processors. The memory stores one or more application programs, and the one or more application programs are adapted to be used by the one or more processors to execute the method steps of the aforementioned method embodiment.

[0146] Although not shown in the drawings, more embodiments further include a non-transitory computer-readable storage medium storing a computer program. When the computer program is executed, the method steps of the aforementioned method embodiments are implemented.

[0147] It can be understood that the system, product, device, medium embodiments and method implementations correspond to each other and can reference each other. Their principles are similar or the same, so they will not be repeated.

[0148] For other technologies, principles, algorithms or models not elaborated in detail in this application, please refer to the existing technology.

[0149] The student management method and system technology proposed in this invention, based on a large language model and applied to management terminals, has many significant advantages and improvements, and is of great significance in improving teaching management efficiency and optimizing resource utilization, as shown in the following aspects:

[0150] (1) Efficient and accurate response to instructions

[0151] Leveraging its powerful natural language processing capabilities, the large language model can accurately identify voice commands received by management terminals. Compared to traditional speech recognition systems, it offers a deeper and more comprehensive understanding of natural language, unaffected by factors such as accent, speaking speed, and ambiguous expressions. For example, even when a teacher issues a command with a regional accent or imprecise wording, the model can accurately parse key information. For example, in a query like "Check the physics quiz scores for Class 2 and Class 3 last semester," the model can accurately extract key information such as "last semester," "Class 2 and Class 3," and "physics quiz scores," ensuring accurate interpretation of the command. This allows the system to quickly and accurately understand the teacher or administrator's intent, significantly improving the accuracy and efficiency of executing teaching management commands.

[0152] (2) Significantly reduce data transmission overhead

[0153] The large language model precisely connects to multiple sub-category databases in the cloud database through multiple data query APIs. When processing user instructions, the model accurately determines the target data query API based on the instructions in the instruction, and then retrieves the required data set from the corresponding sub-category database, avoiding the loading of irrelevant data. Compared with traditional systems that may retrieve data in full from the cloud database, the amount of data transmitted is significantly reduced. Actual tests have shown that when processing student management data requests of similar complexity, the amount of data transmitted can be reduced by 60%-80%, effectively alleviating network bandwidth pressure and improving system response speed. This advantage is particularly prominent in campus environments with limited network bandwidth.

[0154] (3) Optimize resource management during API execution

[0155] During the execution of data query APIs, this solution employs differentiated resource management strategies for different types of APIs. When executing a second target data query API in parallel, the API communication link between the large language model and the cloud database is maintained, avoiding the resource overhead associated with frequent connection establishment and disconnection. Establishing a new link requires complex operations such as network handshakes and identity authentication, consuming significant time and computing resources. Maintaining active links is like creating a stable multi-lane highway, allowing multiple query tasks to efficiently transmit data simultaneously, maintaining a stable transmission rate and significantly shortening overall data acquisition time. When executing the first target data query API sequentially based on dependency relationships, the communication links between these APIs and the cloud database are disconnected in turn, promptly releasing system resources such as network ports and memory space. This reduces security risks, improves system stability, avoids idle resources and waste, and ensures reliable system operation when executing complex dependent tasks.

[0156] (4) Provide powerful data analysis and decision support

[0157] The large language model performs multi-dimensional attribute analysis on the retrieved data set, providing users with rich and targeted options. Taking the student performance data set as an example, it can not only calculate conventional statistics such as average scores and rankings, but also conduct in-depth analysis of performance distribution, progress and regression trends, and scores of various question types. By selecting options such as "View details of students with significant performance improvement" and "Analyze the learning characteristics of students in each score range", users can obtain corresponding visual reports, such as a line graph of performance changes for students with improved performance and a pie chart of the proportion of students in each score range. Compared with traditional manual analysis or simple data statistics software, the large language model can process complex data relationships in a short period of time, provide a comprehensive and in-depth analytical perspective, and provide strong data support for teachers to adjust their teaching strategies and schools to make educational decisions, thereby promoting the transformation of teaching management from experience-driven to data-driven.

[0158] (5) Greatly enhance user interaction experience

[0159] From the moment a user issues a voice command to the moment they obtain a visual student data analysis report, the entire process is simple to operate and interactive. Teachers or administrators can complete data query and analysis simply by expressing their needs in everyday natural language without any complex operations. Compared to the tedious process of traditional student management data query, which requires switching between multiple system interfaces and manually entering complex query conditions, this solution greatly reduces the user threshold. For example, during a break between classes, teachers can easily query the class attendance status through voice, and instantly obtain a visual attendance report to understand the student attendance status. This eliminates the need to spend a lot of time in front of an office computer operating the software, which increases user enthusiasm and promotes the efficient implementation of teaching management work.

[0160] The foregoing has shown and described the method embodiments and system of the present invention, but it is understood by those skilled in the art that various changes, modifications, substitutions and variations may 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 student management method based on a large language model, the method being applied to at least one management terminal, characterized in that: The method comprises: Sending a voice command to the management terminal, the large language model recognizing instruction information contained in the voice command, and retrieving a corresponding data set based on the instruction information; After performing attribute analysis on the corresponding data set, a plurality of optional items corresponding to the attribute analysis results are provided on the human-computer interaction interface of the management terminal; In response to at least one of the multiple options being selected, displaying a visual student data analysis report on the human-computer interaction interface; The indication information is used to indicate at least one of a target student, a target time period, and a target management element, and the target management element includes at least one of attendance management, grade management, file management, and classroom discipline management.

2. A student management method based on a large language model as claimed in claim 1, characterized in that: The management terminal is connected to a cloud database, and the cloud database is used to store pre-collected student management data; The pre-collection of student management data includes: Through the campus card system, classroom smart access control devices or mobile terminal clock-in applications, students' arrival and departure times and class attendance can be obtained in real time; Connect data with the school's academic management system to regularly synchronize students' test scores, homework scores, and regular test scores; Connect to the school's reward and punishment points system to automatically obtain the reward and punishment points data earned by students.

3. A student management method based on a large language model as claimed in claim 2, characterized in that: The cloud database includes multiple sub-category databases, and the large language model provides multiple data query APIs; each data query API corresponds to at least one sub-category database; The large language model recognizes the instruction information contained in the voice command, and retrieves the corresponding data set based on the instruction information, specifically including: The large language model determines at least one target data query API based on the indication information, connects to at least one subcategory database based on the at least one target data query API, and retrieves a corresponding data set from the at least one subcategory database.

4. The student management method based on a large language model according to claim 1, characterized in that: The performing attribute analysis on the corresponding data set specifically includes: Analyze the time range, update frequency, and visibility range of the corresponding data set; A plurality of optional items corresponding to the attribute analysis result are provided on the human-computer interaction interface of the management terminal and are associated with the time range, update frequency and visible range.

5. The student management method based on a large language model according to claim 3, characterized in that: The large language model provides multiple data query APIs, at least a first portion of the multiple data query APIs are dependent on each other, and at least a second portion of the data query APIs can be executed in parallel.

6. A student management method based on a large language model as claimed in claim 5, characterized in that: The large language model recognizes the instruction information contained in the voice command, and retrieves the corresponding data set based on the instruction information, specifically including: When the large language model recognizes that the instruction information includes a dependency instruction relationship, determining a plurality of first target data query APIs from the at least first portion of the data query APIs; and sequentially executing the plurality of first target data query APIs based on the dependency relationship; When the large language model recognizes that the instruction information does not include a dependency instruction relationship, a plurality of second target data query APIs are determined from the at least second portion of the data query APIs; and the second target data query APIs are executed in parallel.

7. The student management method based on a large language model according to claim 5, characterized in that: During the parallel execution of the target data query API, maintaining the API communication link between the large language model and the cloud database; In the process of sequentially executing multiple target data query APIs based on the dependency relationship, the API communication links between the multiple target data query APIs and the cloud database are sequentially disconnected.

8. A student management system based on a large language model, the system comprising a plurality of management terminals and a cloud database communicating with the management terminals; Its characteristics are: The cloud database is used to store pre-collected student management data; Each of the management terminals can call the large language model; The large language model provides multiple data query APIs; There is a dependency relationship between at least a first part of the data query APIs among the multiple data query APIs; The management terminal receives a voice instruction, and the large language model recognizes instruction information contained in the voice instruction; When the large language model recognizes that the instruction information includes a dependency instruction relationship, determining a plurality of first target data query APIs from the at least first portion of the data query APIs; executing the plurality of first target data query APIs sequentially based on the dependency relationship to retrieve corresponding data sets; After performing attribute analysis on the corresponding data set, a plurality of optional items corresponding to the attribute analysis results are provided on the human-computer interaction interface of the management terminal; In response to at least one of the multiple optional items being selected, a visual student data analysis report is displayed on the human-computer interaction interface.

9. A student management system based on a large language model as claimed in claim 8, characterized in that: At least a second part of the data query APIs among the multiple data query APIs can be executed in parallel; When the large language model recognizes that the instruction information does not include a dependency instruction relationship, multiple second target data query APIs are determined from the at least second portion of the data query APIs; and the second target data query APIs are executed in parallel to retrieve the corresponding data sets.

10. A student management system based on a large language model as claimed in claim 8, characterized in that: The cloud database includes multiple sub-category databases, and each data query API corresponds to at least one sub-category database; The large language model queries the API based on the at least one target data and connects to at least one subcategory database, and retrieves a corresponding data set from the at least one subcategory database.

Images