Educational data analysis methods, devices and electronic equipment

Abstract

This invention relates to the field of artificial intelligence technology and discloses a method, apparatus, and electronic device for educational data analysis. The method includes: determining the target dimension corresponding to the educational management problem text based on textual information; constructing an intelligent agent matching the target dimension based on the target dimension; invoking the intelligent agent to execute multiple tasks, obtaining the task execution results for each task, with the multiple tasks used to analyze the educational management problem text; and generating an analysis report for the educational management problem text based on the fusion of the task execution results, the analysis report including the reasons for the formation of the educational management problem text and adjustment strategies. By implementing the method of this application, the problem of low accuracy in analysis reports generated by the artificial intelligence agent when the role and function of the educational management problem do not match the role and function of the artificial intelligence agent is solved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to educational data analysis methods, devices, and electronic equipment. Background Technology

[0002] With the advancement of educational informatization, universities have accumulated a large amount of data related to students, teachers, courses, research, and teaching quality. When utilizing this accumulated data and combining it with artificial intelligence to analyze educational management issues in universities, it is often necessary to conduct cross-dimensional analysis of these issues to obtain analytical reports.

[0003] Currently, AI agents are typically pre-defined with pre-defined roles and functions, and these pre-defined AI agents are used to analyze educational management issues in universities. However, if the educational management issues raised by educational administrators do not match the roles and functions of the AI ​​agent, the pre-defined AI agent will be unable to analyze the educational management issues effectively, resulting in low accuracy in the generated analysis reports. Summary of the Invention

[0004] This invention provides an educational data analysis method, apparatus, and electronic device to address the problem of low accuracy in analysis reports generated by AI agents when there is a mismatch between educational management issues and the roles and functions of AI agents.

[0005] In a first aspect, the present invention provides an educational data analysis method, which includes: determining the target dimension corresponding to the educational management problem text based on the text information corresponding to the educational management problem text; constructing an intelligent agent matching the target dimension based on the target dimension corresponding to the educational management problem text; invoking the intelligent agent to execute multiple tasks and obtaining the task execution results corresponding to each task, wherein the multiple tasks are used to analyze the educational management problem text; and generating an analysis report for the educational management problem text based on the fusion result of the execution results of each task, wherein the analysis report includes the reasons for the formation of the educational management problem text and adjustment strategies.

[0006] The educational data analysis method provided in this embodiment, based on the target dimensions extracted from educational management problems, can dynamically construct intelligent agents that match the target dimensions. Through the dynamically constructed intelligent agents, educational management problems can be analyzed more accurately and comprehensively, thereby obtaining more accurate analysis reports. This solves the problem that the pre-set intelligent agents' analytical capabilities are limited and cannot generate more accurate analysis reports due to the mismatch between the pre-set intelligent agents and the dimensions of educational management problems.

[0007] In one optional implementation, the target dimension corresponding to the educational management problem text is determined based on the text information corresponding to the educational management problem text, including: extracting multiple entity objects from the text information corresponding to the educational management problem text; and searching a preset dimension library based on each entity object to obtain the target dimension corresponding to each entity object. The preset dimension library is used to store the mapping relationship between entity objects and target dimensions.

[0008] By matching entity objects extracted from the text information corresponding to educational management issues with a preset dimension library, heterogeneous entity objects can be mapped to standard target dimensions. This ensures that the target dimensions corresponding to the same entity object are consistent, avoiding subsequent retrieval errors due to differences in expression, and thus affecting the success rate of intelligent agent construction.

[0009] In one optional implementation, the target dimension includes a business dimension and a constraint dimension. Based on the target dimension corresponding to the educational management problem text, an intelligent agent matching the target dimension is constructed, including: retrieving a function library based on the business dimension to obtain functional information matching the business dimension. The function library is used to store the mapping relationship between the business dimension and the functional information. The functional information includes data processing strategies, analysis and decision-making strategies, and domain knowledge. Based on the functional information and the constraint information corresponding to the constraint dimension, target prompt words are constructed. The constraint information includes subject information, time information, and data source information. A general intelligent agent base is obtained, and the target prompt words are configured as prompt words of the general intelligent agent base to obtain the intelligent agent.

[0010] Based on the business dimension, the role and function information corresponding to the agent to be generated can be quickly matched. Then, by combining the role and function information with the constraint information corresponding to the constraint dimension, the target prompt words corresponding to the agent to be generated can be quickly assembled. Then, by simply inputting the customized target prompt words into the general agent base, an agent with professional capabilities and execution boundaries can be quickly obtained. The above method can skip complex steps such as model training and fine-tuning, shortening the agent construction cycle.

[0011] In one optional implementation, there are multiple agents; invoking agents to execute multiple tasks and obtaining task execution results for each task includes: dividing tasks belonging to the same dependency level into the same group based on the dependency relationship between tasks, resulting in multiple groups; determining the target agent corresponding to each group from the multiple agents based on the task description of the tasks in each group; and executing the tasks in the corresponding group based on the target agent corresponding to each group, thereby obtaining task execution results for the tasks in each group.

[0012] By grouping tasks at the same level into the same group, it is possible to clearly determine which tasks can be started simultaneously and which tasks must wait for the preceding group to complete before they can be started, thus avoiding logical confusion caused by disordered scheduling of cross-level tasks from the root.

[0013] In one optional implementation, based on the task descriptions of the tasks in each group, the target intelligent agent corresponding to each group is determined from multiple intelligent agents, including: for each group, semantically matching the task descriptions corresponding to each task in the group with the functional information of all intelligent agents to determine the target intelligent agent corresponding to each task, wherein the functional information is configured as data processing strategy, analysis and decision-making strategy and domain knowledge; and configuring the target intelligent agent corresponding to each task as the target intelligent agent corresponding to the group.

[0014] By semantically matching the task descriptions for each task in the group with the role function information of all agents, the agent corresponding to each task can be accurately and quickly identified.

[0015] In one optional implementation, the task execution results for each group are obtained by executing the tasks within each group based on the target agent corresponding to each group. This includes: for each group, combining each task within the group with the target agent corresponding to the task to obtain multiple combinations for the group, where each combination includes a task and a target agent corresponding to the task; obtaining the information gain for each combination and determining the scheduling priority between the combinations based on the information gain; calling the target agent corresponding to the combination to execute the task corresponding to the combination based on the scheduling priority for each combination to obtain the initial task execution results for each task in each combination; and filtering the initial task execution results for each task in the group based on the confidence level corresponding to the initial task execution results to obtain the task execution results for each task in the group.

[0016] By using different target agents to execute the same task, the execution results for that task can be made more accurate. Subsequently, based on the execution results of each task, more precise analysis reports targeting educational management issues can be generated. Furthermore, information gain can be used to reasonably determine the execution priority among different combinations, ensuring that core tasks are executed first.

[0017] In an optional implementation, the method further includes: after obtaining the task execution results corresponding to each task, if the task execution results lack structured element information, then calling the target agent corresponding to the task execution results to re-execute the task corresponding to the task execution results to obtain a first new task execution result; and / or, if the confidence level of the task execution results is less than a preset confidence threshold, then calling the target agent corresponding to the task execution results to re-execute the task corresponding to the task execution results based on a backup analysis and decision strategy to obtain a second new task execution result; and / or, if there is a conflict between the task execution results and existing task execution results in the shared pool, then adjusting the education management problem to obtain a new education management problem, and continuing to determine the analysis report corresponding to the new education management problem.

[0018] When the task execution result lacks structured information, the target agent corresponding to the task execution result is invoked to re-execute the task corresponding to the task execution result, so that the first new task execution result is complete and correct, avoiding the incomplete task execution result, which in turn leads to an incomplete final analysis report.

[0019] When the confidence level of the task execution result is less than the preset confidence threshold, the target agent corresponding to the task execution result will re-execute the task based on the backup analysis and decision-making strategy. This ensures that the second new task execution result is accurate, and further ensures that the generated analysis report on education management issues is accurate.

[0020] If the task execution result conflicts with the existing task execution results in the shared pool, it indicates that the task execution result may have problems such as incorrect data source or oversight in calculation logic. By triggering adjustments to the education management issue, the output of conflicting task execution results can be avoided from the source, ensuring that the final generated task execution result for the education management issue is consistent with the global business logic.

[0021] In one optional implementation, the process of acquiring multiple tasks includes: based on the text information corresponding to the educational management problem text, splitting the total task of analyzing the educational management problem text into multiple initial tasks; based on the dependency relationship between the multiple initial tasks, determining the target initial tasks at the same dependency level and the semantic similarity between pairs of target initial tasks; merging two target initial tasks with a semantic similarity greater than a preset similarity threshold into one task to obtain multiple tasks.

[0022] For any two tasks at the same dependency level, the semantic similarity between the two tasks can be used to merge two similar tasks at the same dependency level with relatively accurate results, avoiding the occurrence of two similar tasks at the same dependency level, and thus avoiding the problem of wasted computing power caused by subsequent intelligent agents executing similar tasks.

[0023] Secondly, the present invention provides an educational data analysis device, comprising: a first determining module, used to determine the target dimension corresponding to the educational management problem text based on the text information corresponding to the educational management problem text; a constructing module, used to construct an intelligent agent matching the target dimension based on the target dimension corresponding to the educational management problem text; a first invoking module, used to invoke the intelligent agent to execute multiple tasks and obtain the task execution results corresponding to each task, the multiple tasks being used to analyze the educational management problem text; and a fusion module, used to generate an analysis report for the educational management problem text based on the fusion result of the execution results of each task, the analysis report including the formation reasons of the educational management problem text and adjustment strategies.

[0024] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the educational data analysis method of the first aspect or any corresponding embodiment described above.

[0025] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the educational data analysis method of the first aspect or any corresponding embodiment described above.

[0026] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the educational data analysis method of the first aspect or any corresponding embodiment described above. Attached Figure Description

[0027] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0028] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the first type of educational data analysis method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a second process for an educational data analysis method according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the process for detecting the results of task execution according to an embodiment of the present invention; Figure 5 This is a flowchart illustrating a specific educational data analysis method according to an embodiment of the present invention; Figure 6 This is a structural block diagram of an educational data analysis device according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0031] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0032] With the advancement of educational informatization, universities have accumulated a wealth of data related to students, teachers, courses, research, and teaching quality. This accumulated data comes from diverse sources, including academic affairs systems, quality evaluation systems, research management systems, and financial resource systems, and encompasses structured tables, semi-structured documents, and unstructured text. This data can be used to analyze educational management problems arising in the university's management process and to formulate improvement strategies to address these problems.

[0033] However, educational management issues cannot be comprehensively analyzed from a single dimension. For example, fluctuations in teaching quality may involve multiple factors such as changes in faculty structure, student learning characteristics, curriculum design, and assessment methods. Traditional statistical or visualization tools often only provide results from one dimension and cannot form a comprehensive cross-dimensional analysis and attribution. With the rapid development of artificial intelligence agents, the roles and functions of AI agents can be pre-defined, and these pre-defined AI agents can be used to analyze educational management issues in universities.

[0034] However, if the educational management issues raised by education administrators do not match the role and function of the AI ​​agent, the pre-defined AI agent will be unable to analyze the educational management issues effectively, resulting in low accuracy of the generated analysis reports.

[0035] In view of this, this application dynamically generates intelligent agents that match the dimensional information contained in the text of educational management issues. These agents are then invoked to execute multiple tasks for analyzing the text, yielding task execution results for each task. This achieves a relatively comprehensive and accurate analysis of the educational management issue text and generates precise analysis reports based on the task execution results. Furthermore, the dynamic generation of intelligent agents avoids limitations in their analytical capabilities. When universities add new business areas, only corresponding data interfaces need to be added to the underlying library, without requiring system architecture reconstruction. This significantly reduces maintenance costs and enables unlimited expansion of business capabilities.

[0036] As an optional application scenario of this invention, such as Figure 1 As shown, application 101 is installed in terminal device 110, and user 130 can interact with application 101 through terminal device 110 and / or access device of terminal device 110.

[0037] For example, application 101 can be any application that provides services related to educational data analysis. For instance, application 101 could be an educational data analysis application, such as a text-to-text application, a text-to-graphics application, etc. Figure 1 In the application scenario shown, if application 101 is active, the terminal device 110 can display the interface 102 of application 101. The interface 102 may include various pages that application 101 can provide, such as interactive pages, settings pages, query pages, etc.

[0038] In some embodiments, terminal device 110 is communicatively connected to server 120 to provide services to application 101. Terminal device 110 may be a mobile terminal, fixed terminal, or portable terminal, etc., including but not limited to mobile phones, desktop computers, laptop computers, multimedia tablets, e-book devices, gaming devices, or any combination thereof, including accessories and peripherals of these devices or any combination thereof. In some embodiments, terminal device 110 may also support any type of interface, and server 120 may be various types of computing systems or servers capable of providing computing power, including but not limited to mainframes, edge computing nodes, computing devices in cloud environments, etc.

[0039] It should be noted that, Figure 1This is merely an example of an application scenario and does not limit the scope of protection of this invention.

[0040] The embodiments of the present invention will now be described with reference to the accompanying drawings. It should be understood that the pages shown in the drawings are merely examples, and various page designs are possible in practice. The various graphic elements on the page may have different arrangements and different visual representations; one or more elements may be omitted or replaced, and one or more other elements may also be present, without any limitation in the embodiments of the present invention. Furthermore, the embodiments described below primarily pertain to terminal device 110. It should be understood that the actions described relative to terminal device 110 can be performed by application 101 on terminal device 110, or can be performed by application 101 in conjunction with its server (e.g., server 120).

[0041] According to an embodiment of the present invention, an embodiment of an educational data analysis method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0042] This embodiment provides an educational data analysis method that can be used in terminal devices. Figure 2 This is a flowchart of an educational data analysis method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Based on the text information corresponding to the educational management problem text, determine the target dimension corresponding to the educational management problem text.

[0043] As shown above, Figure 1 Application 101 can provide an interactive page through which users can input educational management question text into the terminal device. Correspondingly, the terminal device responds to the user's input and retrieves the educational management question text.

[0044] For example, a text on educational management issues could be: "Analyze the impact of the Computer Science Department's research funding investment on undergraduate teaching quality over the past three years, with data based on the Academic Affairs Office's data." Another example could be: "Analyze the reasons for the decline in teaching quality in the Humanities Department over the past three years and provide suggestions for improvement."

[0045] The target dimension can represent the structured features extracted from the text of an educational management issue. For a given educational management issue text, it can correspond to one target dimension or multiple target dimensions. Following the previous example, if the educational management issue text is: "Analyze the impact of the Computer Science Department's research funding investment on undergraduate teaching quality over the past three years, with data based on the Academic Affairs Office," then the target dimensions corresponding to this educational management issue text could include a business dimension represented by teaching quality, a subject dimension represented by the Computer Science Department, a time dimension represented by the past three years, and a data source dimension represented by the Academic Affairs Office.

[0046] As a concrete example, a pre-trained dimension recognition model can be used to process educational management problem texts to identify the target dimensions corresponding to the texts. This dimension recognition model can be trained based on a large language model architecture, a machine learning model architecture, or even a combination of multiple model architectures. No specific limitations are imposed here, as long as the model can identify the target dimensions corresponding to the educational management problem texts.

[0047] Step S202: Based on the target dimension corresponding to the educational management problem text, construct an intelligent agent that matches the target dimension.

[0048] This approach does not rely on weight adjustments, retraining, or parameter fine-tuning of the underlying large language model. Instead, it rapidly instantiates intelligent agents with specialized processing capabilities corresponding to the target dimension through a collaborative mechanism of dynamic configuration and prompt word engineering. For example, based on the target dimension corresponding to the educational management problem text, the capabilities of target external tools such as search engines, databases, and application programming interface calls, as well as target prompt words, can be determined. The target external tools are configured for the general intelligent agent base, and the target prompt words are configured as system prompt words for the general intelligent agent base, thereby obtaining an intelligent agent matching the target dimension. In this application, one target dimension can correspond to one intelligent agent; of course, multiple target dimensions can also correspond to one intelligent agent.

[0049] As a specific example, we can construct indicator agents corresponding to indicator dimensions, profile agents corresponding to profile dimensions, quality agents corresponding to quality dimensions, causal agents corresponding to causal dimensions, and report agents related to reporting dimensions.

[0050] Step S203: Invoke the intelligent agent to execute multiple tasks and obtain the task execution results corresponding to each task. The multiple tasks are used to analyze the text of educational management issues.

[0051] The multiple tasks here can be derived from the overall task of generating an analysis report corresponding to an educational management issue text. As a specific example, the educational management issue text can be segmented, its intent extracted, and key elements extracted to break down the generation of the analysis report into multiple tasks.

[0052] Following the previous example, the educational management issue text "Analyzing the reasons for the decline in teaching quality in the School of Humanities over the past three years and providing suggestions for improvement" can be broken down into the following five tasks: Task 1: Obtaining the trends of teaching quality indicators in the School of Humanities over the past three years; Task 2: Generating profiles of teachers, students, and courses in the School of Humanities and analyzing the possible influencing factors of quality fluctuations; Task 3: Conducting quality analysis on the pass rate and achievement rate of courses in the School of Humanities; Task 4: Conducting causal attribution for Tasks 1, 2, and 3 to identify the key factors leading to the fluctuations; Task 5: Combining the resource constraints information of the School of Humanities, generating improvement strategies and action suggestions.

[0053] As a concrete example, we can first determine the target agent corresponding to the task from the dynamically constructed agent based on the task description; then determine the execution priority of the task based on the dependencies between tasks; finally, call the target agent corresponding to the task to process the task according to the execution priority, thereby obtaining the task execution result.

[0054] Following the previous example, taking the five tasks corresponding to "analyzing the reasons for the decline in teaching quality in the School of Humanities over the past three years and providing suggestions for improvement" as an example, we set Task 1 to correspond to the first intelligent agent, Task 2 to the second intelligent agent, Task 3 to the third intelligent agent, Task 4 to the fourth intelligent agent, and Task 5 to the fifth intelligent agent. Moreover, we set that the execution of Task 4 depends on the execution results of Task 1, Task 2, and Task 3, and the execution of Task 5 depends on the execution result of Task 4. Therefore, the first, second, and third intelligent agents execute their respective tasks in parallel. Then, the fourth intelligent agent processes Task 4 based on the execution results of Task 1, Task 2, and Task 3 to generate the execution result of Task 4, and the fifth intelligent agent processes Task 5 based on the execution result of Task 4 to generate the execution result of Task 5.

[0055] Step S204: Based on the fusion results of the execution results of each task, generate an analysis report for the text on educational management issues. The analysis report includes the reasons for the formation of the text on educational management issues and adjustment strategies.

[0056] Here, a pre-trained task execution result fusion model can be used to generate an analysis report on texts related to educational management issues based on the results of each task execution. The training process of the task execution result fusion model is similar to that of the dimensionality recognition model, and will not be repeated here.

[0057] Following the previous example, a task execution result fusion model can be used to fuse the task execution results corresponding to Task 1, Task 2, Task 3, Task 4, and Task 5 to generate an analysis report on texts related to educational management issues. Alternatively, a dedicated intelligent agent can be configured to fuse the task execution results corresponding to Task 1, Task 2, Task 3, Task 4, and Task 5 to generate an analysis report on texts related to educational management issues.

[0058] In generating the analysis report, the effect size analysis of the influencing factors in the task execution results corresponding to each task can be performed first. Statistical methods or causal inference models can be used to quantify the degree of influence of each factor, providing a classification of strong and weak influencing factors to understand the causes of the educational management problem text. Further evaluation of the potential improvement after adjustment or intervention of different influencing factors yields candidate adjustment strategies. Then, resource constraints at the university level, such as budget, faculty staffing, course offering cycles, and venue capacity, are introduced to verify the feasibility of each candidate strategy under existing constraints. Solutions exceeding resource limitations are eliminated to obtain the final adjustment strategy. Additionally, the adjustment strategy can be generated in a structured manner, including: problem background, influencing factors, causal chain, candidate strategies, priorities, and feasibility explanation.

[0059] When an analysis report corresponding to an educational management problem text proves to have long-term value through practice, this report can be archived as a historical case study of educational management problem texts. When faced with similar educational management problem texts in the future, educational administrators can directly access these historical cases, reducing redundant calculations and manual analysis, and improving decision-making efficiency and consistency. Through this mechanism, this application not only answers the question "What are the causes of educational management problem texts?" but also further addresses the key question of "What measures should be taken?", achieving a complete closed loop from data analysis to strategy adjustment.

[0060] As a concrete example, analysis reports corresponding to educational management issues can be displayed through dashboards, interactive charts, and reports, making it easier for managers to understand them quickly.

[0061] The educational data analysis method provided in this embodiment, based on the target dimensions extracted from educational management problem texts, can dynamically construct intelligent agents that match the target dimensions. The dynamically constructed intelligent agents can analyze educational management problem texts more accurately and comprehensively, thereby obtaining more accurate analysis reports. This solves the problem that the pre-set intelligent agents' analytical capabilities are limited and cannot generate more accurate analysis reports due to the mismatch between the pre-set intelligent agents and the dimensions of the educational management problem texts.

[0062] This embodiment provides an educational data analysis method that can be used in terminal devices. Figure 3 This is a flowchart of an educational data analysis method according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps: Step S301: Based on the text information corresponding to the educational management problem text, determine the target dimension corresponding to the educational management problem text.

[0063] Specifically, step S301 includes: Step S3011: Extract multiple entity objects from the text information corresponding to the educational management problem text.

[0064] Entities can be people, events, things, indicators, or concepts with clear educational management business meanings involved in the text information corresponding to the educational management problem text; they can be the core elements constituting the educational management problem text. As a specific example, an entity recognition model can be used to process the text information corresponding to the educational management problem text to obtain multiple entity objects corresponding to the educational management problem text.

[0065] Following the previous example, taking "Analyzing the impact of the Computer Science Department's research funding investment on undergraduate teaching quality in the past three years, with data based on the Academic Affairs Office" as an example, the entity objects corresponding to this educational management issue text can be "Computer Science Department", "Research Funding", "Teaching Quality", "Past Three Years" and "Academic Affairs Office".

[0066] Step S3012: Based on each entity object, the preset dimension library is searched to obtain the target dimension corresponding to each entity object. The preset dimension library is used to store the mapping relationship between entity objects and target dimensions.

[0067] For example, the entity objects corresponding to the business dimension can be scientific research, finance, academic affairs, student affairs, assets, etc.; the entity objects corresponding to the subject dimension can be the whole school, college, major, class, and individual, etc.; the entity objects corresponding to the time dimension can be the past three years, semester, month, etc.; the entity objects corresponding to the data source dimension can be the academic affairs office, etc.

[0068] Since the preset dimension library stores the mapping relationship between entity objects and target dimensions, the target dimension corresponding to the entity object can be obtained by querying the preset dimension library through the entity object.

[0069] By matching entity objects extracted from the text information corresponding to educational management issues with a preset dimension library, heterogeneous entity objects can be mapped to standard target dimensions. This ensures that the target dimensions corresponding to the same entity object are consistent, avoiding subsequent retrieval errors due to differences in expression, and thus affecting the success rate of intelligent agent construction.

[0070] Step S302: Based on the target dimension corresponding to the educational management problem text, construct an intelligent agent that matches the target dimension.

[0071] As a specific example, the target dimension can include the business dimension and the constraint dimension. Among them, the constraint dimension can be the subject dimension, time dimension, and data source dimension as mentioned above.

[0072] Specifically, step S302 includes: Step S3021: Retrieve the function library based on the business dimension to obtain the function information that matches the business dimension. The function library is used to store the mapping relationship between the business dimension and the function information. The function information includes data processing strategies, analysis and decision-making strategies, and domain knowledge.

[0073] The function library can be a fine-grained, modular library of capability atoms maintained at the underlying level. Specifically, it can include tool atoms (i.e., data processing strategies), algorithm atoms (i.e., analysis and decision-making strategies), and knowledge atoms (i.e., domain knowledge). Each tool in the tool atom (i.e., data processing strategy) is bound to a corresponding prompt word micro-instruction; each algorithm in the algorithm atom is bound to a corresponding prompt word micro-instruction; and each piece of knowledge in the knowledge atom is bound to a corresponding prompt word micro-instruction. As a concrete example, the tools in the tool atom can be SQL generators, Excel parsers, and OCR recognizers. For instance, the prompt word micro-instruction bound to an SQL generator could be "convert natural language to Hive SQL." The algorithms in the algorithm atom can be trend prediction, attribution inference, and anomaly detection. For instance, the prompt word micro-instruction bound to trend prediction could be "analyze growth rates using time series models." The knowledge in the knowledge atom can be a dictionary of financial indicators, teaching evaluation standards, etc.

[0074] By searching the function library based on the business dimension, we can obtain data analysis strategies (tool atoms), analysis and decision-making strategies (algorithm atoms), and domain knowledge (knowledge atoms) that match the business dimension, thereby obtaining the functional information corresponding to the business dimension.

[0075] Step S3022: Based on the functional information and the constraint information corresponding to the constraint dimension, construct the target prompt words. The constraint information includes subject information, time information, and data source information.

[0076] The main information here can be the information corresponding to the main dimension shown above, the time information can be the information corresponding to the time dimension shown above, and the data source information can be the information of the data source dimension shown above.

[0077] Here, we can first obtain the prompt word template; then, fill the prompt word template with the functional information and the constraint information corresponding to the constraint dimension to obtain the target prompt word. As a specific example, the prompt word template can include a Role Definition field, a Skill Set field, a First Constraint field, and a Second Constraint field. The Role Definition field is used to fill in the role information of the dynamically generated agent; the Skill Set field is used to fill in the functional information shown above, namely, the tool atoms extracted from the atom library and their bound prompt word micro-instructions, the algorithm atoms and their bound prompt word micro-instructions, and the knowledge atoms and their bound prompt word micro-instructions; the First Constraint field is used to fill in the subject information corresponding to the subject dimension and the time information corresponding to the time dimension; the Second Constraint field is used to fill in the data source information corresponding to the data source dimension. Among them, the role information corresponding to the Role Definition field can be obtained by summarizing the information corresponding to the Skill Set field, the First Constraint field, and the Second Constraint field.

[0078] Step S3023: Obtain the general intelligent agent base and configure the target prompt word as the prompt word of the general intelligent agent base to obtain the intelligent agent.

[0079] Prompt words serve as the interaction interface between the user and the agent. They can convey information such as "what to do" and "how to do it" to the agent through natural language, structured descriptions, or examples. This helps to bridge the gap between the generality of model training data and the personalization of user needs, guiding the agent to activate corresponding knowledge and capabilities from massive training parameters.

[0080] As a specific example, the educational management problem text is set as "analyze the trend of scientific research funding investment." Entity recognition and mapping are performed on the text information corresponding to the educational management problem text to obtain business dimensions including "financial" and "trend." Based on "financial" and "trend," a function library is searched to obtain a financial database reading tool and a linear regression algorithm. Then, the financial database reading tool and its bound prompt word micro-instructions, and the linear regression algorithm and its bound prompt word micro-instructions are filled into the prompt word template to obtain the target prompt word. The target prompt word is configured as the prompt word of the general intelligent agent base, and then an intelligent agent with financial data reading permissions and focused on trend analysis is instantiated.

[0081] Based on the business dimension, the functional information corresponding to the agent to be generated can be quickly matched. Then, by combining the functional information with the constraint information corresponding to the constraint dimension, the target prompt words corresponding to the agent to be generated can be quickly assembled. Then, by simply inputting the customized target prompt words into the general agent base, an agent with professional capabilities and execution boundaries can be quickly obtained. The above method can skip complex steps such as model training and fine-tuning, shortening the agent construction cycle.

[0082] Step S303: Invoke the intelligent agent to execute multiple tasks and obtain the task execution results corresponding to each task. The multiple tasks are used to analyze the text of educational management issues.

[0083] Specifically, step S303 includes: Step S3031: Based on the dependencies between tasks, tasks belonging to the same dependency level are grouped into the same group, resulting in multiple groups. Please refer to the previous text for details on dependency order and dependency levels.

[0084] Following the previous example, the execution of Task 4 depends on the execution results of Tasks 1, 2, and 3, and the execution of Task 5 depends on the execution result of Task 4. Therefore, Tasks 1, 2, and 3 are at the same dependency level, as are Task 4 and Task 5. Accordingly, Tasks 1, 2, and 3 can be grouped into one group, Task 4 into another, and Task 5 into yet another. That is, a group can contain one task or multiple tasks.

[0085] Step S3032: Based on the task descriptions of the tasks in each group, determine the target agent corresponding to each group from multiple agents.

[0086] As mentioned above, there can be multiple dynamically constructed intelligent agents, and different intelligent agents have different functions. Therefore, by using the task descriptions of the tasks in each group, the target intelligent agent for each group can be selected from the multiple dynamically constructed intelligent agents. That is, the target intelligent agent is the intelligent agent among the multiple intelligent agents.

[0087] In some optional implementations, step S3032 above includes: Step a1: For each group, semantically match the task descriptions corresponding to each task in the group with the functional information of each agent to determine the target agent corresponding to each task. The functional information includes data processing strategy, analysis and decision-making strategy, and domain knowledge.

[0088] Step a2: Configure the target agent corresponding to each task as the target agent corresponding to the group.

[0089] As mentioned earlier, a group can contain multiple tasks, and these tasks may belong to the same dependency level but are not identical to each other. Therefore, different tasks correspond to different target agents. Thus, the target agent for a group can be the target agent corresponding to each task within that group.

[0090] In determining the target agent for each task within a group, the task description can be vectorized to obtain a task text vector. Furthermore, the functional information of all dynamically constructed agents can be vectorized to obtain multiple functional text vectors. The cosine or Euclidean distance between the task text vector and each functional text vector can then be calculated. Agents whose cosine or Euclidean distance meets a preset screening threshold are then selected as the target agents for that task. Therefore, a task can have one or more target agents.

[0091] Following the previous example, for Task 1: Obtain the teaching quality indicator trends of the School of Humanities over the past three years, the corresponding target intelligent agent can be the indicator intelligent agent; for Task 2: Generate profiles of teachers, students, and courses in the School of Humanities and analyze the possible influencing factors of quality fluctuations, the corresponding target intelligent agent can be the profile intelligent agent; for Task 3: Conduct quality analysis on the course pass rate and achievement rate of the School of Humanities, the corresponding target intelligent agent can be the quality intelligent agent; for Task 4: Conduct causal attribution on Tasks 1, 2, and 3 to identify the key factors causing fluctuations, the corresponding target intelligent agent can be the causal intelligent agent; for Task 5: Combine the resource constraint information of the School of Humanities to generate improvement strategies and action suggestions, the corresponding target intelligent agent can be the report generation intelligent agent.

[0092] By semantically matching the task descriptions for each task in the group with the functional information of all agents, the agent corresponding to each task can be accurately and quickly identified.

[0093] Step S3033: Based on the target agent of each group, execute the task in the corresponding group to obtain the task execution result of each group.

[0094] As a specific example, the execution priority of a group can be determined based on its dependency hierarchy, and the target agent of each group can be called to execute the corresponding task according to the execution priority, thereby obtaining the task execution result of each group.

[0095] Following the previous example, tasks 1, 2, and 3 correspond to a group, called Group 1; task 4 corresponds to a group, called Group 2; and task 5 corresponds to a group, called Group 3. Clearly, the execution priority among Group 1, Group 2, and Group 3 can be: Group 1 is executed first, then Group 2, and finally Group 3.

[0096] In some optional implementations, step S3033 above includes: Step b1: For each group, combine each task in the group with the target agent corresponding to the task to obtain multiple combinations for the group. Each combination includes a task and a target agent corresponding to the task.

[0097] Step b2: Obtain the information gain corresponding to each combination, and determine the scheduling priority between each combination based on the information gain.

[0098] Step b3: Based on the scheduling priority of each combination, call the target agent in the combination to execute the task in the combination, and obtain the initial task execution result of each task in the combination.

[0099] Step b4: Based on the confidence level corresponding to the initial task execution results, filter the initial task execution results corresponding to each task in the group to obtain the task execution results corresponding to each task in the group.

[0100] As mentioned earlier, a task can correspond to multiple target agents. Therefore, the task can be combined with each target agent to obtain multiple combinations. For example, a task may correspond to a first target agent, a second target agent, and a third target agent. Thus, the task can be combined with the first target agent, the second target agent, and the third target agent to obtain a first combination including the task and the first target agent, a second combination including the task and the second target agent, and a third combination including the task and the third target agent.

[0101] For example, a group may include three tasks: Task 1, Task 2, and Task 3. Task 1 corresponds to one target agent, Task 2 corresponds to two target agents, and Task 3 corresponds to three target agents. Therefore, this group can have 6 possible combinations.

[0102] Information gain can be used to measure the effective information increment and value contribution brought to the analysis of the entire educational management problem text after the corresponding tasks are performed. As a specific example, the information gain corresponding to each combination can be calculated using the following formula.

[0103]

[0104] in, Used to represent the information gain corresponding to the t-th combination. This is used to represent the estimated confidence level corresponding to the t-th combination. The estimated confidence level can be the probability, pre-generated based on the matching degree between the task corresponding to the t-th combination and the target agent corresponding to the t-th combination, that the target agent will successfully complete the task corresponding to the t-th combination and produce a valid result. This is used to represent the weight corresponding to the t-th combination, and the weight can be determined based on the task corresponding to the t-th combination. This is used to represent the j-th specific data element produced by the task corresponding to the t-th combination. This is used to represent the set of all data elements involved in the task corresponding to the t-th combination. This represents the number of downstream tasks that can be unlocked directly upon completion of the task in the t-th combination. Used to represent the unlock reward coefficient.

[0105] For a given combination, after the tasks within that combination are executed by the target agent within that combination, an initial task execution result is obtained corresponding to the tasks in that combination. As mentioned earlier, a task can correspond to multiple combinations, and correspondingly, a task can correspond to multiple initial task execution results. Therefore, the final task execution result corresponding to the task can be selected from multiple initial task execution results. As a specific example, after the target agent in the combination is executed by the tasks in that combination, the target agent not only returns the initial task execution result corresponding to that task, but also returns the confidence level corresponding to that initial task execution result; therefore, the final task execution result corresponding to the task can be selected from multiple initial task execution results based on the relationship between the confidence levels corresponding to the initial task execution results.

[0106] By using different target agents to perform the same task, the execution results for that task can be made more accurate. Subsequently, based on the execution results of each task, a more precise analysis report targeting educational management issues can be generated. Furthermore, information gain can be used to reasonably determine the execution priority among different combinations, ensuring that core tasks are executed first.

[0107] In some alternative implementations, the process of acquiring multiple tasks includes: Step c1: Based on the text information corresponding to the educational management problem text, the overall task of analyzing the educational management problem text is broken down into multiple initial tasks.

[0108] Step c2: Based on the dependencies between multiple initial tasks, determine the target initial tasks at the same dependency level and the semantic similarity between any two target initial tasks.

[0109] Step c3: Merge two initial target tasks with semantic similarity greater than a preset similarity threshold into one task, resulting in multiple tasks.

[0110] As a concrete example, we can first extract multiple entity objects from the text information corresponding to the educational management problem text, and then combine these entity objects to break down the overall task of generating the educational management problem text into multiple tasks. Following the previous example, we first extract the entity objects from "analyze the reasons for the decline in teaching quality of the School of Humanities in the past three years and suggestions for improvement," such as School of Humanities, past three years, teaching quality, reasons for decline, and suggestions for improvement. Then, based on School of Humanities, past three years, and teaching quality, we obtain Task 1: Obtain the trend of teaching quality indicators of the School of Humanities in the past three years; Task 2: Generate profiles of teachers, students, and courses in the School of Humanities, and analyze the possible influencing factors of quality fluctuations; Task 3: Conduct quality analysis of the course pass rate and achievement rate of the School of Humanities; and then, combining School of Humanities, past three years, teaching quality, reasons for decline, and suggestions for improvement, we obtain Task 4: Conduct causal attribution on Task 1, Task 2, and Task 3 to identify the key factors leading to the fluctuations; and Task 5: Combine the resource constraint information of the School of Humanities to generate improvement strategies and action suggestions.

[0111] Of course, based on the target dimension corresponding to the educational management problem text, the overall task of generating an analysis report corresponding to the educational management problem text can be broken down into multiple tasks.

[0112] The dependency hierarchy here can be the order of execution of the initial tasks. It is used to distinguish which tasks can be carried out in parallel and which tasks must be started after the preceding tasks are completed. Moreover, tasks at the same dependency hierarchy have no sequential dependency relationship with each other and can be executed synchronously.

[0113] Following the previous example, the execution of task 4 depends on the execution results of tasks 1, 2, and 3, and the execution of task 5 depends on the execution result of task 4. Therefore, tasks 1, 2, and 3 are in the same dependency level, task 4 is in the same dependency level, and task 5 is in the same dependency level.

[0114] The specific process for calculating the semantic similarity between two initial tasks can be as follows: first, the task descriptions of the two initial tasks are vectorized to obtain text vectors corresponding to the two initial tasks; then, the cosine distance or Euclidean distance between the two text vectors is calculated to obtain the semantic similarity between the two initial tasks. Furthermore, the specific vectorization method used in this application can be any suitable vectorization method, and no specific limitation is made to the vectorization method here.

[0115] As a specific example, the preset similarity threshold can be 0.95. Of course, the preset similarity threshold is not limited to 0.95. The specific value of the preset similarity threshold can be flexibly set according to the actual situation. This application does not make specific limitations on this.

[0116] For any two tasks at the same dependency level, the semantic similarity between the two tasks can be used to merge two similar tasks at the same dependency level with relatively accurate results, avoiding the occurrence of two similar tasks at the same dependency level, and thus avoiding the problem of wasted computing power caused by subsequent intelligent agents executing similar tasks.

[0117] In some alternative implementations, the method further includes: Step d1: After obtaining the task execution results corresponding to each task, if the task execution results lack structured element information, the target agent corresponding to the task execution results is called to re-execute the task corresponding to the task execution results to obtain the first new task execution results.

[0118] As a concrete example, the target agent can be configured to output the task execution results in a structured form. For example, the task execution results output in a structured form may include the following structured elements: indicators such as name, definition, and calculation formula; evidence such as data source, SQL query, and original snapshot; conclusions such as quantitative results or qualitative judgments; confidence level; and reproducibility information such as SQL statement, data table name, and time range.

[0119] After obtaining the task execution result, an integrity check can be performed on the result to determine if it lacks any of the aforementioned structured information. If any one of these structured information elements is missing, the task execution result fails the integrity check. In this case, the target agent corresponding to the task execution result can be invoked to re-execute the task, resulting in a new, first-level task execution result.

[0120] When the task execution result lacks structured information, the target agent corresponding to the task execution result is invoked to re-execute the task corresponding to the task execution result, so that the first new task execution result is complete and correct, avoiding the incomplete task execution result, which in turn leads to an incomplete final analysis report.

[0121] In some alternative implementations, the method further includes: Step e1: After obtaining the task execution results corresponding to each task, if the confidence level of the task execution result is less than the preset confidence threshold, the target agent corresponding to the task execution result is called to re-execute the task corresponding to the task execution result based on the backup analysis and decision strategy to obtain the second new task execution result.

[0122] The alternative analysis decision here can be one of the algorithm atoms shown above. Please refer to the previous text for details on the algorithm atoms.

[0123] When the confidence level of the task execution result is less than the preset confidence threshold, the target agent corresponding to the task execution result re-executes the task based on the backup analysis and decision strategy. This ensures that the second new task execution result is accurate, and further ensures that the generated analysis report on educational management issues is accurate.

[0124] In some alternative implementations, the method further includes: Step f1: After obtaining the task execution results corresponding to each task, if there is a conflict between the task execution results and the existing task execution results in the shared pool, the education management problem text is adjusted to obtain a new education management problem text, and the analysis report corresponding to the new education management problem text is determined.

[0125] If a conflict exists between the task execution result and existing task execution results in the shared pool (for example, the newly calculated "student-teacher ratio" deviates from the "historical student-teacher ratio" in the shared pool by more than a threshold), it indicates that the task execution result may have problems such as incorrect data source or oversight in calculation logic. By triggering adjustments to the education management issue text, conflicting task execution results can be avoided from the source, ensuring that the final generated task execution result for the education management issue text remains consistent in the global business logic.

[0126] As a specific example, such as Figure 4 As shown, taking a task as an example, the process of detecting the task execution result may include steps S401 to S408. Among them, Step S401: Obtain the task execution result corresponding to the task.

[0127] Step S402: Perform an integrity check on the task execution result to determine whether the task execution result is complete. If the task execution result is complete, proceed to steps S404 to S408; if the task execution result is incomplete, proceed to step S403.

[0128] Step S403: Execute the missing certificate loop, that is, call the target agent corresponding to the task execution result to re-execute the task corresponding to the task execution result, and obtain the first new task execution result.

[0129] Step S404: Determine if there is a conflict between the task execution result and the existing task execution results in the shared pool. If there is a conflict, proceed to step S405; if there is no conflict, proceed to steps S406 to S408.

[0130] Step S405: Execute the conflict loop, that is, clarify the educational management problem text and continue to determine the analysis report corresponding to the new educational management problem text.

[0131] Step S406: Determine whether the confidence level of the task execution result is higher than a preset confidence threshold. If the confidence level of the task execution result is higher than the preset confidence threshold, proceed to step S408; if the confidence level of the task execution result is lower than the preset confidence threshold, proceed to step S407.

[0132] Step S407: Execute the low-confidence loop, that is, call the target agent corresponding to the task execution result to re-execute the task corresponding to the task execution result based on the backup analysis and decision strategy to obtain the second new task execution result.

[0133] Step S408: Store the task execution result and transmit the task execution result to the downstream intelligent agent.

[0134] Step S304: Based on the fusion results of the various task executions, generate an analysis report for the educational management problem texts. The analysis report includes the causes of the educational management problem texts and adjustment strategies. For details, please refer to [link to relevant documentation]. Figure 2 Step S204 of the illustrated embodiment will not be described again here.

[0135] The educational data analysis method provided in this embodiment can clearly determine which tasks can be started simultaneously and which tasks must wait for the preceding group to complete before they can be started, by dividing tasks at the same level into the same group, thus avoiding logical confusion caused by disordered scheduling of cross-level tasks from the root.

[0136] As a specific application embodiment of the present invention, such as Figure 5The diagram shown is a flowchart of a specific educational data analysis method, specifically including steps S501 to S506. Among them, Step S501: The user inputs the text of the education management question into the terminal device.

[0137] Step S502: Based on the text information corresponding to the educational management problem text, determine the target dimension corresponding to the educational management problem text; based on the target dimension corresponding to the educational management problem text, dynamically construct an intelligent agent that matches the target dimension.

[0138] Step S503: Decompose the educational management problem text into multiple initial tasks, merge the tasks based on semantic similarity, and finally divide the multiple tasks into multiple groups based on the dependencies between the tasks.

[0139] Step S504: Based on semantic similarity, match the target agent corresponding to each group.

[0140] Step S505: Utilize the target agent corresponding to each group to execute the tasks in the respective group, thereby obtaining the task execution results for all tasks.

[0141] Step S506: Merge the task execution results of all tasks to generate an analysis report corresponding to the educational management problem text.

[0142] This embodiment also provides an educational data analysis device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0143] This embodiment provides an educational data analysis device, such as... Figure 6 As shown, it includes: The first determining module 601 is used to determine the target dimension corresponding to the educational management problem text based on the text information corresponding to the educational management problem text.

[0144] Module 602 is used to construct an intelligent agent that matches the target dimension based on the target dimension corresponding to the text of the education management problem.

[0145] The first calling module 603 is used to call the intelligent agent to execute multiple tasks and obtain the task execution results corresponding to each task. The multiple tasks are used to analyze the text of educational management issues.

[0146] The fusion module 604 is used to generate an analysis report on educational management problem texts based on the fusion results of the execution results of each task. The analysis report includes the causes of the educational management problem texts and adjustment strategies.

[0147] In some optional implementations, the first determining module 601 is further configured to extract multiple entity objects from the text information corresponding to the educational management problem text; and to retrieve the target dimension corresponding to each entity object by searching a preset dimension library. The preset dimension library is used to store the mapping relationship between entity objects and target dimensions.

[0148] In some optional implementations, the target dimension includes a business dimension and a constraint dimension; the construction module 602 is also used to retrieve the function library based on the business dimension to obtain the function information that matches the business dimension. The function library is used to store the mapping relationship between the business dimension and the function information. The function information includes data processing strategies, analysis and decision-making strategies, and domain knowledge; based on the function information and the constraint information corresponding to the constraint dimension, target prompt words are constructed. The constraint information includes subject information, time information, and data source information; a general intelligent agent base is obtained, and the target prompt words are configured as prompt words of the general intelligent agent base to obtain the intelligent agent.

[0149] In some optional implementations, the first calling module 603 is further configured to divide tasks belonging to the same dependency level into the same group based on the dependency relationship between the tasks, thereby obtaining multiple groups; determine the target agent corresponding to each group from multiple agents based on the task description of the tasks in each group; execute the tasks in the corresponding group based on the target agent corresponding to each group, and obtain the task execution results corresponding to the tasks in each group, thereby obtaining the task execution results of all tasks.

[0150] In some optional implementations, the first calling module 603 is further configured to semantically match the task descriptions corresponding to each task in each group with the functional information of each agent to determine the target agent corresponding to each task, wherein the functional information is configured as data processing strategy, analysis and decision-making strategy and domain knowledge; and configure the target agent corresponding to each task as the target agent corresponding to the group.

[0151] In some optional implementations, the first invocation module 603 is further configured to, for each group, combine each task in the group with the target agent corresponding to the task to obtain multiple combinations corresponding to the group, each combination including a task and a target agent corresponding to the task; obtain the information gain corresponding to each combination, and determine the scheduling priority between the combinations based on the information gain; based on the scheduling priority corresponding to each combination, invoke the target agent in the combination to execute the task in the combination, and obtain the initial task execution result corresponding to the task in each combination; based on the confidence level corresponding to the initial task execution result, filter the initial task execution results corresponding to each task in the group to obtain the task execution result corresponding to each task in the group.

[0152] In some alternative embodiments, the device further includes: The second calling module is used to call the target agent corresponding to the task execution result to re-execute the task corresponding to the task execution result if the task execution result lacks structured element information, so as to obtain the first new task execution result.

[0153] The third calling module is used to call the target agent corresponding to the task execution result to re-execute the task corresponding to the task execution result based on the backup analysis and decision strategy if the confidence of the task execution result is less than the preset confidence threshold, so as to obtain the second new task execution result.

[0154] The adjustment module is used to adjust the education management problem text to obtain a new education management problem text if there is a conflict between the task execution result and the existing task execution results in the shared pool, and then continue to determine the analysis report corresponding to the new education management problem text.

[0155] In some alternative implementations, the device for acquiring multiple tasks includes: The splitting module is used to break down the overall task of analyzing educational management problem texts into multiple initial tasks based on the text information corresponding to the educational management problem texts.

[0156] The second determination module is used to determine the target initial tasks at the same dependency level and the semantic similarity between pairs of target initial tasks based on the dependency relationships between multiple initial tasks.

[0157] The merging module is used to merge two initial target tasks with a semantic similarity greater than a preset similarity threshold into one task, resulting in multiple tasks.

[0158] The educational data analysis device provided in this embodiment of the invention can execute the educational data analysis method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the various modules and units described above are the same as in the corresponding embodiments described above, and will not be repeated here.

[0159] Figure 7 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. The electronic device may be the terminal device or server described above.

[0160] The following is a detailed reference. Figure 7 The diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 701, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 702 or a program loaded from memory 708 into random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device. The processor 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0161] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 7 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0162] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 709, or installed from a memory 708, or installed from a ROM 702. When the computer program is executed by the processor 701, it performs the functions defined in the educational data analysis method of the embodiments of the present invention.

[0163] Figure 7 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0164] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the educational data analysis method shown in the above embodiments is implemented.

[0165] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0166] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. An educational data analysis method, characterized in that, The method includes: Based on the textual information corresponding to the educational management problem text, the target dimension corresponding to the educational management problem text is determined; Based on the target dimension corresponding to the educational management problem text, an intelligent agent matching the target dimension is constructed; The intelligent agent is invoked to execute multiple tasks, and the task execution results corresponding to each task are obtained. The multiple tasks are used to analyze the educational management problem text. Based on the fusion results of the execution results of each task, an analysis report is generated for the educational management problem text. The analysis report includes the reasons for the formation of the educational management problem text and adjustment strategies.

2. The method according to claim 1, characterized in that, The process of determining the target dimension corresponding to the educational management problem text based on the text information corresponding to the educational management problem text includes: Extract multiple entity objects from the text information corresponding to the educational management problem text; The preset dimension library is searched based on each entity object to obtain the target dimension corresponding to each entity object. The preset dimension library is used to store the mapping relationship between the entity objects and the target dimensions.

3. The method according to any one of claims 1 or 2, characterized in that, The target dimension includes a business dimension and a constraint dimension; the construction of an intelligent agent matching the target dimension based on the target dimension corresponding to the educational management problem text includes: The function library is retrieved based on the business dimension to obtain the function information that matches the business dimension. The function library is used to store the mapping relationship between the business dimension and the function information. The function information includes data processing strategies, analysis and decision-making strategies, and domain knowledge. Based on the functional information and the constraint information corresponding to the constraint dimension, target prompt words are constructed. The constraint information includes subject information, time information, and data source information. Obtain the general intelligent agent base and configure the target prompt word as the prompt word of the general intelligent agent base to obtain the intelligent agent.

4. The method according to claim 1, characterized in that, There are multiple intelligent agents; the process of invoking the intelligent agents to execute multiple tasks and obtaining the task execution results corresponding to each task includes: Based on the dependencies between the various tasks, the tasks belonging to the same dependency level are divided into the same group, resulting in multiple groups; Based on the task descriptions of the tasks in each group, the target agent corresponding to each group is determined from among the multiple agents; Based on the target agent corresponding to each group, the task in the corresponding group is executed to obtain the task execution result corresponding to the task in each group.

5. The method according to claim 4, characterized in that, The step of determining the target agent corresponding to each group from among multiple agents based on the task description of the task in each group includes: For each group, the task description corresponding to each task in the group is semantically matched with the functional information of each agent to determine the target agent corresponding to each task. The functional information includes data processing strategy, analysis and decision-making strategy and domain knowledge. Configure the target agent corresponding to each task as the target agent corresponding to the group.

6. The method according to claim 4, characterized in that, The step of executing the tasks within the corresponding groups based on the target agents of each group, and obtaining the task execution results for each group, includes: For each group, each task in the group is combined with the target agent corresponding to the task to obtain multiple combinations corresponding to the group. Each combination includes one task and one target agent corresponding to the task. Obtain the information gain corresponding to each of the combinations, and determine the scheduling priority among the combinations based on the information gain; Based on the scheduling priority corresponding to each of the combinations, the target agent in the combination is invoked to execute the task in the combination, and the initial task execution result corresponding to the task in each combination is obtained; Based on the confidence level corresponding to the initial task execution result, the initial task execution results corresponding to each task in the group are filtered to obtain the task execution results corresponding to each task in the group.

7. The method according to any one of claims 4 to 6, characterized in that, After obtaining the task execution results corresponding to each of the aforementioned tasks, the method further includes: If the task execution result lacks structured element information, the target agent corresponding to the task execution result is invoked to re-execute the task corresponding to the task execution result, and a first new task execution result is obtained; And / or, If the confidence level of the task execution result is less than the preset confidence threshold, then the target agent corresponding to the task execution result is invoked to re-execute the task corresponding to the task execution result based on the backup analysis and decision strategy to obtain a second new task execution result; And / or, If there is a conflict between the task execution result and the existing task execution results in the shared pool, the education management problem text is adjusted to obtain an adjusted education management problem text, and the analysis report corresponding to the adjusted education management problem text is then determined.

8. The method according to claim 1, characterized in that, The process of acquiring the multiple tasks includes: Based on the text information corresponding to the educational management problem text, the overall task of analyzing the educational management problem text is broken down into multiple initial tasks; Based on the dependencies between the multiple initial tasks, the target initial tasks at the same dependency level and the semantic similarity between each pair of the target initial tasks are determined. Two initial target tasks with a semantic similarity greater than a preset similarity threshold are merged into one task to obtain the multiple tasks.

9. An educational data analysis device, characterized in that, The device includes: The first determining module is used to determine the target dimension corresponding to the educational management problem text based on the text information corresponding to the educational management problem text; The construction module is used to construct an intelligent agent that matches the target dimension based on the target dimension corresponding to the educational management problem text; The first invocation module is used to invoke the intelligent agent to execute multiple tasks and obtain the task execution results corresponding to each task. The multiple tasks are used to analyze the educational management problem text. The fusion module is used to generate an analysis report on the educational management problem text based on the fusion results of the execution results of each of the tasks. The analysis report includes the reasons for the formation of the educational management problem text and adjustment strategies.

10. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the educational data analysis method according to any one of claims 1 to 8.

Images