A multi-agent system applied to educational statistics and a use method thereof

Abstract

This application belongs to the field of educational information technology, specifically relating to a multi-agent system and its usage method applied to educational statistics. The multi-agent system includes: first to fourth agents; the first agent acquires user needs information, generates corresponding keyword information, and sends it to the corresponding agent; it can also output the processing results of other agents to the user; the second agent acquires relevant raw theoretical knowledge data and feeds it back to the first agent; the third agent acquires relevant practical cases in educational statistics and sends them to the first agent; and the fourth agent acquires relevant software data and sends it to the first agent. The system and method of this application enable students to increase interaction with agents in traditional teaching environments, forming a "teacher-student-computer" three-way interaction through independent exploration and collaboration, thereby completing the entire teaching process from data collection, cleaning, analysis to result interpretation.

Description

Technical Field

[0001] This application belongs to the field of educational information technology, specifically relating to a multi-agent system for educational statistics and its usage method. Background Technology

[0002] In the field of education, especially in courses like educational statistics that require a balance between theory and practice, teaching and practice have long faced several inherent challenges. As an interdisciplinary subject integrating statistical principles and educational research, educational statistics aims to cultivate students' ability to analyze and solve practical educational problems using data-driven thinking. However, traditional teaching models often rely on one-way theoretical lectures, static case analyses, and basic software demonstrations. This approach falls short in addressing the ever-evolving educational research methods and complex, dynamic educational scenarios. Furthermore, teachers' knowledge inevitably lags behind the demands of the times. Educational statistics is a dynamic process, with statistical methods constantly evolving, while teachers, constrained by time and space, find it difficult to update their own knowledge base.

[0003] Specifically, existing teaching systems or methods typically have the following limitations: 1. Insufficient personalization and adaptability: It is difficult to provide dynamically adjusted teaching content, difficulty, and pathways based on learners' different knowledge bases, cognitive styles, and learning progress. When students are dealing with tasks of varying complexity, from descriptive statistics to inferential statistics and multi-level models, they often cannot receive targeted guidance and real-time feedback.

[0004] 2. Weak Contextualization and Interactivity: The application scenarios of educational statistics (such as academic evaluation and assessment of the effectiveness of teaching interventions) are highly systematic and complex. Traditional teaching models struggle to construct interactive simulation environments.

[0005] 3. Lack of Intelligent Assistance and Collaborative Teaching: In a lesson preparation environment, a teaching team composed of several teachers should jointly design the teaching plan. However, in reality, it is often completed by one person, making team collaboration and division of labor difficult. Students' guidance is limited by the time and space constraints of the classroom, and they often lack teacher assistance during practical exercises outside of class, leading to problems such as emotional frustration.

[0006] 4. Limited teaching resources and assessment models: Teaching content and case studies are outdated, failing to fully utilize the ever-growing resources of educational statistical methods. Furthermore, assessments of learning outcomes often focus on final reports or exam scores, lacking formative and multi-dimensional evaluations of practical steps, analytical processes, logical connections, and collaborative interactions.

[0007] In recent years, multi-agent system technology has demonstrated great potential in multiple fields due to its advantages in distributed problem solving, complex system simulation, and human-machine collaboration. This technology, through the interaction and collaboration among multiple autonomous and social agents, can generate complex collective intelligent behaviors, making it highly suitable for building flexible and dynamic learning support environments in education and teaching, and also applicable to teacher-student-computer three-way interactions.

[0008] Therefore, introducing the concepts and technologies of multi-agent systems into the teaching process of educational statistics, and designing a system and method that can provide personalized learning paths, immersive scenario simulations, agent-assisted collaborative learning, and process-based comprehensive evaluation, has significant practical implications and innovative value. This is not only a beneficial exploration in the field of educational technology, but also an important way to improve the quality of educational statistics talent training and address future challenges in educational data science. This invention aims to overcome the aforementioned shortcomings of existing technologies and provide an innovative solution.

[0009] The knowledge bases used in each submodule of this system are constructed using RAG technology. RAG stands for Retrieval-Augmented Generation, a technique that combines retrieval systems and text generation to supplement large models with proprietary knowledge. It retrieves relevant information from a large knowledge base and feeds this information, along with the question, into the large model for processing. In this way, RAG helps the large model generate more accurate, relevant, and richer answers. Simply put, RAG involves retrieval before generation, giving the large model a solid foundation and supplementing it with proprietary knowledge. The core idea of ​​RAG is to compensate for the shortcomings of generative models in handling knowledge-intensive tasks through the organic combination of retrieval and generation. Traditional generative models often generate incorrect or irrelevant answers when faced with complex problems due to a lack of sufficient knowledge. RAG, however, obtains relevant background information through its retrieval module and can refer to this information during the generation process to produce more credible and accurate answers. Summary of the Invention

[0010] To address at least one technical problem existing in the prior art, this application provides a multi-agent system and its usage method for educational statistics.

[0011] In a first aspect, this application discloses a multi-agent system for educational statistics, comprising: A first intelligent agent, the first intelligent agent being used for: Obtain user-described needs information regarding educational statistics questions; and After processing the demand information, keyword information corresponding to user needs is generated. These user needs include theoretical knowledge learning needs related to educational statistics, case study learning needs, and software download needs. Based on the keyword information, a corresponding intelligent agent capable of handling the user's request is identified, and the keyword information is sent to that intelligent agent; and It can output the results obtained by other intelligent agents in processing corresponding keyword information to the user; The second intelligent agent is used to respond to the received keyword information, obtain the corresponding original theoretical knowledge data, and feed back the theoretical knowledge data to the first intelligent agent; The third intelligent agent is used to respond to the received keyword information, obtain the corresponding educational statistics practice cases, and send the educational statistics practice cases to the first intelligent agent; The fourth intelligent agent is used to respond to the received keyword information, obtain the corresponding software data, and send the software data to the first intelligent agent.

[0012] Optionally, the first intelligent agent includes: The interaction module is used to receive user request information and output the results obtained by other intelligent agents from processing corresponding keyword information to the user. The first execution module is used to extract corresponding keyword information from the demand information based on preset keyword data, and then send the keyword information to the corresponding intelligent agent according to the internally preset keyword-intelligent agent correspondence rules.

[0013] Optionally, when the first execution module fails to extract keyword information from the requirement information, the first execution module is further configured to: First guidance information is generated based on the internally preset first guidance response template and sent to the interaction module. The first guidance information is used to guide the user to re-enter the demand information that is more consistent with their educational statistics question. Accordingly, the interaction module is also used for: The first guidance information is output to the user, and the system continues to receive the user's request information based on the first guidance information.

[0014] Optionally, when there are multiple keyword information items, and it is determined that at least two intelligent agents need to process different keyword information items respectively, the first execution module is further configured to: Based on the priority of theoretical knowledge learning needs, case study learning needs, and software download needs, keyword information is sent to the corresponding intelligent agents for processing; and The processing results fed back by other intelligent agents are sent to the interaction module according to the priority of theoretical knowledge learning needs, case operation learning needs, and software download needs, so as to output them to the user.

[0015] Optionally, the second intelligent agent includes: The theoretical knowledge base stores various theoretical knowledge about educational statistics, including basic knowledge, various theoretical concepts, formulas and principles, methods and approaches, corresponding statistical exercises, and reference information for various theories. The second execution module is used to retrieve theoretical text fragments corresponding to keyword information from the theoretical knowledge base, process the retrieved theoretical text fragments to obtain theoretical knowledge data that meets the user's needs, and then feed the theoretical knowledge data back to the first intelligent agent. The processing of the retrieved theoretical text fragments includes at least simplification, format unification, privacy information removal, and illegal and non-compliant content avoidance.

[0016] Optionally, when the second execution module fails to retrieve a theoretical text fragment corresponding to the keyword information from the theoretical knowledge base, the second execution module is further configured to: A second guidance message is generated based on a pre-set second guidance response template and sent to the interaction module. This second guidance message guides the user to re-enter requirement information that better matches their theoretical knowledge learning needs; and / or The system searches for publicly available relevant information using a pre-built search tool, provides a brief description of the information, and then sends the brief description and a link to the relevant information back to the first intelligent agent for output to the user.

[0017] Optionally, the third intelligent agent includes: The case knowledge base stores various practical cases related to educational statistics; The third execution module is used to retrieve case text fragments corresponding to keyword information from the case knowledge base, perform structured processing on the case text fragments to obtain practical case data that meets user needs, and then feed the practical case data back to the first intelligent agent (101). The practical case data obtained by structured processing includes at least case exercises, data software operation steps, and data result analysis. In addition, when the third execution module fails to retrieve a case text fragment corresponding to the keyword information from the case knowledge base, the third execution module is also used to: The third guidance information is generated according to the internally preset third guidance response template and sent to the interaction module. The third guidance information is used to guide the user to re-enter the requirement information that is more consistent with the user's case operation learning needs, or to guide the user to request a new practical case, which is generated by the third intelligent agent. Accordingly, the third intelligent agent also includes: The case replication module is used to associate with the case knowledge base, replicate practical cases in the case knowledge base, generate new practical cases, and store them. Replicating practical cases in the case knowledge base includes at least the following operations: Analyze the data structure of the corresponding practical cases, extract the variable relationships and statistical analysis logic, and retain the variable coding rules of the original data; Based on the analysis results, data samples matching the user's case study learning needs are extracted, and fictitious data that conforms to the logic of educational statistics is generated. Without changing the number of variables, variable types, and core analysis methods of the original practical case data, the variables and analysis objectives of the original practical case data are reasonably replaced and expanded.

[0018] Optionally, in the theoretical knowledge base and case knowledge base, the content of the corresponding theoretical knowledge and practical cases is stored in the form of text fragments. The text fragments are divided according to the length of the text, and the next paragraph reproduces 10%-15% of the content of the previous paragraph. Furthermore, each text fragment in the theoretical knowledge base and case knowledge base is converted into vector form; Accordingly, the second and third execution modules can respectively convert keyword information into vector distances and measure the similarity of content by calculating the distance between vectors, thereby matching in the corresponding theoretical knowledge base and case knowledge base to obtain the corresponding text fragments.

[0019] Optionally, the fourth intelligent agent includes: The fourth execution module is used to retrieve matching software data based on keyword information and feed the software data back to the first intelligent agent. The software data includes at least a URL link, download assistance steps, and download version suggestions.

[0020] Secondly, this application also discloses a method for using a multi-agent system applied to educational statistics, comprising the following steps: Step 1: Obtain the user's description of the needs regarding educational statistics issues through the first intelligent agent, process the needs information to generate keyword information corresponding to the user's needs, determine the corresponding intelligent agent that can handle the user's needs based on the keyword information, and send the keyword information to the intelligent agent. The user's needs include theoretical knowledge learning needs, case operation learning needs, and software download needs related to educational statistics. Step 2: The second intelligent agent responds to the received keyword information, obtains the corresponding original theoretical knowledge data, and feeds back the theoretical knowledge data to the first intelligent agent; and / or The third intelligent agent receives the keyword information, obtains the corresponding educational statistics practice case, and sends the educational statistics practice case to the first intelligent agent; and / or The fourth intelligent agent responds to the received keyword information, obtains the corresponding software data, and sends the software data to the first intelligent agent; Step 3: The first agent outputs the results obtained by other agents in processing the corresponding keyword information to the user.

[0021] This application has at least the following beneficial technical effects: 1) The multi-agent system and its usage method applied to educational statistics in this application enable students to increase the forms of interaction with intelligent agents in the traditional teaching environment. Through independent exploration and collaboration, a three-way interaction of "teacher-student-computer" is formed. In this interaction mode, the entire teaching process from data collection, cleaning, analysis to result interpretation is completed. 2) Contextualized Interactive Simulation: Students' past learning has been limited to theory, and there is often a gap between case studies and their own practice. In particular, the unchanging case studies in textbooks often lead to students becoming rigid. The case study assistant agent (i.e., the third agent) can simulate and generate new questions based on cases in the knowledge base using built-in data tables, which to some extent makes up for the shortcomings of contextualized interaction. 3) Cross-time and space collaborative lesson preparation and intelligent after-class assistance: This system integrates real-time Q&A, practical guidance and emotional support functions. It answers students' practical questions after class through voice / text interaction (such as software operation steps and statistical model selection). At the same time, it identifies emotional frustration signals through behavioral data (such as the number of repeated operation failures and interaction tone) and sends encouraging words in a timely manner to avoid students' emotional frustration. 4) The multi-agent system and its application method in educational statistics presented in this application can have beneficial effects on personalized dynamic adaptation of students: the memory database collects user interaction information in real time, including but not limited to knowledge base and learning progress. Teachers can use this data to construct a personal learning profile of each user in real time. The intelligent agent will also update its long-term memory and feedback based on the data to adapt to different user styles. This can, to some extent, compensate for the current lack of personalized courses. To a certain extent, it solves the problem that the traditional "one-size-fits-all" teaching model cannot adapt to the different knowledge bases, cognitive styles, and learning progress of students.

[0022] 5) The system and method of this application are capable of real-time crawling and filtering of the latest educational statistical methods and resources (such as cutting-edge statistical models, practical cases in the field of education, and research paper data). Teaching content and case databases can be manually updated at any time, avoiding the phenomenon of outdated knowledge systems remaining stagnant for extended periods, leading to teaching falling behind social realities. This can, to some extent, solve the problems of lagging teaching resource updates and a limited case database. Simultaneously, the memory database records user interaction information in real time, allowing teachers to review these records and utilize this interaction process for formative assessment. To a certain extent, this can break away from the singular assessment model that emphasizes results while neglecting process, filling the gap in formative assessment of practical processes, logical chains, and collaborative interactions. Attached Figure Description

[0023] Figure 1 This is a diagram showing the composition of the multi-agent system applied to educational statistics in this application; Figure 2 This is a diagram showing the configuration of the first agent in the multi-agent system applied to educational statistics in this application; Figure 3 This is a diagram showing the structure of the second agent (theoretical learning assistant) in the multi-agent system applied to educational statistics in this application; Figure 4 This is a diagram showing the structure of the third agent (case assistant) in the multi-agent system applied to educational statistics in this application; Figure 5 This is a diagram showing the structure of the multi-agent system applied to educational statistics in this application, specifically the fourth agent (software download assistant). Figure 6 This is a flowchart illustrating the knowledge base retrieval process in the multi-agent system of this application; Figure 7 This is a structural diagram of the knowledge base in the multi-agent system of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some, but not all, embodiments of this application. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0025] The following is in conjunction with the appendix Figure 1-7 This application provides a more detailed description of the multi-agent system for educational statistics and its usage.

[0026] Firstly, this application discloses a multi-agent system for educational statistics, such as... Figure 1 As shown, the multi-agent system includes a first agent 101, a second agent 102 (also called a theoretical learning assistant), a third agent 103 (also called a case assistant), and a fourth agent 104 (also called a software download assistant). It should be noted that the second to fourth agents 104 are all capable of operating in single-agent mode.

[0027] The first intelligent agent 101 is used to acquire user needs information about educational statistics problems described in natural language, process the needs information to generate keyword information corresponding to the user's needs, determine the appropriate intelligent agent that can handle the user's needs based on the keyword information, and send the keyword information to that intelligent agent. It can also output the results obtained by other intelligent agents processing the corresponding keyword information to the user. The user's needs include theoretical knowledge learning needs related to educational statistics, case study learning needs, and software download needs.

[0028] Specifically, such as Figure 2 As shown, the first intelligent agent 101 may include an interaction module 1011 and a first execution module 1012. The interaction module 1011 is used to receive the user's request information and output the results obtained by other intelligent agents in processing the corresponding keyword information to the user. The first execution module 1012 is used to extract the corresponding keyword information from the request information according to the preset keyword data, and then send the keyword information to the corresponding intelligent agent according to the internal preset keyword and intelligent agent correspondence rules.

[0029] It should also be noted that when the first execution module 1012 fails to extract keyword information from the requirement information, the first execution module 1012 is also used for: The first guidance information is generated based on the internally preset first guidance response template and sent to the interaction module 1011. The first guidance information is used to guide the user to re-enter the demand information that is more consistent with their educational statistics question, such as: "Do you need theoretical explanation, case practice, or statistical software related services? You can add specific requirements."

[0030] Correspondingly, the interaction module 1011 is also used for: The system outputs the initial guidance information to the user and continues to receive further input from the user based on that information. This process is then repeated until a match is found.

[0031] Furthermore, when there are multiple keyword information items, and it is determined that at least two agents are needed to process different keyword information items respectively, the first execution module 1012 is also used to: Based on the priority of theoretical knowledge learning needs, case study learning needs, and software download needs, keyword information is sent to the corresponding intelligent agents for processing; and The processing results fed back by other intelligent agents are sent to the interaction module 1011 according to the priority of theoretical knowledge learning needs, case operation learning needs, and software download needs, so as to be output to the user.

[0032] Furthermore, the interaction module 1011 may also include a text acquisition unit and a voice input unit, which are used to acquire text information and voice information entered by the user regarding their needs, respectively.

[0033] Correspondingly, the first execution module 1012 may also include a text parsing unit and a speech parsing unit, the purpose of which is to convert information entered by the user in various forms into natural language; wherein, the text parsing unit is used to extract keyword information from the text information entered by the user, and the speech parsing unit is used to extract keyword information from the speech information entered by the user. Furthermore, before the speech parsing unit performs parsing, it is also used to determine whether the speech entered by the user is a pre-preset standard format speech. If it is, it is parsed directly; if not, the entered speech is first converted into the standard format speech before parsing.

[0034] It is understood that the above embodiments are a summary description of the role of the first intelligent agent 101. As the overall control and scheduling assistant of the multi-agent system for educational statistics, the first intelligent agent 101 is the only user interaction entry point. The following will describe it in more detail from multiple perspectives, including core responsibilities, demand judgment rules, scheduling logic, and result aggregation specifications.

[0035] 1.1 Core Responsibilities: 1) Unified interaction entry point: As the only external node of the multi-agent system, it uniformly receives all educational statistics-related needs from users (including theories, cases, software tools, etc.), avoiding the problem of chaotic operation of multiple sub-agents (i.e., the second to fourth agents); 2) Accurate demand identification: Through keyword matching and intent analysis, quickly determine the type of user demand (pure theory, pure case study, pure software tool, or cross-type combination demand). 3) Sub-agent scheduling: Automatically distribute tasks to corresponding sub-agents according to demand type (theoretical learning / case learning / scientific research software download assistant), supporting single-sub-agent invocation and multi-sub-agent collaborative invocation; 4) Result aggregation and feedback: Collect the response results of sub-agents, integrate them according to the logical priority and module structure of "theory → case → tool", and output a unified and coherent solution; 5) Fault tolerance and fallback handling: When a sub-agent does not respond, its needs are vague, or it exceeds its functional scope, it will proactively ask for supplementary information or directly provide basic answers to ensure service continuity.

[0036] 1.2 Requirements Judgment Rules: 1) If it contains any of the keywords "theory, knowledge point, principle, formula, test point, concept" and has no other core keywords, then it is a purely theoretical requirement; 2) If any of the keywords "case study," "imitation," "problem-solving," "exercise," or "data analysis" are included, then the requirement is purely a case study. 3) If the request contains any of the keywords "download, software, SPSS, R, Stata, installation, tutorial", it is a pure software requirement; 4) If two or more types of keywords are present, it is considered a cross-type combination requirement, and should be prioritized as "theoretical learning - case operation - software download".

[0037] 1.3 Scheduling Logic: 1) Pure type requirements: Directly call the corresponding sub-agent without any extra redundant operations; 2) Cross-type requirements: Sub-agents are scheduled in order of priority to ensure a coherent learning process; 3) After a sub-agent is invoked, it must wait for a response synchronously and must not miss any feedback from any sub-agent.

[0038] 1.4 Result Aggregation Specification: 1) Presented in a structured format of "module + core content," with cross-type requirements ordered by "theory-case study-tools," as shown in the example below: [Requirement Answers] Theoretical explanation (from the theoretical learning assistant): XXX; Case content (from Case Study Assistant): XXX; Software support (from software download assistant): XXX.

[0039] 1.5 Fault Tolerance and Safety Net: 1) If the requirements are vague (e.g., "How to analyze data"), then ask follow-up questions such as "Do you need theoretical explanations, case studies, or statistical software-related services? Please provide more specific requirements." 2) If the sub-agent does not respond: the message "The current module has not yet obtained a solution. We will reschedule or manually provide a basic solution for you" will be returned. (It should be noted that the solution can also be generated by the corresponding sub-agent itself.) 3) Beyond the scope of the sub-entity's functions: then directly provide a basic compliance answer, explaining "This requirement exceeds the sub-assistant's functions, please provide the core information: XXX".

[0040] 1.6 Constraints: 1) It does not directly answer specialized questions within the responsibilities of sub-agents, but is only responsible for scheduling and aggregation; 2) No scheduling errors occur (such as assigning case requirements to theoretical sub-agents); 3) The feedback language is concise and professional, suitable for educational statistics scenarios, and does not deviate from the user's learning needs.

[0041] Furthermore, based on the above descriptions of the first intelligent agent 101 from multiple perspectives, the applicable scenarios (preceding node switching trigger conditions) of the first intelligent agent 101 need to be summarized or supplemented as follows: 1) Initial user question scenario: When the preceding node does not clearly define the type of user need (such as "educational statistics related questions", "how to learn ANOVA"), and there is no specific sub-agent pointing to it; 2) Cross-sub-agent requirement scenario: When user requirements involve two or more types of sub-agent functions (such as "related analysis theory + case + software download"), and multiple sub-agents need to coordinate and schedule; 3) Scenarios that the sub-agent cannot handle: When the preceding sub-agent receives a request that is beyond its own capabilities (such as a theoretical agent encountering a software download requirement) or that it cannot answer; 4) Scenarios with ambiguous requirements / requiring follow-up questions: When the user's question is unclear (e.g., "How to analyze the data"), and the preceding nodes cannot determine the type of requirement; 5) Result integration requirement scenario: When multiple sub-agents have responded separately and need to be integrated according to the learning logic or a unified answer format is required; 6) System backup: If the preceding node becomes disordered, the system will ensure normal operation.

[0042] Furthermore, the second intelligent agent 102 is used to respond to the received keyword information, obtain the corresponding original theoretical knowledge data, and feed the theoretical knowledge data back to the first intelligent agent 101.

[0043] Specifically, such as Figure 3As shown, the second intelligent agent 102 may include a theoretical knowledge base 1021 and a second execution module 1022.

[0044] The theoretical knowledge base 1021 stores various theoretical knowledge about educational statistics, including basic knowledge, various theoretical concepts, formulas and principles, methods and corresponding statistical exercises, as well as reference information corresponding to various theories. The second execution module 1022 is used to retrieve theoretical text fragments corresponding to keyword information from the theoretical knowledge base, process the retrieved theoretical text fragments to obtain theoretical knowledge data that meets the user's needs, and then feed the theoretical knowledge data back to the first intelligent agent 101. The processing of the retrieved theoretical text fragments includes at least simplification, format unification, privacy information removal, and avoidance of illegal and non-compliant content.

[0045] Further, see Figure 7 As shown, the content entered into the theoretical knowledge base 1021 is segmented according to the length of the text, with each subsequent segment replicating 10%-15% of the content from the previous segment, to facilitate retrieval by the second execution module 1022. Furthermore, the theoretical knowledge base 1021 converts each text segment into vector form for more precise retrieval, enabling the capture of the text's contextual relationships and semantic information. See also... Figure 6 As shown, the second execution module 1022 can convert keyword information into vector distance and measure the similarity of content by calculating the distance between vectors, thereby matching in the corresponding theoretical knowledge base 1021 to obtain the corresponding theoretical text fragment.

[0046] Furthermore, when the second execution module 1022 fails to retrieve a theoretical text fragment corresponding to the keyword information from the theoretical knowledge base 1021, the second execution module 1022 is also used to: The second guidance information is generated based on the internally preset second guidance response template and sent to the interaction module 1011. This second guidance information (e.g., as mentioned earlier: "The current module has not yet obtained an answer; it will reschedule or manually provide a basic answer") is used to guide the user to re-enter information that better matches their theoretical knowledge learning needs; and / or The system searches for publicly available relevant information using a pre-built search tool, provides a brief description of the information, and then sends the brief description and the link to the relevant information back to the first intelligent agent (101) for output to the user.

[0047] Similarly, the second agent 102 will be described in more detail here from multiple perspectives, including response logic, workflow, skills, and limitations.

[0048] 2.1 Response Logic: When users ask questions about educational statistics, first offer emotional value, such as encouragement, followed by a detailed explanation. When users feel overwhelmed by the knowledge, offer comfort and encouragement. For example, the agent needs to provide theoretical knowledge about educational statistics, such as its history, applications, and core functions. For complex theories, you need to respond in simple, easy-to-understand language that is engaging and interactive. If users find the theory too obscure, use a lighthearted, slightly humorous tone, and try to explain complex concepts using stories or appropriate metaphors.

[0049] 2.2 Workflow: Step 1: Problem Understanding and Response Analysis 1) Carefully understand the content retrieved from the theoretical knowledge base and the questions entered by the user, and determine whether the retrieved content is the answer to the user's question; 2) If you cannot understand the user's question, for example, if the user's question is too simple or does not contain necessary information, you need to ask the user again until you are sure that you have understood the user's question and needs.

[0050] Step Two: Answer User Questions 1) After careful evaluation, it was determined that the user's question was completely unrelated to educational statistics and SPSS knowledge, and therefore the answer should be refused; 2) If no content is retrieved from the knowledge base, you can use a phrase like, "I'm sorry, the knowledge I've already acquired does not include content related to your question, so I cannot provide an answer at this time. If you have other questions related to educational statistics, I will try to help you answer them." 3) If the retrieved content is related to the user's question, only the relevant parts of the knowledge base should be extracted. The retrieved content should then be organized, summarized, integrated, and optimized. The answers provided to the user must be accurate and concise, without needing to specify the data source.

[0051] 2.3 Skills: 1) Skill 1: Using PBL (Project-Based Learning) to enhance learning practice - Design thought-provoking and challenging problem scenarios based on the course content and learning progress of educational statistics; - Guide teachers and students to analyze problems, identify the educational statistics knowledge and methods needed to solve them, and encourage them to explore independently and collaborate in groups; - Provide necessary guidance and feedback in a timely manner during the problem-solving process for teachers and students, and help them summarize their experiences and deepen their understanding and application of knowledge.

[0052] Skill 2: Providing theoretical knowledge - When teachers and students ask questions about educational statistics, relevant knowledge is extracted from the dataset to provide detailed and accurate answers based on the topic of the question.

[0053] - If the dataset does not contain relevant knowledge, use tools to search for comprehensive and authoritative information, and organize it into easy-to-understand content for teachers and students.

[0054] 2.4 Limitations: 1) Questions that are prohibited from being answered -Personal privacy information, including but not limited to real name, phone number, address, account password and other sensitive information; - Illegal and irregular content: including but not limited to politically sensitive topics, pornography, violence, gambling, infringement, and other content that violates laws, regulations, and ethical standards.

[0055] 2) Style - The answer must be accurate, concise, and easy to understand; a professional and definitive response is required.

[0056] 3) Language - The answer should be in the same language as the user's input.

[0057] 4) Length of the answer - Concise and clear or detailed and rich.

[0058] 5) Only provide content related to educational statistics learning, SPSS applications, and PBL learning methods; refuse to answer irrelevant questions.

[0059] 6) The theoretical knowledge and case studies provided must be accurate, clear, and conform to the professional standards of educational statistics.

[0060] 7) The SPSS operation steps in the case should be detailed and easy to understand, so that teachers and students can imitate and practice them.

[0061] 8) The design of PBL problems should be closely integrated with the practical application of educational statistics and have certain practical significance.

[0062] Furthermore, the third intelligent agent 103 responds to the received keyword information, obtains the corresponding educational statistics practice cases, and sends the educational statistics practice cases to the first intelligent agent (101).

[0063] Specifically, such as Figure 4 As shown, the third intelligent agent 103 may include a case knowledge base 1031 and a third execution module 1032.

[0064] The case knowledge base 1031 stores various practical cases related to educational statistics. The third execution module 1032 is used to retrieve case text fragments corresponding to keyword information from the case knowledge base, perform structured processing on the case text fragments to obtain practical case data that meets user needs, and then feed the practical case data back to the first intelligent agent 101. The structured practical case data includes at least case exercises, data software operation steps, and data result analysis.

[0065] Similarly, the case knowledge base 1031 can segment the content of the corresponding theoretical knowledge and practical cases according to the length of the text, with each subsequent segment replicating 10%-15% of the content of the previous segment, thus storing the theoretical knowledge and practical cases in the form of text fragments; furthermore, the case knowledge base 1031 can also convert each text fragment of the corresponding theoretical knowledge and practical cases into vector form.

[0066] Correspondingly, the third execution module 1032 can convert keyword information into vector distances and measure the similarity of content by calculating the distance between vectors, thereby matching in the case knowledge base 1031 to obtain the corresponding case fragments.

[0067] Furthermore, when the third execution module 1032 fails to retrieve a case text fragment corresponding to the keyword information from the case knowledge base 1031, the third execution module 1032 is also used to: The third guidance information is generated according to the internally preset third guidance response template and sent to the interaction module 1011. The third guidance information is used to guide the user to re-enter the requirement information that is more consistent with the user's case operation learning needs, or to guide the user to request a new practical case, which is generated by the third intelligent agent 103.

[0068] Correspondingly, the third intelligent agent 103 also includes a case imitation module 1033, which is used to associate with the case knowledge base 1031, imitate the practical cases in the case knowledge base, generate new practical cases, and store them.

[0069] It should be noted that the third agent 103 aims to help students master the practical skills of SPSS software through interaction; it provides students with new exercise cases by utilizing statistical practice simulation content in the knowledge base, and ensures that the cases contain complete and accurate SPSS operation steps. Therefore, the third agent 103 will also be described in more detail here from the perspectives of skills and limitations.

[0070] 3.1 Skills: 1) Skill 1: Provide SPSS practical exercises - Identify needs: When students request to practice SPSS exercises, first identify the type of need (such as descriptive statistics, correlation analysis, comparative analysis, etc.).

[0071] - Retrieval and Extraction: By calling the corresponding workflow (e.g., SPSScase), retrieve matching statistical practice simulation cases from it, and prioritize extracting all original SPSS operation steps contained in the case (covering key aspects such as data entry, variable setting, analysis tool selection, operation sequence, and option settings) to ensure that the steps are complete and without omission.

[0072] - Adjustments and Changes: Make reasonable adjustments to the retrieved case background (such as data scenario, research topic) and the complete operational steps (such as changing data sources or adjusting analysis objectives) to ensure that the adjusted operational step framework remains unchanged, key operational steps are fully preserved, and educational statistics and SPSS logic are strictly followed.

[0073] 2) Skill 2: Generating SPSS Educational Statistics Practice Cases -Prioritize extracting existing steps: When students need to generate a certain type of SPSS educational statistics case, first search for similar cases in the knowledge base (case knowledge base), extract their complete operation steps as the basis for generation, and ensure that the steps are logically coherent and the step names are accurate; - Balance between innovation and integrity: When generating new cases, combine them with new educational scenarios (such as performance analysis of different grade levels, comparison of the effects of different teaching methods, etc.) while retaining the integrity of the original case's step framework. Only adjust the data scenario, analysis objectives, or variable settings to avoid omitting key operational steps. - Supplement the principles and logical steps: For the extraction operation steps, a clear explanation of the operation principle is required (such as "why we need to select analysis-descriptive statistics here") to ensure that students understand the statistical logic behind the steps. At the same time, the step description should be detailed to the specific operation of "click the menu bar-select the XX button-set the XX parameter".

[0074] 3) Skill 3: Generating SPSS Educational Statistics Practical Case Data Files - Call workflow retrieval: Call SPssdatabase to process the case based on the student's required case name; - Retrieve the corresponding SPSS case data file and provide a file download link; - This skill only provides case links and does not make any additional associations.

[0075] 4) Skill 4: Generating new cases by imitating existing ones - Identify needs: This skill can be used when users require additional practice questions beyond the case knowledge base; - Data Analysis and Application: From the uploaded Excel spreadsheet (College Students' Digital Literacy Database), fully analyze its data structure (including variable names, variable types, sample size, data distribution characteristics, etc.), extract variable relationships and statistical analysis logic (such as the correlation between "digital literacy score" and "grade" and "gender"), and retain the original data's variable coding rules (such as using 1 / 2 / 3 to represent primary school, junior high school, and senior high school for "grade"). Based on the analysis results, extract data samples that match the user's needs (ensuring that the sample size and variable coverage are consistent with the original file, and the data range is reasonable), and generate fictitious data that conforms to the logic of educational statistics (such as replacing actual data with simulated values ​​under new scenarios, but keeping the variable types and distribution characteristics unchanged).

[0076] - Structural Adjustment and Logic Preservation: Reasonably replace and expand the original case data variables and analysis objectives (e.g., add a "subject interest" variable, adjust the analysis objective to "correlation analysis between digital literacy and academic performance"), but strictly prohibit changing the number of variables, variable types, and core analysis methods of the original data (e.g., if the original case used a t-test, the new case should also use a t-test, and cannot be changed to ANOVA). Ensure that the generated new data does not contain any real sensitive information from the original database (e.g., specific names of people, school names, real questionnaire data), and that the data are logically consistent (e.g., the mean of "digital literacy score" should conform to the common range of educational statistics).

[0077] -Reply format and document provision: --Exercise Case: Describe the case background, such as "A university conducted a survey on the current status of college students' digital literacy. The data includes variables such as grade, gender, digital literacy score (0-100 points), and major"; --SPSS operation requirements: Complete operation steps, such as: Data entry steps: Import the "College Students' Digital Literacy Database" (fictional data), with "Grade" as a categorical variable, "Gender" as a binary variable, and "Digital Literacy Score" as a continuous variable; Descriptive statistics steps: Generate the mean and standard deviation of each variable through "Analyze - Descriptive Statistics - Description"; Correlation analysis steps: Through "Analysis - Correlation - Bivariate", select "Digital Literacy Score" and calculate the Pearson correlation coefficient with "Grade" and "Gender"; --Data file: Provide an Excel file (fictional data version) of the "College Students' Digital Literacy Database", containing the following variable table 1: Table 1 Variable Table for Exercise Cases

[0078] - Prohibited items: Do not change the statistical methods used in the original question (e.g., if the original question used a t-test, the new question must not change it to analysis of variance). Do not generate cases that contradict the characteristics of Excel data (e.g., if Excel data is bivariate, do not generate multifactor ANOVA questions); Non-educational scenarios (such as commercial data or medical data) should not be included. The original data in Excel is not disclosed; all new data is a reasonable fabrication based on the characteristics of the original data. Do not lower or raise the difficulty of the original question (e.g., if the original question only requires calculation, the new question must not require complex interpretation of the result).

[0079] 3.2 Limitations: 1) Extraction priority: The complete operation steps of the corresponding case must be extracted from the knowledge base first, and the key operation steps in the original case must not be deleted after the adjustment. 2) Completeness of steps: The output SPSS operation must include all necessary operation steps (such as data entry, variable definition, analysis tool calling, result export, etc.), and the order of steps, operation names, and parameter settings must be accurate.

[0080] 3) Logical consistency: All operational steps must conform to the actual operational logic of educational statistics and SPSS software, and there must be no contradictions in the steps or logical gaps.

[0081] Furthermore, such as Figure 5 As shown, the fourth intelligent agent 104 includes a fourth execution module 1041, which is used to respond to the received keyword information, retrieve matching software data, and feed the software data back to the first intelligent agent 101. The software data includes at least a website link, download assistance steps, and download version suggestions.

[0082] Specifically, in the fourth execution module 1041, intent recognition is first performed on keywords. After the keywords are recognized, the keyword information is matched to obtain the user's intent, which determines whether to invoke the tool and knowledge base retrieval modules. After the intent of the keyword information is recognized, the link reading tool is invoked according to the invocation command and the recognized intent to read external network links and match them with the keywords. The matched URL links are output, and the results include at least the software's official website URL. Further, the fourth execution module 1041 matches the user's needs and the keyword information from the intent retrieval module with the content in the knowledge base to match content results based on the user's device requirements, needs, and intent. The results include at least download assistance steps and download version suggestions. Finally, the fourth execution module 1041 adapts the results matched by the knowledge base retrieval module according to the prompts of this module and the user's needs to form an analysis and processing plan. If the recognition fails, the agent will provide an apologetic message.

[0083] The Fourth Agent 104 is a research software download platform assistant designed specifically for university faculty and students. It is able to assist faculty and students in successfully downloading research-related software with a professional, patient, and meticulous attitude, and provides comprehensive and practical software usage knowledge and skills learning support. Here, we will also provide a more detailed description of the Fourth Agent 104 from the perspectives of skills and limitations.

[0084] 4.1 Skills 1) Skill 1: Software Download Service When university faculty and students request to download a research software, first confirm the software name and version information, then accurately provide them with the download link (located in the knowledge base) or a reliable way to obtain it. If multiple channels exist, explain the characteristics of each channel. If teachers or students report problems during the download process, such as network failures, insufficient permissions, or download interruptions, inquire in detail about the specific circumstances of the problem, including information such as the operating system and network environment. Provide effective solutions based on the actual situation and offer guidance on the operating steps.

[0085] 2) Skill 2: Knowledge and skills transfer - For software downloaded by teachers and students, provide detailed and easy-to-understand usage instructions, including introductions to the basic functional modules of the software, demonstrations of operation procedures (with illustrated examples if necessary), and explanations of commonly used settings. - Share practical tips for using this software in scientific research scenarios, such as improving data processing efficiency, optimizing experimental simulation parameters, and quickly generating scientific charts, to help teachers and students improve their efficiency and better serve scientific research.

[0086] 4.2 Restrictions - Our services focus solely on providing knowledge and skills related to the download and use of research software by university faculty and students; we will not answer irrelevant questions. - The information provided must be accurate, reliable, and comply with relevant laws, regulations, and software usage guidelines.

[0087] - Replies should be concise, clear, and logically structured, highlighting key points. If complex procedures are involved, they should be explained step-by-step.

[0088] Furthermore, in the multi-agent system for educational statistics in this application, each agent also includes a memory database terminal. The memory database terminal is activated immediately after the user interaction module responds, and records the information of the user interaction process for later updates and iterations of the multi-agent system and future research.

[0089] Among them, the memory database of the first intelligent agent 101 is automatically activated when the interaction module is triggered. It is used to collect user interaction data (interaction type, content, duration, result, etc.) and record it into this database. On the one hand, it can be transferred to the memory bank for storage and collection. On the other hand, this data can help teachers use for scientific research and explore the interaction influencing factors between students and intelligent agents.

[0090] Specifically, the collected data is shown in Table 2 below: Table 2 Data collected from the memory database

[0091] Secondly, this application also discloses a method for using a multi-agent system applied to educational statistics, comprising the following steps: Step 1: Obtain the user's description of the needs regarding educational statistics through the first intelligent agent 101, process the needs information to generate keyword information corresponding to the user's needs, determine the corresponding intelligent agent that can handle the user's needs based on the keyword information, and send the keyword information to the intelligent agent. The user's needs include theoretical knowledge learning needs, case operation learning needs, and software download needs related to educational statistics. Step 2: The second intelligent agent 102 responds to the received keyword information, obtains the corresponding original theoretical knowledge data, and feeds back the theoretical knowledge data to the first intelligent agent 101; and / or The third intelligent agent 103 responds to the received keyword information, obtains the corresponding educational statistics practice cases, and sends the educational statistics practice cases to the first intelligent agent 101; and / or The fourth intelligent agent 104 responds to the received keyword information, obtains the corresponding software data, and sends the software data to the first intelligent agent 101; Step 3: The first intelligent agent 101 outputs the results obtained by other intelligent agents in processing the corresponding keyword information to the user.

[0092] The above embodiments are used to illustrate the technical solutions of this application, and not to limit its protection scope; any equivalent substitutions or modifications made without departing from the technical concept of this application shall fall within the protection scope of this application.

Claims

1. A multi-agent system for educational statistics, characterized in that, include: A first intelligent agent (101) is used for: Obtain user-described needs regarding educational statistics issues; as well as After processing the demand information, keyword information corresponding to user needs is generated. These user needs include theoretical knowledge learning needs related to educational statistics, case study learning needs, and software download needs. Based on the keyword information, a corresponding intelligent agent capable of handling the user's request is identified, and the keyword information is sent to that intelligent agent; and It can output the results obtained by other intelligent agents in processing corresponding keyword information to the user; The second intelligent agent (102) is used to respond to the received keyword information, obtain the corresponding original theoretical knowledge data, and feed back the theoretical knowledge data to the first intelligent agent (101). The third intelligent agent (103) is used to respond to the received keyword information, obtain the corresponding educational statistics practice case, and send the educational statistics practice case to the first intelligent agent (101). The fourth intelligent agent (104) is used to respond to the received keyword information, obtain the corresponding software data, and send the software data to the first intelligent agent (101).

2. The multi-agent system applied to educational statistics according to claim 1, wherein, The first intelligent agent (101) includes: The interaction module (1011) is used to receive the user's request information and output the results obtained by other intelligent agents in processing the corresponding keyword information to the user; The first execution module (1012) is used to extract the corresponding keyword information from the demand information according to the preset keyword data, and then send the keyword information to the corresponding intelligent agent according to the internal preset keyword and intelligent agent correspondence rules.

3. The multi-agent system for educational statistics according to claim 2, characterized in that, When the first execution module (1012) fails to extract keyword information from the requirement information, the first execution module (1012) is further configured to: First guidance information is generated based on the internally preset first guidance response template and sent to the interaction module (1011). The first guidance information is used to guide the user to re-enter the demand information that is more consistent with their educational statistics question. Accordingly, the interaction module (1011) is also used for: The first guidance information is output to the user, and the system continues to receive the user's request information based on the first guidance information.

4. The multi-agent system for educational statistics of claim 3, wherein, When there are multiple keywords, and it is determined that at least two agents are needed to process different keywords, the first execution module (1012) is further configured to: Based on the priority of theoretical knowledge learning needs, case operation learning needs, and software download needs, keyword information is sent to the corresponding intelligent agents for processing; as well as The processing results fed back by other intelligent agents are sent to the interaction module (1011) according to the priority of theoretical knowledge learning needs, case operation learning needs, and software download needs, so as to be output to the user.

5. The multi-agent system for educational statistics according to claim 1, characterized in that, The second intelligent agent (102) includes: The theoretical knowledge base (1021) stores various theoretical knowledge about educational statistics, including basic knowledge, various theoretical concepts, formulas and principles, methods and corresponding statistical exercises, as well as reference information corresponding to various theories. The second execution module (1022) is used to retrieve theoretical text fragments corresponding to keyword information from the theoretical knowledge base, process the retrieved theoretical text fragments to obtain theoretical knowledge data that meets the user's needs, and then feed the theoretical knowledge data back to the first intelligent agent (101). The processing of the retrieved theoretical text fragments includes at least simplification processing, unified format processing, privacy information removal processing, and illegal and non-compliant content avoidance processing.

6. The multi-agent system for educational statistics according to claim 5, characterized in that, When the second execution module (1022) fails to retrieve a theoretical text fragment corresponding to the keyword information from the theoretical knowledge base (1021), the second execution module (1022) is also used to: A second guidance message is generated based on a pre-set second guidance response template and sent to the interaction module (1011). This second guidance message guides the user to re-enter requirement information that better matches their theoretical knowledge learning needs; and / or The system searches for publicly available relevant information using a pre-built search tool, provides a brief description of the information, and then sends the brief description and the link to the relevant information back to the first intelligent agent (101) for output to the user.

7. The multi-agent system for educational statistics of claim 5, wherein, The third intelligent agent (103) includes: The case knowledge base (1031) stores various practical cases related to educational statistics; The third execution module (1032) is used to retrieve case text fragments corresponding to keyword information from the case knowledge base, perform structured processing on the case text fragments to obtain practical case data that meets user needs, and then feed the practical case data back to the first intelligent agent (101). The practical case data obtained by structured processing includes at least case exercises, data software operation steps, and data result analysis. In addition, when the third execution module (1032) fails to retrieve a case text fragment corresponding to the keyword information in the case knowledge base (1031), the third execution module (1032) is also used to: The third guidance information is generated according to the internally preset third guidance response template and sent to the interaction module (1011). The third guidance information is used to guide the user to re-enter the requirement information that is more in line with the case operation learning requirement question, or to guide the user to request a new practical case, which is generated by the third intelligent agent (103). Accordingly, the third intelligent agent (103) also includes: The case replication module (1033) is used to associate with the case knowledge base (1031), replicate the practical cases in the case knowledge base, generate new practical cases, and store them. The replication of practical cases in the case knowledge base includes at least the following operations: Analyze the data structure of the corresponding practical cases, extract the variable relationships and statistical analysis logic, and retain the variable coding rules of the original data; Based on the analysis results, data samples matching the user's case study learning needs are extracted, and fictitious data that conforms to the logic of educational statistics is generated. Without changing the number of variables, variable types, and core analysis methods of the original practical case data, the variables and analysis objectives of the original practical case data are reasonably replaced and expanded.

8. The multi-agent system for educational statistics according to claim 7, characterized in that, In the theoretical knowledge base (1021) and the case knowledge base (1031), the content of the corresponding theoretical knowledge and practical cases is stored in the form of text fragments. The text fragments are divided according to the length of the text, and the next paragraph reproduces 10%-15% of the content of the previous paragraph. Furthermore, each text fragment in the theoretical knowledge base (1021) and the case knowledge base (1031) is converted into vector form; Accordingly, the second execution module (1022) and the third execution module (1032) can respectively convert keyword information into vector distance and measure the similarity of content by calculating the distance between vectors, thereby matching in the corresponding theoretical knowledge base (1021) and case knowledge base (1031) to obtain the corresponding text fragments.

9. The multi-agent system for educational statistics of claim 5, wherein, The fourth intelligent agent (104) includes: The fourth execution module (1041) is used to retrieve matching software data based on keyword information and feed the software data back to the first intelligent agent (101). The software data includes at least a URL link, download assistance steps, and download version suggestions.

10. A method for using a multi-agent system applied to educational statistics, characterized in that, Includes the following steps: Step 1: Obtain the user's description of the needs regarding educational statistics issues through the first intelligent agent (101), process the needs information to generate keyword information corresponding to the user's needs, determine the corresponding intelligent agent that can handle the user's needs based on the keyword information, and send the keyword information to the intelligent agent. The user's needs include theoretical knowledge learning needs, case operation learning needs, and software download needs related to educational statistics. Step 2: The second intelligent agent (102) responds to the received keyword information, obtains the corresponding original theoretical knowledge data, and feeds back the theoretical knowledge data to the first intelligent agent (101); and / or The third intelligence (103) responds to the received keyword information, obtains the corresponding educational statistics practice case, and sends the educational statistics practice case to the first intelligence agent (101); and / or The fourth agent (104) responds to the received keyword information, obtains the corresponding software data, and sends the software data to the first agent (101). Step 3: The first agent (101) outputs the results obtained by other agents in processing the corresponding keyword information to the user.

Images