Education intelligent analysis method and system, electronic equipment and storage medium

Abstract

The invention discloses an education intelligent analysis method and system, electronic equipment and a storage medium, and relates to the technical field of education intelligent analysis, and the method comprises the steps: carrying out the multi-stage semantic analysis of natural language query, and outputting a special structured expression for representing the query intention in the education field; determining a data source, a data aggregation mode and a result visualization type required by intelligent education analysis according to semantic elements contained in the special structured expression in the education field, and forming a data analysis execution plan; and according to the data analysis execution plan, carrying out one-hop or limited multi-hop graph relation expansion in the education knowledge graph by taking a standard node in the education standard system as a core, obtaining an education data node set associated with the query intention, and then executing a statistical aggregation operation to generate an intelligent analysis result. According to the method, natural language query in the education field can be accurately understood, an executable analysis plan can be stably generated, and education standard system data can be effectively integrated.

Description

Technical Field

[0001] This invention relates to the field of educational intelligent analysis technology, and in particular to an educational intelligent analysis method, system, electronic device, and storage medium. Background Technology

[0002] Current educational data analysis systems face technical challenges such as inaccurate understanding of natural language queries and difficulties in integrating across different standards systems. Users desire to directly obtain cross-disciplinary and cross-grade analysis results using natural language, but existing systems have limited ability to interpret complex educational intentions and struggle to accurately capture multi-dimensional semantic elements in queries. Furthermore, educational data is often scattered across different standards systems, lacking a unified knowledge representation method, resulting in inconsistent and uninterpretable analysis results. These issues severely impact the practicality and user experience of intelligent educational analysis systems.

[0003] Existing technologies primarily employ two approaches to address the aforementioned problems. One is template-based natural language processing, which matches user input with predefined query templates and then converts it into structured query statements. The other is to directly use large language models for end-to-end query parsing, converting natural language into database query or analysis commands. Some systems also attempt to construct knowledge graphs for the education domain, enhancing semantic understanding capabilities through entity recognition and relation extraction. These methods, to varying degrees, provide data analysis capabilities for natural language interactions.

[0004] However, existing technologies have significant drawbacks. Template-based methods lack flexibility and struggle to cover the diverse query expressions used in the education field. While methods using large language models offer broad coverage, their output stability is poor, they are prone to ambiguity in complex query scenarios, and they lack sufficient consideration for education-specific standards. Existing knowledge graph applications are mostly limited to simple entity queries, lacking deep integration with education standards systems and failing to support complex cross-system, multi-hop analysis needs. These limitations restrict the feasibility of existing solutions in real-world educational environments.

[0005] To address the shortcomings of existing technologies, there is an urgent need for an intelligent analysis method capable of accurately understanding natural language queries in the education field, stably generating executable analysis plans, and effectively integrating data from the education standards system. This invention presents an innovative solution in this context. Summary of the Invention

[0006] The technical problem to be solved by this invention is to address the shortcomings of existing technologies, and specifically provides an educational intelligent analysis method, system, electronic device, and storage medium, as detailed below: 1) In a first aspect, the present invention provides an educational intelligent analysis method, the specific technical solution of which is as follows: Receive natural language queries from users; Perform multi-stage semantic parsing on natural language queries, output a structured expression specific to the education domain that represents the query intent, and generate the confidence score of the structured expression specific to the education domain. The structured expression specific to the education domain contains parsed semantic elements. Based on the semantic elements contained in the specialized structured expressions in the field of education, determine the data sources, data aggregation methods, and result visualization types required for educational intelligent analysis, and formulate a data analysis execution plan; According to the data analysis execution plan, the graph relationship is expanded in the education knowledge graph with standard nodes in the education standard system as the core, using one-hop or limited multi-hop graph relationships, to obtain a set of education data nodes associated with the query intent; Perform statistical aggregation operations on the set of educational data nodes to generate intelligent analysis results, which include at least one of the following: charts, text summaries, and statistical data.

[0007] The beneficial effects of the educational intelligence analysis method provided by this invention are as follows: Multi-stage semantic parsing effectively improves the accuracy of understanding natural language queries, outputting a specialized structured expression representing the query intent in the education domain and generating confidence scores to ensure high reliability and evaluability of the parsing results. This parsing method combines dictionary matching, rule matching, fuzzy matching, and large language models to reliably capture semantic elements in complex educational intents, avoiding ambiguity and errors. Based on the semantic elements in the specialized structured expression in the education domain, the data sources, data aggregation methods, and result visualization types required for intelligent educational analysis are precisely determined, forming a structured data analysis execution plan to ensure the orderliness and executability of the analysis process. By expanding the graph relationships in the educational knowledge graph with standard nodes in the educational standards system as the core through one-hop or limited multi-hop methods, the set of educational data nodes associated with the query intent can be efficiently obtained, supporting deep data integration across standards systems. Finally, statistical aggregation operations are performed on the set of educational data nodes to generate intelligent analysis results including charts, text summaries, and statistical data, providing comprehensive and intuitive analysis outputs to meet diverse educational analysis needs. Overall, the technical solution enhances the interpretability, stability, and engineering feasibility of educational intelligent analysis, and solves the problems of inaccurate query understanding and difficulties in cross-system integration in existing technologies.

[0008] Based on the above scheme, the educational intelligence analysis method of the present invention can be further improved as follows.

[0009] Furthermore, it also includes: during the execution of the data analysis execution plan, pushing the task progress of each step in the data analysis execution plan to the user in real time through the WebSocket interface, and pushing a task snapshot containing the current execution status to the user when the user initiates a subscription.

[0010] The beneficial effects of adopting the above-mentioned further solution are as follows: Existing educational analysis systems lack a real-time progress feedback mechanism, making it impossible for users to understand the current status of long-running tasks. This invention uses a WebSocket interface to push the task progress of each step in the data analysis execution plan to users in real time, enabling them to continuously perceive the analysis progress. When a user subscribes, a task snapshot containing the current execution status is pushed, providing a complete instantaneous status view. This design enhances observability, allowing users to understand the task execution status promptly, avoiding the uncertainty of long waiting times, and providing detailed status information for problem diagnosis, effectively improving user experience and transparency.

[0011] Furthermore, it also includes: after all data analysis execution plans have been completed, sending the intelligent analysis results to the business front end via a REST interface so that the business front end can render and display them.

[0012] The beneficial effects of adopting the above-mentioned further solution are as follows: Existing educational analysis systems suffer from inconsistent interfaces and unstable data transmission during the result delivery stage, affecting the front-end display effect. This invention, after the entire data analysis execution plan is completed, sends the structurally complete intelligent analysis results to the business front-end through a standardized REST interface. This design ensures that analysis results such as charts, text summaries, and statistical data can be transmitted stably in a unified format, enabling the business front-end to accurately parse and render the data, effectively improving the reliability of result delivery and the consistency of front-end display, and enhancing the data consumption experience for end users.

[0013] Furthermore, multi-stage semantic parsing includes dictionary matching, rule matching, fuzzy matching, and collaboration with large language models.

[0014] The beneficial effects of adopting the above-mentioned further solution are as follows: Existing educational analysis systems often rely on a single technology for parsing natural language queries, leading to inaccurate or unstable understanding of complex educational intentions. This invention comprehensively utilizes the advantages of each method through multi-stage semantic parsing, which combines dictionary matching, rule matching, fuzzy matching, and a large language model. Dictionary matching ensures accurate identification of domain terms, rule matching guarantees correct parsing of grammatical structures, fuzzy matching handles expression variations, and the large language model provides deep semantic understanding. This collaborative working mode significantly improves the accuracy and coverage of structured expressions specific to the educational domain, effectively handles diverse natural language queries, and provides a reliable foundation for subsequent analysis processes.

[0015] 2) Secondly, the present invention also provides an intelligent educational analysis system, the specific technical solution of which is as follows: It includes a receiving module, a semantic parsing module, a plan generation module, a graph relationship expansion module, and an intelligent analysis result generation module; The receiving module is used to: receive natural language queries input by the user; The semantic parsing module is used to: perform multi-stage semantic parsing on natural language queries, output a structured expression specific to the education domain that represents the query intent, and generate the confidence score of the structured expression specific to the education domain, wherein the structured expression specific to the education domain contains parsed semantic elements; The plan generation module is used to: determine the data sources, data aggregation methods, and result visualization types required for educational intelligent analysis based on the semantic elements contained in the specialized structured expressions in the education field, and to form a data analysis execution plan; The graph relationship extension module is used to: based on the data analysis execution plan, extend the graph relationship in the education knowledge graph with standard nodes in the education standard system as the core, and obtain a set of education data nodes associated with the query intent. The intelligent analysis result generation module is used to perform statistical aggregation operations on the set of educational data nodes and generate intelligent analysis results, which include at least one of the following: charts, text summaries, and statistical data.

[0016] Based on the above solution, the educational intelligent analysis system of the present invention can be further improved as follows.

[0017] Furthermore, it also includes a task progress push module, which is used to: push the task progress of each step in the data analysis execution plan to the user in real time through the WebSocket interface during the execution of the data analysis execution plan, and push a task snapshot containing the current execution status to the user when the user initiates a subscription.

[0018] Furthermore, it also includes a rendering and display module, which is used to send the intelligent analysis results to the business front end through a REST interface after all data analysis execution plans have been completed, so that the business front end can render and display them.

[0019] Furthermore, multi-stage semantic parsing includes dictionary matching, rule matching, fuzzy matching, and collaboration with large language models.

[0020] 3) In a third aspect, the present invention also provides an electronic device, the electronic device including a processor coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor, so as to enable the electronic device to implement any of the above-mentioned educational intelligent analysis methods.

[0021] 4) In a fourth aspect, the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it implements any of the above-mentioned educational intelligent analysis methods.

[0022] It should be noted that the beneficial effects of the technical solutions of the second to fourth aspects of the present invention and their corresponding possible implementations can be found in the above description of the technical effects of the first aspect and its corresponding possible implementations, and will not be repeated here. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments of the present invention will be briefly introduced below: Figure 1 This is a flowchart illustrating an educational intelligent analysis method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an educational intelligent analysis system according to an embodiment of the present invention. Detailed Implementation

[0024] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0025] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0026] like Figure 1 As shown, an educational intelligence analysis method according to an embodiment of the present invention includes the following steps: S1. Receive natural language queries input by the user, specifically: S10. Provide a user input interface where users can input natural language queries through input components in a graphical user interface. The graphical user interface is typically a web page or mobile application containing text input fields and a submit button. The input fields are designed to support multi-line text, allowing users to enter longer queries. Additionally, the interface may integrate speech recognition functionality, converting voice input into text via a browser or device API. The input interface also includes prompts to guide users in entering valid natural language queries, such as example queries or formatting requirements.

[0027] S11. When the user completes input and triggers the submit operation, the front-end application uses a client-side scripting language such as JavaScript to listen for the event. The script retrieves the text content from the input field and performs basic validation. Validation includes checking whether the input is empty, contains illegal characters, or exceeds the maximum length limit. If the input is invalid, the front-end application displays an error message and prompts the user to re-enter the information. After successful validation, the front-end application stores the natural language query text in a temporary variable, ready for transmission.

[0028] S12. The front-end application encapsulates the natural language query as a JSON payload, for example, using key-value pairs such as {"query":"the natural language query text entered by the user"}. The front-end application uses an HTTP client library such as the Fetch API or Axios to initiate a POST request, sending the payload to a specific REST endpoint of the back-end service. The request header sets the content type to application / json and includes an authentication token to ensure secure transmission. The request process is handled asynchronously to avoid blocking the user interface and provides a loading indicator to enhance the user experience.

[0029] S13. The backend service uses a web framework such as Spring Boot or Flask to expose REST endpoints and listen for incoming HTTP requests. When a request arrives, the framework parses the request headers to verify authentication and permissions, and then extracts the JSON payload from the request body. The backend service deserializes the JSON data, obtains the natural language query text, and performs further validation, such as checking encoding consistency or preventing SQL injection attacks. After successful validation, the natural language query is stored in a memory buffer or message queue, marked as pending, and triggers subsequent calls to the semantic parsing module.

[0030] Natural language queries refer to queries expressed by users in everyday language, such as sentences in Chinese or English, rather than structured query languages ​​like SQL. This query method is more in line with human habits, but requires parsing to understand its intent. Natural language queries are typically input in text form, containing the user's specific needs for educational data analysis, such as "showing the top ten students in math" or "comparing the average English scores of different classes." By receiving natural language queries, subsequent semantic parsing and analysis processes are initiated to generate intelligent analysis results.

[0031] S2. Perform multi-stage semantic parsing on the natural language query, outputting an education-specific structured representation of the query intent, and generating the confidence score of the education-specific structured representation. The education-specific structured representation includes parsed semantic elements. Multi-stage semantic parsing includes dictionary matching, rule matching, fuzzy matching, and collaboration with a large language model. The specific implementation process is as follows: S20. Perform dictionary matching, loading a predefined education domain dictionary containing standardized terms and phrases, such as subject names like mathematics or English, educational entities like students or classes, and operation types like comparison or statistics. The dictionary is stored in a hash table structure for fast lookup. Natural language queries are processed by word segmentation, using algorithms such as maximum matching to divide the query text into sequences of words or phrases. Each segmentation result is precisely matched against a dictionary entry to identify domain-related concepts. Successfully matched terms are marked as candidate semantic elements, and the matching position and frequency are recorded. For example, for the query "show the top ten students in math," dictionary matching identifies "mathematics" as a subject entity, "grades" as an attribute, and "students" as an entity. The matching process also handles synonym mapping, unifying variant expressions such as "grades" and "scores" into standard terms through a thesaurus.

[0032] S21. Application rule matching uses a predefined set of syntactic rules built upon common query patterns in the education field. For example, a rule might be defined as "display [subject][attribute]" or "compare [entity] [metrics]". The rule engine parses the syntactic structure of the natural language query, using dependency parsing or regular expression matching to extract semantic elements that conform to the rules. The rule matching process considers the syntactic order and dependencies of the query. For example, in the query "compare the average English scores of different classes", the rule identifies "compare" as the operation, "class" as the entity, "English" as the subject, and "average score" as the statistical metric. The rule set is stored in XML or JSON format, supporting dynamic updates and expansion. The matching results are merged with the dictionary output to form a preliminary semantic framework, and element types and roles are labeled.

[0033] S22. Perform fuzzy matching to handle imprecise expressions, spelling errors, or near-identical terms in natural language queries. String similarity algorithms such as edit distance or cosine similarity are used to calculate the similarity between query segments and dictionary entries. Fuzzy matching sets a similarity threshold; for example, terms with an edit distance less than or equal to 2 are considered matches, and the most relevant standard term is returned. For example, in the query "show the top ten math scores", "scores" is corrected to "achievements" through fuzzy matching. Fuzzy matching also utilizes contextual information, such as evaluating the influence of adjacent words using an n-gram model, to improve matching accuracy. The matching results are integrated with the output of previous steps, missing or ambiguous elements are supplemented, and the similarity score for each match is recorded as input for subsequent confidence calculations.

[0034] S23. A large language model, such as a pre-trained Transformer-based model, is invoked, with the input being the natural language query and the parsing results from the first three steps. The large language model performs deep semantic analysis, handling complex or ambiguous queries and outputting enhanced semantic parsing. The model employs a dynamic routing mechanism, selecting a lightweight or heavyweight model instance based on query complexity, context length, and latency budget. For example, a lightweight model is used for simple queries to reduce latency, while a heavyweight model is used for complex queries to improve quality. The large language model output includes entity links, relation extraction, and intent classification, such as parsing "top ten" as a ranking operation and constraints. The model generates probability scores representing the reliability of the parsing and integrates them with the results from other stages.

[0035] S24 generates a structured representation specific to the education domain. It integrates the output of all parsing stages to construct a structured data representation, using JSON or a custom domain-specific language format. This education-specific structured representation contains parsed semantic elements, such as entities, attributes, operations, and constraints, and is organized according to educational standards. For example, the query "display the top ten students in math" is transformed into a structured representation containing element types such as "Subject: Math", "Attribute: Grade", "Operation: Sort", "Constraint: 10", and "Entity: Student". This structured representation ensures machine readability and executability, supporting subsequent workflow orchestration.

[0036] S25 calculates the confidence score of the specialized structured representation. A weighted formula is used to combine the scores from multiple stages of analysis to generate the overall confidence score. The confidence score formula is defined as follows: in, This represents the confidence level of the specialized structured representation, with a value ranging from 0 to 1. The score represents the dictionary match score, based on the number and importance of the matched terms; The score represents the rule matching score, based on rule compliance and coverage. The score represents the fuzzy match score, based on similarity value and contextual consistency. This represents the probability score output by the large language model; , , , These are weighting coefficients, satisfying... The weights are adjusted through training on historical data to optimize confidence accuracy. After confidence calculation, calibration is performed, using the sigmoid function for smoothing to reduce the impact of extreme values.

[0037] Query intent refers to the deeper goals or purposes expressed by users through natural language queries, such as wanting to obtain analytical results for specific educational data. Query intent reflects the user's information needs and drives the execution of corresponding data analysis tasks. During the parsing process, it is necessary to accurately identify query intent and translate it into executable instructions to ensure that subsequent analysis processes meet user expectations.

[0038] Among them, the education-specific structured representation is a formalized data structure used to represent the meaning of natural language queries in the education domain. This structured representation uses a domain-specific language, contains standardized elements and relationships, and is easy to process and analyze. For example, it may include entity types, attributes, operators, and constraints, encapsulate parsed semantic information, and support workflow orchestration and data aggregation.

[0039] The confidence score of the specialized structured representation is a numerical metric that measures the accuracy and reliability of the educational domain-specific structured representations output during the parsing process. The confidence score is calculated based on the matching quality during the parsing phase, the model output probability, and other factors, and is used to assess the credibility of the parsing results. A high confidence score indicates that the parsing result is likely correct, while a low confidence score may trigger subsequent processing such as backtracking or calibration to improve output stability.

[0040] The parsed semantic elements are discrete units of meaning extracted from natural language queries, forming the basis of structured expressions specific to the education domain. These elements include entities, attributes, relationships, operations, and conditions, such as subject names, statistical indicators, time ranges, or comparison operations. The parsed semantic elements are organized in a structured manner, supporting precise data analysis and ensuring that query intent is fully expressed within the domain context.

[0041] S3. Based on the semantic elements contained in the specialized structured expressions for the education field, determine the data sources, data aggregation methods, and result visualization types required for educational intelligent analysis, and formulate a data analysis execution plan. The specific implementation process is as follows: S30. Parse the semantic elements in the education-specific structured representation. Load the education-specific structured representation, stored in JSON format, containing key-value pairs representing semantic elements. The parsing process uses a standard JSON parsing library, such as Jackson or Gson, to traverse all fields and extract key semantic elements, such as entity type, attributes, operations, and constraints. For example, for the semantic element "Subject:Mathematics", the entity type is identified as "Subject" and the attribute value as "Mathematics"; for the semantic element "Operation:Sorting", the operation type is identified as "Sorting". The parsing process also handles nested structures, such as conditional statements or compound queries, ensuring that all semantic elements are fully extracted and categorized and stored in internal data structures, such as hash tables or object lists. This step provides the foundational data for subsequent decisions and verifies the completeness and consistency of the semantic elements.

[0042] S31 involves mapping data sources based on semantic elements, maintaining a data source mapping table stored in YAML or JSON format, and associating semantic elements with physical data sources. The data source mapping table contains entries; for example, subject semantic elements map to subject standard nodes in the knowledge graph, and grade semantic elements map to student grade database tables. The data source mapping table is queried based on the parsed semantic elements, using string matching or regular expressions to determine the required data source. For example, if the semantic element contains "entity: student" and "attribute: grade", the student information database and grade table are selected as data sources; if the semantic element contains "subject: English", the English-related standard nodes in the knowledge graph are selected. The mapping process also considers data source priority and availability, verifying the data source status through a health check API to ensure that the selected data source is accessible and has sufficient permissions.

[0043] S32. Determine the data aggregation method based on semantic elements. Access a predefined data aggregation method library, which contains aggregation functions such as average, sum, count, maximum, minimum, as well as grouping and sorting operations. The data aggregation method library exists as a code module or configuration file, for example, using SQL aggregation functions or custom algorithms. Match the data aggregation method using a rule engine based on the operations and statistical requirements in the semantic elements. For example, if the semantic element contains "operation: average", select the average aggregation function; if the semantic element contains "constraint: top ten", select sorting aggregation and set the constraint to 10. The determination process also involves parameter settings, such as grouping fields or filtering conditions, and defining the aggregation logic using formulas. For example, for average aggregation, the formula is expressed as: ,in, This represents the value of the i-th data point. This indicates the total number of data points. Record the selected data aggregation method and its parameters, and ensure they are compatible with the data source.

[0044] S33. Determine the visualization type of the results based on semantic elements. Refer to a visualization type mapping table, which is based on best practices in education, to associate semantic elements with appropriate visualization formats. The visualization type mapping table is stored in JSON format and contains rules such as comparison operations corresponding to bar charts, trend analysis corresponding to line charts, and proportion distribution corresponding to pie charts. Query this mapping table based on the operations and attributes in the semantic elements, and use a pattern matching algorithm to determine the visualization type of the results. For example, if the semantic elements contain "Operation: Comparison" and "Entity: Class", select a bar chart; if the semantic elements contain "Attribute: Grade" and "Operation: Time Trend", select a line chart. The determination process also considers user context, such as device type or historical preferences, and optimizes the selection through A / B testing. Output the visualization type of the results and its configuration parameters, such as chart title, axis labels, or color scheme.

[0045] Step S34 involves integrating information to form a data analysis execution plan. The outputs of the first four steps are combined into a complete data analysis execution plan. This plan is in JSON format and includes fields such as a list of data sources, a sequence of data aggregation methods, the type of result visualization, and execution metadata. A template engine or code generator is used to construct the data analysis execution plan, ensuring structural consistency. For example, the data analysis execution plan might include: specifying the knowledge graph and database tables in the data source section; defining the average calculation and grouping operations in the data aggregation method section; and specifying the bar chart rendering in the result visualization type section. Error handling logic is also added, such as rolling back to the data source or default aggregation method, and execution timeout and logging parameters are set. The data analysis execution plan is serialized into a file or message and stored in a distributed cache or message queue for subsequent use by the aggregation analysis module.

[0046] In this context, the data source refers to the original source of the data required in the educational intelligence analysis process. This data may be stored in various systems, such as relational databases, knowledge graphs, data warehouses, or external APIs. Data sources include specific datasets, such as student grade sheets, course information databases, and subject standard nodes. Appropriate data sources are selected based on the query intent to extract relevant data, ensuring that the analysis is based on accurate and complete original information.

[0047] Data aggregation refers to methods used to statistically summarize or transform data obtained from data sources, such as calculating averages, summing, counting, grouping, sorting, or filtering. Data aggregation transforms raw data into meaningful statistical results, such as class averages or subject rankings, to support in-depth analysis and decision-making. These methods are typically implemented using mathematical functions or algorithms, taking into account data characteristics and query requirements.

[0048] Among them, result visualization types refer to the way analysis results are presented in the form of graphics or charts, such as bar charts, line charts, pie charts, scatter plots, or text summaries. Result visualization types help users intuitively understand data analysis results, improve readability and interactivity, and the most appropriate visualization form should be selected based on the analysis content and user habits to effectively convey information.

[0049] The data analysis execution plan is a structured sequence of instructions that details the entire process of educational intelligent analysis, including data source selection, data aggregation steps, visualization rendering sequence, and execution parameters. The data analysis execution plan ensures the analysis process is orderly, efficient, and repeatable, and outputs stable domain intelligence analysis results. It is typically stored in a machine-readable format, such as JSON or XML, for easy parsing and execution.

[0050] S4. Based on the data analysis execution plan, expand the graph relationships in the education knowledge graph using standard nodes in the education standard system as the core, either in one hop or with a limited number of hops, to obtain a set of education data nodes associated with the query intent. The specific implementation process is as follows: S40. Parse the graph query requirements in the data analysis execution plan. Read the data analysis execution plan and extract the specified standard node identifiers and relation types. The data analysis execution plan is stored in JSON format and contains standard node fields, such as "standard_node_id":"MATH_ALGEBRA_001", and relation constraint fields. The parsing process uses a JSON parsing library to traverse the execution plan structure and identify the core elements that need to be queried in the educational knowledge graph. Verify the validity of the standard node identifiers by querying the graph metadata to confirm the existence and accessibility of the nodes. The parsing process also extracts extended depth parameters, such as the specific values ​​of one hop or finite multi-hop, as well as relation types or node categories that need to be excluded.

[0051] S41. Initiate a graph traversal query centered on the standard node. Construct the traversal query statement using a graph query language such as Cypher or Gremlin. The query statement starts from the standard node, specifies the relationship direction and type, and sets the maximum expansion depth. For example, for the standard node "MATH_ALGEBRA_001", the query might be defined as: starting from this node, expanding outwards along relationship types such as "belongs to system", "contains discipline", and "corresponding unit". When executing the query, parameterization is used to prevent injection attacks, and timeout limits are set to ensure query performance. The traversal process considers the characteristics of the education standard system, prioritizing the retention of relationship paths relevant to education analysis and filtering out irrelevant connections.

[0052] S42. Perform one-hop or finite multi-hop relation expansion. Based on the query, perform a breadth-first or depth-first graph traversal algorithm in the educational knowledge graph. One-hop expansion starts from a standard node and only visits directly adjacent nodes; finite multi-hop expansion allows visiting nodes at greater distances, but limits the number of hops to a preset value, such as three or five hops. During the traversal, record the access path and store each encountered node and its attributes. For each visited node, evaluate its relevance to the query intent using a relevance scoring formula: in, The relevance score of a node to the query intent. Indicates the node type matching degree. Indicates the relevance of node content. and It is a weighting coefficient, and satisfies Based on this score, low-relevance nodes are filtered out, and high-relevance nodes are retained to enter the candidate set.

[0053] S43. Organize all candidate nodes obtained during the traversal into an educational data node set. This set is stored in a list structure, with each element containing a node identifier, node type, node attributes, and the relationship path with the original standard nodes. The set is deduplicated to ensure that each node appears only once. Simultaneously, nodes are sorted according to their topological importance in the educational knowledge graph, and node weights are calculated using centrality algorithms such as degree centrality or proximity centrality. The organization process also verifies the data integrity of the nodes, checking for the existence of necessary attributes, such as student node grade data or course node lesson hour information.

[0054] S44. Verify the match between the educational data node set and the query intent. Compare the educational data node set with the original query intent to assess the completeness and accuracy of the set's coverage. The verification process uses semantic similarity calculation to compare the semantic range represented by the node set with the expected range of the query intent. Calculate the match index: in, Indicates the percentage of match. This indicates the number of nodes directly related to the query intent. This represents the total number of nodes in the education data node set. If the match is below the threshold, the expansion process will be re-executed, adjusting the relationship type or expansion depth parameters. After successful validation, the education data node set is marked as valid and ready for subsequent statistical aggregation operations.

[0055] Educational knowledge graphs are a data representation of knowledge in the education field organized in a graph structure. They consist of nodes and edges, where nodes represent educational entities and edges represent relationships between entities. Educational knowledge graphs integrate educational data from multiple sources, including curriculum content, student information, teaching resources, and assessment standards, forming an interconnected knowledge network. This type of graph supports complex semantic queries and relational reasoning, providing a structured data foundation for intelligent educational analysis.

[0056] In this context, standard nodes within the education standards system serve as core anchors in the education knowledge graph, representing standardized concepts within the official education standards system. These nodes typically correspond to curriculum standards, competency indicators, or learning objectives, such as "basic algebra in mathematics" or "English reading ability." Standard nodes possess unique identifiers and standardized attribute descriptions, serving as the starting point for graph expansion and ensuring that data analysis remains consistent with education standards.

[0057] The education data node set is a group of related nodes obtained through graph relationship expansion, and these nodes are directly associated with the query intent. The education data node set contains various types of educational entities, such as student nodes, class nodes, and course nodes, each with its own attributes and relationships. This set provides the data foundation for subsequent statistical aggregation operations, ensuring that the analysis results comprehensively reflect the query requirements.

[0058] S5. Perform statistical aggregation operations on the set of educational data nodes to generate intelligent analysis results. The intelligent analysis results include at least one of the following: charts, text summaries, and statistical data. The specific implementation process is as follows: S50. Extract attribute data from the educational data node set. Traverse each node in the educational data node set, access the node's attribute fields, and extract numerical and categorical data relevant to statistical aggregation. For example, for student type nodes, extract grade attribute values; for course type nodes, extract class hour attribute values ​​and subject classification information. Create a temporary data table to store the extracted attributes; the table structure includes fields such as node identifier, attribute name, attribute value, and data type. The extraction process performs data cleaning, handling missing and outlier values, using interpolation methods to fill in blank data, and applying outlier detection algorithms to identify abnormal records. The cleaned data is converted to a unified format, such as converting numerical data to floating-point numbers and categorical data to standard encoding, preparing standardized input data for subsequent statistical aggregation operations.

[0059] S51. Perform statistical calculations defined by the data aggregation method. Based on the data aggregation method specified in the data analysis execution plan, perform corresponding statistical operations on the extracted attribute data. Data aggregation methods include basic statistical calculations such as average, summation, count, maximum, and minimum values, as well as complex calculations such as group aggregation, sorting, and percentage calculations. Using the statistical calculation engine, apply the data aggregation method to the dataset item by item. For example, for average aggregation, calculate the arithmetic mean of the specified attribute; for group aggregation, calculate the statistics for each group after grouping according to the specified categorical attribute. The calculation process is expressed using mathematical formulas, such as the average calculation: in, This represents the average value. Indicates the total number of data points. Indicates the first The values ​​of each data point. After performing all data aggregation methods, the calculation results are stored in an intermediate data structure, including a summary of the original data and statistical output.

[0060] S52. Generate chart data according to the data analysis execution plan. Based on the visualization type determined in the data analysis execution plan, convert statistical calculation results into chart data. The chart data uses the format required by standard chart libraries, such as the JSON structure supported by ECharts or Chart.js. Map statistical results to chart elements, such as mapping grouped averages to bar height data in a bar chart, and mapping time series data to point coordinates in a line chart. Configure chart style parameters during the generation process, including color scheme, axis labels, legend position, and animation effects. For example, for a bar chart comparing class scores, generate a data structure containing a list of class names and an array of corresponding average scores, while setting the chart title to "Comparison of Average Scores of Each Class," the X-axis label to "Class Name," and the Y-axis label to "Average Score." The chart data ensures it contains all the information required for complete rendering and does not rely on additional front-end processing.

[0061] S53. Generate text summaries based on statistical calculation results. Using text generation templates and rules, key statistical findings are transformed into coherent natural language descriptions. The text generation process identifies salient features in the statistical results, such as maximum and minimum values, trend changes, and anomaly patterns, and selects appropriate language templates to organize this information. Education-specific expressions are applied, using educational terminology and suitable writing styles. For example, for performance analysis, the text summary might describe it as "The average score for fifth-grade math was 85, with the highest score of 95 achieved by Class 2 of Grade 3, and overall performance improved by 5% compared to the previous month." The text generation process also considers level of detail control, adjusting the summary length and information density according to user type and context to ensure the summary is both comprehensive and concise.

[0062] S54. Organize and encapsulate the intelligent analysis results. Integrate the generated chart data, text summaries, and statistical data into a unified intelligent analysis result structure. The intelligent analysis results are organized in JSON format, containing independent fields to store different types of output content. Verify the integrity and consistency of each component, checking the reasonableness of the data range and the correctness of the text syntax. Add metadata information during the encapsulation process, such as generation timestamps, data version identifiers, and result confidence scores. Finally, the intelligent analysis results are serialized into a standard JSON string, ready to be returned to the business front end via a REST interface. Record key metrics during the generation process, such as processing time, data coverage, and result quality assessment, for performance monitoring and optimization of the analysis process.

[0063] Charts are a form of data analysis that presents results graphically, using visual elements such as bars, lines, pie charts, or dots to express data relationships and patterns. Charts transform numerical information into intuitive visual representations, helping users quickly understand data distribution, trends, and comparative results. In educational intelligence analytics, charts can display student performance distribution, subject comparisons, or changes in learning progress.

[0064] Text summarization is an output format that presents analytical conclusions in the form of natural language paragraphs. It summarizes the core findings and key insights of data analysis through concise textual descriptions. Text summarization automatically extracts important information and organizes it into a coherent expression, enabling users to quickly grasp the key points of the analysis without needing to interpret the raw data. In educational intelligent analysis, text summarization can describe the overall performance of a class, the strengths and weaknesses of a subject, or an overview of learning trends.

[0065] Statistical data is an output format that presents analytical results in numerical form, encompassing various statistical indicators and measures. It provides precise numerical information, supports quantitative analysis and detailed comparisons, and typically includes statistical measures such as mean, sum, count, percentage, and standard deviation. In educational intelligence analysis, statistical data can display average scores, pass rates, ranking order, or score distribution ranges.

[0066] Optionally, the above technical solution also includes: S6. During the execution of the data analysis execution plan, the task progress of each step in the data analysis execution plan is pushed to the user in real time through the WebSocket interface. When the user subscribes, a task snapshot containing the current execution status is pushed to the user. The specific implementation process is as follows: S60. Establish a WebSocket connection on the same port as the HTTP service. Configure a WebSocket endpoint on the HTTP server, which shares the same network port and domain name as the main HTTP service. When a user accesses the educational intelligent analysis interface, the front-end application initiates a connection request using the WebSocket client library. The connection establishment process follows the WebSocket protocol handshake procedure, including an Upgrade header field in the initial HTTP request to upgrade the protocol to WebSocket. Verify user authentication, using the same token mechanism as the REST interface to ensure security. After a successful connection, the server maintains a WebSocket session object, associating it with the user's identity and the current analysis task. The front-end application listens for connection status events, handles network exceptions and automatic reconnection logic, ensuring connection stability.

[0067] S61. Track the execution status of each step in the data analysis execution plan. During the execution of the data analysis execution plan, create a status tracker for each step. The status tracker records the step's start time, end time, processing progress, and result code. Define a unified progress calculation model, assigning weight coefficients to each step to reflect its relative importance in the overall task. When a step's status changes, update the internal state storage and calculate the overall task progress. The overall progress calculation formula is expressed as: in, This indicates the overall progress of the data analysis execution plan, with a value ranging from 0 to 100%; This indicates the total number of steps in the data analysis execution plan; Indicates the first The weight coefficients of each step satisfy the following conditions: ; Indicates the first Individual progress at each step is also expressed as a percentage. The progress of each step is evaluated periodically. The value is based on the ratio of processed data to total data or the ratio of executed operations to total operations.

[0068] S62. Real-time task progress is pushed via the WebSocket interface. The progress push mechanism is triggered when the status of each step in the data analysis execution plan changes. The push process constructs a standard-format progress message, including the current step identifier, step progress percentage, overall progress percentage, and timestamp. The progress message is sent to the corresponding user using text frame format via an established WebSocket connection. An adaptive strategy is used for push frequency; messages are guaranteed at the beginning and end of each step, and triggered at fixed intervals or progress change thresholds during execution. For push failures, when a WebSocket connection error is detected, undelivered progress messages are stored in a persistent queue and resent after the connection is restored. Upon receiving the progress message, the front-end application updates the progress bar component and status indicators to provide visual feedback to the user.

[0069] S63. Handle user-initiated task snapshot subscription requests. Define a dedicated snapshot request message type on the WebSocket connection. When a user initiates a subscription through the interface, the front-end application sends a snapshot request message containing the task identifier to the server. After receiving the request, the server verifies the user's permissions and task ownership, ensuring that the user can only access snapshots of tasks they initiated. Collect current data to analyze the execution status of all steps in the execution plan, including the final results of completed steps and the real-time progress of ongoing steps. Also obtain task-level information, such as resource usage, the scale of data processed, and the estimated remaining time, and integrate this information to construct a complete task snapshot of the execution status.

[0070] S64. Generate and push task snapshots of execution status. Organize the collected status information into structured snapshot objects. Snapshot objects are in JSON format, containing task metadata fields such as task identifier, creation time, and overall progress, as well as a step status array field detailing the execution status of each step. Use a serialization library to convert the snapshot objects into JSON strings and push them to subscribing users via a WebSocket connection. Snapshot data ensures integrity and consistency, containing sufficient information for users to fully understand the task status. After receiving the snapshots, the front-end application parses them and displays detailed information on a dedicated monitoring interface, including a step list, progress charts, and performance metrics. A snapshot history query function is also provided, allowing users to view task status records from past points in time, supporting longitudinal comparison and analysis.

[0071] The steps in the data analysis execution plan refer to multiple ordered sub-tasks that constitute the complete educational intelligent analysis process. These steps are arranged in execution order and include semantic parsing, workflow orchestration, graph relationship expansion, and statistical aggregation. Each step has clear input and output definitions and specific processing logic. The entire educational intelligent analysis task is completed by coordinating the sequential execution of these steps.

[0072] The execution status refers to the current status information of each step in the data analysis execution plan during its operation. The execution status includes metadata such as step identifier, start time, end time, percentage of progress completed, success or failure flags, and error messages. The execution status reflects the real-time progress of the analysis task, allowing users to understand the current processing stage and the estimated remaining time.

[0073] The task snapshot in execution status is a complete record of the data analysis execution plan captured at a specific point in time. It contains the latest execution status of all steps, as well as overall task progress, resource usage, and performance metrics. The task snapshot provides a comprehensive, instantaneous view of the analysis task, supporting progress monitoring and problem diagnosis.

[0074] Optionally, the above technical solution also includes: S7. After the data analysis execution plan is fully executed, the intelligent analysis results are sent to the business front-end via the REST interface so that the business front-end can render and display them. The specific implementation process is as follows: S70. Verify the completion status of the data analysis execution plan. Monitor the execution status of all steps in the data analysis execution plan. When the successful completion of the final step is detected, trigger the completion status verification process. The verification process checks the execution record of each step to confirm that there are no failed steps and that all necessary outputs have been generated. Query the task status database to obtain the task completion timestamp and final execution summary. The verification process also includes an integrity check to ensure that the intelligent analysis results contain all necessary components, such as chart data, text summaries, and statistical data. Generate a verification report, recording details of passed and failed items. Only tasks that fully pass verification will enter the results sending process.

[0075] S71. Prepare and encapsulate the intelligent analysis results. Load the various components of the intelligent analysis results from the results store, including chart data, text summaries, and statistical data. Chart data retains its original JSON structure and contains complete visualization configuration information; text summaries are stored as UTF-8 encoded strings; statistical data is organized as a collection of key-value pairs. Encapsulate these components into a unified response object, organized according to a predefined result pattern. Metadata information is added during the encapsulation process, including task identifier, generation timestamp, result version number, and content checksum. Perform serialization on the encapsulated response object, converting it to a JSON-formatted string representation to ensure encoding consistency and data integrity.

[0076] S72. Construct the REST interface response, initialize the HTTP response object, set the status code to 200 to indicate success, and configure the content type header to application / json; the character set to utf-8. Set the serialized intelligent analysis result string as the response body content. The response construction process also includes setting security-related headers, such as the CORS header to allow cross-domain access from the business frontend domain, and the cache control header to prevent intermediate nodes from caching sensitive data. Calculate the response body size and set the Content-Length header, enabling chunked transmission encoding for large data volumes. After the response object is constructed, record the response summary information, including the response size, construction timestamp, and target frontend identifier.

[0077] S73. Send intelligent analysis results via REST interface. Using an HTTP server framework such as Spring Boot or Express.js, the constructed response object is returned to the business frontend through a predefined REST endpoint. The sending process follows the HTTP protocol specification, outputting the status line, response headers, and response body in sequence. It handles various network transmission scenarios, including connection timeouts, client interruptions, and bandwidth limitations. For large intelligent analysis results, it supports gzip compression to reduce the amount of data transmitted and automatically detects compression formats supported by the client. During the sending process, it monitors the transmission status, recording the start time, end time, and actual number of bytes transmitted to ensure complete delivery of results.

[0078] S74. Handle reception acknowledgments and error retries from the business frontend. After sending, wait for a reception acknowledgment response from the business frontend. After successfully receiving and parsing the intelligent analysis result, the business frontend returns a success response containing the task identifier and reception status. Record this acknowledgment information and update the task status to "delivered". If no acknowledgment response is received within the timeout period, or if an error response is received, initiate a retry mechanism. The retry process uses an exponential backoff strategy: the first retry waits for 1 second, the second waits for 2 seconds, and the waiting time doubles for each subsequent retry until the maximum number of retries is reached. Maintain a retry queue and persistently store the task information to be retried to ensure that retry tasks are not lost even if the system restarts.

[0079] S75. Clean up temporary resources and update task status. After confirming that the business front-end has successfully received the intelligent analysis results, perform resource cleanup operations. The cleanup process includes releasing memory resources occupied during result generation, deleting temporary storage files, and closing database connections that are no longer needed. Update the task management database, changing the task status from "execution completed" to "result returned," and record the result return time and receipt confirmation time. Also collect performance metrics for this execution, including result preparation time, network transmission time, and total processing time, and store them in the analysis log for subsequent performance optimization. The cleanup operation ensures efficient resource utilization and reserves sufficient resources for subsequent analysis tasks.

[0080] The following example details the implementation process of the educational intelligent analysis method. Based on a user query of "displaying the top ten students in math," the example demonstrates the complete technical flow from natural language input to result rendering, specifically including: S101. Receive natural language query input from the user. The user enters the query "Display the top ten students in math" into the text input box through the web interface of the education analytics platform. After capturing the user input, the front-end application sends the query text to the back-end service via an HTTP POST request. The back-end service's REST interface receives and verifies the query content. After confirming that the query is not empty and the format is valid, it stores the query text in the task queue, assigns a unique task identifier T001, and immediately returns a task reception confirmation response to the front-end.

[0081] S102. Perform multi-stage semantic parsing on the natural language query. The multi-stage semantic parsing process is initiated. First, dictionary matching is performed, using an education domain dictionary to identify the subject entity corresponding to "mathematics," the attribute corresponding to "grade," the entity corresponding to "student," and the sorting constraint corresponding to "top ten." Next, rule matching is performed, applying the grammatical rule "display [subject][attribute][constraint][entity]" to confirm the completeness of the query structure. Then, fuzzy matching is performed to check for possible spelling variations, such as the similarity between "mathematics" and "mathematics class." Finally, in conjunction with a large language model, a pre-trained Transformer model is used to deeply understand the query semantics, confirming that "top ten" precisely means taking the top 10 results in descending order of grades. The parsed output is a dedicated structured expression for the education domain, using JSON format containing complete semantic elements: {"operation":"display","subject":"mathematics","attribute":"grade","entity":"student","sorting":"descending","constraint":10}. The confidence score of the generated dedicated structured expression is 0.92, calculated based on the weighted average of the parsing quality at each stage.

[0082] S103. Develop a data analysis execution plan based on the specific structured representations used in the education field. Analyze the semantic elements in these representations to determine that the student grade database is the primary data source, and select the mathematics grade table as the specific dataset. Based on sorting and constraint semantic elements, determine that the data aggregation method is to sort by grade value in descending order and limit the return to 10 records. Based on display operation semantic elements, determine that the result visualization type is a table format. Integrate these decisions to form a complete data analysis execution plan, including data source location information, aggregation operation sequence, visualization configuration parameters, and execution priority settings.

[0083] S104. Perform graph relationship expansion in the educational knowledge graph. According to the data analysis execution plan, locate the standard node M001 for the mathematics subject in the educational knowledge graph. Using this standard node as the core, perform a one-hop expansion along the "belongs to subject" relationship direction to obtain all student nodes associated with the mathematics subject. The expansion process traverses the relationship edges in the knowledge graph, collecting student node identifiers that meet the conditions to form an educational data node set. The set contains a list of student nodes that meet the conditions, and each node contains basic information such as student identifier, name attribute, and grade attribute.

[0084] S105. Perform statistical aggregation operations on the educational data node set. Load the score attribute values ​​of each node in the educational data node set, and sort all math scores in descending order according to the aggregation method defined in the data analysis execution plan. Take the student records corresponding to the top 10 highest scores and generate statistical results containing name and score fields. Based on this result, generate three forms of intelligent analysis results: the chart data is prepared as a bar chart format, showing the comparison of the scores of the top ten students; the text summary is automatically generated as a descriptive text "The list of the top ten students in math scores is as follows..."; the statistical data is organized into a set of values ​​including average score, highest score, and score range.

[0085] S106. Real-time task progress is pushed via the WebSocket interface. During the execution of the data analysis execution plan, progress information is pushed to the user through the established WebSocket connection. When the semantic parsing step is completed, the progress message displays "Semantic parsing completed, progress 25%"; when the graph expansion step is completed, "Data acquisition completed, progress 60%" is pushed; when the aggregation operation is completed, "Analysis and calculation completed, progress 90%" is pushed. The user initiates a task snapshot request through the subscription button on the interface, and a task snapshot containing detailed status of each step is immediately returned, displaying the currently executing operation and the estimated remaining time.

[0086] S107. Return intelligent analysis results via REST interface. After the data analysis execution plan has been fully executed, the intelligent analysis results are encapsulated into a standard JSON response and sent to the business frontend via the REST interface. The response body contains three main parts: chart data, text summary, and statistical data, as well as task metadata such as generation time and result version. After receiving the results, the business frontend uses a chart library to render a performance comparison bar chart, displays a summary description in the text area, and displays statistical values ​​in the sidebar, completing a comprehensive visualization of the results.

[0087] In another embodiment, this invention discloses an educational intelligent analysis method centered on a semantic parser, a large model, and a four-step analysis pipeline. The method includes: performing multi-stage semantic parsing on natural language queries using a semantic parser. This multi-stage semantic parsing includes dictionary matching, rule matching, fuzzy matching, and collaboration with a large language model to produce a specialized structured expression for the education domain. The parsing confidence is then calibrated and gating. Subsequently, this specialized structured expression for the education domain is used as a constraint context and executed according to the four-step analysis pipeline, which includes semantic parsing, workflow orchestration, deep analysis, and visualization. During execution, the large language model is dynamically routed, selecting either a lightweight or heavyweight model based on query complexity to meet quality and latency objectives. Finally, stable domain-specific intelligent analysis results are generated for rendering.

[0088] The process involves: orchestrating workflows based on specialized structured expressions for the education domain; determining data sources, data aggregation methods, and result visualization types; and forming a data analysis execution plan. Following this plan, the data analysis knowledge graph is expanded using standard nodes from the education standards system as the core, performing one-hop or limited multi-hop expansions to obtain a set of education data nodes associated with the query intent. Statistical aggregation operations are then performed on this set of education data nodes to generate domain-specific intelligent analysis results, which include chart data, node link data, text summaries, and statistical analysis. During the execution of the data analysis execution plan, the progress of each step is pushed in real-time via a WebSocket interface on the same port as the HTTP service. A snapshot of the current execution status is pushed when a user subscribes. Finally, after the entire data analysis execution plan is completed, the generated domain-specific intelligent analysis results are returned via a REST interface for rendering by the business front-end.

[0089] The large language model employs dynamic routing between lightweight and heavyweight models, adaptively selecting the model type based on the completeness, ambiguity, context length, and latency budget of the specialized structured expressions for the education domain. When the confidence level of the semantic parsing result falls below a set threshold, gating and fallback operations are performed, including enhancing model capabilities or implementing dimensionality reduction constraints to ensure stable structured output. In cases of sparse data, domain-specific intelligent analysis results return empty structure placeholders to maintain stable analysis within the domain.

[0090] The educational knowledge graph uses standard nodes in the educational standards system as core anchors, extending one-hop relationships to educational system nodes, subject system nodes, sub-subject nodes, lesson unit nodes, lesson nodes, and lesson activity / assignment nodes. The method persistently stores process data and diagnostic information, including education-specific structured representations, confidence levels, routing decisions, execution summaries, and cost measurements.

[0091] In the confidence calibration process of structured expressions specific to the education field, a weighted comprehensive evaluation method is adopted. The confidence calculation formula is expressed as: in, This indicates the confidence level of a structured representation specific to the field of education. Indicates the dictionary matching score. This indicates the score for matching the rule. Indicates the fuzzy matching score. This represents the probability score output by the large language model. to The weighting coefficients for each stage score are defined, with a total weighting coefficient of 1. The optimal weighting configuration is obtained through training with historical analytical data. Confidence calibration ensures the accuracy and reliability of the analytical results, providing a basis for subsequent gating decisions.

[0092] In the above embodiments, although the steps are numbered S1, S2, etc., they are only specific embodiments given by the present invention. Those skilled in the art can adjust the execution order of S1, S2, etc. according to the actual situation. The scheme after adjusting the order is also within the protection scope of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0093] like Figure 2 As shown, an educational intelligent analysis system 200 according to an embodiment of the present invention includes a receiving module 201, a semantic parsing module 202, a plan generation module 203, a graph relationship expansion module 204, and an intelligent analysis result generation module 205; The receiving module 201 is used to: receive natural language queries input by the user; The semantic parsing module 202 is used to: perform multi-stage semantic parsing on natural language queries, output a structured expression specific to the education domain that represents the query intent, and generate the confidence score of the structured expression specific to the education domain, wherein the structured expression specific to the education domain contains parsed semantic elements; The plan generation module 203 is used to: determine the data sources, data aggregation methods and result visualization types required for educational intelligent analysis based on the semantic elements contained in the specialized structured expressions in the education field, and form a data analysis execution plan; The graph relationship extension module 204 is used to: based on the data analysis execution plan, extend the graph relationship in the education knowledge graph with standard nodes in the education standard system as the core, and obtain a set of education data nodes associated with the query intent; The intelligent analysis result generation module 205 is used to: perform statistical aggregation operations on the set of educational data nodes to generate intelligent analysis results, which include at least one of charts, text summaries, and statistical data.

[0094] Optionally, the above technical solution also includes a task progress push module, which is used to: push the task progress of each step in the data analysis execution plan to the user in real time through the WebSocket interface during the execution of the data analysis execution plan, and push a task snapshot containing the current execution status to the user when the user initiates a subscription.

[0095] Optionally, the above technical solution also includes a rendering and display module, which is used to: after all the data analysis execution plans have been executed, send the intelligent analysis results to the business front end through a REST interface so that the business front end can render and display them.

[0096] Optionally, in the above technical solution, multi-stage semantic parsing includes dictionary matching, rule matching, fuzzy matching, and collaboration with a large language model.

[0097] In another embodiment, the system includes a semantic parsing module, a workflow orchestration module, an aggregation analysis module, a visualization module, and a task and push module. The semantic parsing module contains a semantic parser that performs multi-stage semantic parsing and outputs a specialized structured representation and confidence level specific to the education domain. The workflow orchestration module generates a data analysis execution plan based on the specialized structured representation for the education domain. The aggregation analysis module performs relation expansion and statistical aggregation in the education knowledge graph, generating domain-specific intelligent analysis results. The visualization module processes and renders the results. The task and push module provides WebSocket push and snapshot subscription capabilities on the same port as HTTP.

[0098] It should be noted that the beneficial effects of the educational intelligent analysis system 200 provided in the above embodiments are the same as those of the educational intelligent analysis method described above, and will not be repeated here. Furthermore, the system provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to the actual situation to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiments, and will not be repeated here.

[0099] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned educational intelligent analysis methods.

[0100] An embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements any of the above-mentioned educational intelligent analysis methods. Through the combination of parsing, constraint, and execution, it improves the interpretability, stability, and engineering feasibility of complex educational intentions in cross-standard systems and multi-analysis task scenarios.

[0101] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0102] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An educational intelligence analysis method, characterized in that, include: Receive natural language queries from users; The natural language query is subjected to multi-stage semantic parsing, outputting an education-specific structured expression representing the query intent, and generating the confidence score of the education-specific structured expression, wherein the education-specific structured expression contains parsed semantic elements; Based on the semantic elements contained in the specialized structured expressions in the education field, the data sources, data aggregation methods, and result visualization types required for educational intelligent analysis are determined, and a data analysis execution plan is formed. According to the data analysis execution plan, the graph relationship is expanded in the education knowledge graph with standard nodes in the education standard system as the core, using one-hop or limited multi-hop graph relationships, to obtain a set of education data nodes associated with the query intent; Perform statistical aggregation operations on the set of educational data nodes to generate intelligent analysis results, which include at least one of charts, text summaries, and statistical data.

2. The educational intelligence analysis method according to claim 1, characterized in that, Also includes: During the execution of the data analysis execution plan, the task progress of each step in the data analysis execution plan is pushed to the user in real time through the WebSocket interface, and a task snapshot containing the current execution status is pushed to the user when the user initiates a subscription.

3. The educational intelligence analysis method according to claim 2, characterized in that, Also includes: After the data analysis execution plan is fully executed, the intelligent analysis results are sent to the business front end via a REST interface so that the business front end can render and display them.

4. The educational intelligence analysis method according to any one of claims 1 to 3, characterized in that, The multi-stage semantic parsing includes dictionary matching, rule matching, fuzzy matching, and collaboration with a large language model.

5. An educational intelligent analysis system, characterized in that, It includes a receiving module, a semantic parsing module, a plan generation module, a graph relationship expansion module, and an intelligent analysis result generation module; The receiving module is used to: receive natural language queries input by the user; The semantic parsing module is used to: perform multi-stage semantic parsing on the natural language query, output a structured expression specific to the education domain that represents the query intent, and generate the confidence score of the structured expression specific to the education domain, wherein the structured expression specific to the education domain contains parsed semantic elements; The plan generation module is used to: determine the data source, data aggregation method and result visualization type required for educational intelligent analysis based on the semantic elements contained in the specialized structured expression in the education field, and form a data analysis execution plan; The graph relationship expansion module is used to: expand the graph relationship in the educational knowledge graph with standard nodes in the educational standard system as the core, based on the data analysis execution plan, and obtain a set of educational data nodes associated with the query intent; The intelligent analysis result generation module is used to: perform statistical aggregation operations on the set of educational data nodes to generate intelligent analysis results, which include at least one of charts, text summaries, and statistical data.

6. The educational intelligent analysis system according to claim 5, characterized in that, It also includes a task progress push module, which is used to: push the task progress of each step in the data analysis execution plan to the user in real time through the WebSocket interface during the execution of the data analysis execution plan, and push a task snapshot containing the current execution status to the user when the user initiates a subscription.

7. The educational intelligent analysis system according to claim 6, characterized in that, It also includes a rendering and display module, which is used to send the intelligent analysis results to the business front end through a REST interface after all the data analysis execution plans have been executed, so that the business front end can render and display them.

8. An educational intelligent analysis system according to any one of claims 5 to 7, characterized in that, The multi-stage semantic parsing includes dictionary matching, rule matching, fuzzy matching, and collaboration with a large language model.

9. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement an educational intelligent analysis method according to any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements an educational intelligent analysis method according to any one of claims 1 to 4.

Images