Education system data cross-domain query method based on knowledge graph

Abstract

The invention discloses a knowledge graph-based education system data cross-domain query method. The method comprises the following steps: S1, constructing an education domain multi-level ontology; s2, recording association relationships among the partitions through a semantic connection strategy; s3, synchronously updating association information between the partitions by utilizing a semantic connection strategy; s4, the structured query statement comprises to-be-retrieved target partition information and query semantics; s5, generating a cross-partition query path; s6, executing parallel query operation on the positioned knowledge graph partitions based on a distributed computing framework, and integrating query results of the partitions according to a cross-partition query path to form a preliminary query result set; and S7, performing semantic optimization on the initial query result set by utilizing the semantic reasoning capability of the multi-level ontology, and integrating the optimized query results into a final output result. According to the method, the uniformity and query association efficiency of the education data are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of education systems, and in particular to a cross-domain query method for education system data based on a knowledge graph. Background Art

[0002] With the development of big data and artificial intelligence technology, the field of education has gradually adopted intelligent means to improve the efficiency and accuracy of teaching, evaluation and resource management. However, the data in the education system is usually distributed in different subsystems and regions, in various forms, including structured, semi-structured and unstructured data. The multi-source heterogeneous characteristics make cross-domain query and data integration an important bottleneck in the development of educational informatization, which seriously restricts the optimal allocation of educational resources and the promotion of intelligent applications.

[0003] At present, most education systems use independent databases or data warehouses to manage data. Existing systems usually implement basic query functions through manual configuration or rule definition, which has the following problems: First, due to the diverse sources of education data and the lack of unified semantic standards, existing technologies are unable to perform deep semantic modeling and standardized expression of data, resulting in difficulty in effectively utilizing the correlation and contextual semantics between data; Second, traditional query methods are mostly based on keyword matching or simple field retrieval. When faced with cross-disciplinary, cross-stage and cross-regional data query needs in the field of education, the query results are often inaccurate and incomplete, and cannot meet the requirements of complex education scenarios; Third, existing distributed query systems usually focus on data access performance rather than semantic optimization. For queries involving multi-dimensional semantic associations, their efficiency and accuracy are significantly insufficient.

[0004] To sum up, the existing technologies have significant deficiencies in the semantic unification, dynamic update, cross-domain query efficiency and result optimization of educational data, and are difficult to meet the needs of the development of modern educational informatization. The defects of the existing technologies directly affect the intelligence level and service capabilities of the education system. An innovative method is urgently needed to solve the above problems and provide more efficient and accurate technical support for data management and application of the education system. Summary of the invention

[0005] One purpose of the present invention is to propose a cross-domain query method for education system data based on knowledge graph, which significantly improves the uniformity of education data and the query association efficiency.

[0006] A cross-domain query method for education system data based on a knowledge graph according to an embodiment of the present invention includes the following steps:

[0007] S1. Based on the specific needs of the education field, define the core concepts, attributes and their hierarchical relationships, and construct a multi-level ontology in the education field;

[0008] S2. Based on the constructed multi-level ontology, the multi-source heterogeneous data in the field of education are partitioned and stored according to subject areas, educational stages and regions. An independent knowledge graph storage structure is established for each partition, and the association relationship between the partitions is recorded through the semantic connection strategy;

[0009] S3. Set up a dynamic update mechanism for each partition in the knowledge graph, update the graph nodes and relationships in the partition in real time according to the newly added or changed education data, and use the semantic connection strategy to synchronously update the association information between partitions;

[0010] S4. Receive a natural language query request from a user, perform semantic analysis on the query request, identify the query intent and generate a structured query statement, the structured query statement includes target partition information to be retrieved and query semantics;

[0011] S5. Based on the target partition information and query semantics in the query request, the query scheduling algorithm is used to locate the relevant partitions in the distributed storage structure of the knowledge graph, and the semantic association relationship between the target partitions is dynamically inferred through the semantic connection strategy to generate a cross-partition query path;

[0012] S6. Perform parallel query operations on the located knowledge graph partitions based on the distributed computing framework, integrate the query results of each partition according to the cross-partition query path, and form a preliminary query result set;

[0013] S7. Use the semantic reasoning capabilities of multi-level ontologies to semantically optimize the preliminary query result set, eliminate duplicate results, resolve semantic conflicts, and complete implicit related data, and integrate the optimized query results into the final output results.

[0014] Optionally, the S1 specifically includes the following steps:

[0015] S11. Determine the specific needs in the field of education and define the core concept set C in the education system. The core concepts include education-related subject themes, teaching objectives, learning objects, and resource types. For each core concept, define its attribute set A (c i ), each attribute a in the attribute set j Represents concept c i Characteristics information;

[0016] S12. Construct a hierarchical relationship set between concepts by analyzing the logical relationships in the field of education:

[0017] R = {r 1 ,r 2 ,…,r k};

[0018] Among them, each relationship r k Including the hierarchical relationship between concepts Dependencies and semantic association

[0019] S13. According to the core concept set C and attribute set A (c i ) and the relationship set R to establish the subject domain ontology Domain ,Discipline domain ontology includes the knowledge model of a specific discipline;

[0020] S14. Divide the education system into primary school, junior high school and high school according to the characteristics of the education stage, and construct the education stage ontology Stage , the education stage ontology defines the core concepts of different stages C Stage , attribute set A Stage and the relation set R Stage ;

[0021] S15. Define regional ontology based on the characteristics of educational resources and standardization requirements in different regions Region , combining educational resources with regional characteristics, the concept set C in the regional ontology Region including education policies and regional resource distribution;

[0022] S16. Associate the subject domain ontology, education stage ontology and region ontology to form a multi-level ontology set:

[0023] Ontology Multi = {Ontology Domain ,Ontology Stage ,Ontology Region};

[0024] S17. Based on multi-level ontology collection Ontology Multi Perform semantic structured modeling on multi-source heterogeneous data in the education system, map the fields in the data with the concepts, attributes and relationships in the multi-level ontology, and form a unified ontology basic framework.

[0025] Optionally, the S2 specifically includes the following steps:

[0026] S21. Classify the multi-source heterogeneous data set D in the field of education based on a multi-level ontology set, where the data classification is based on subject areas, educational stages, and regional characteristics;

[0027] S22. For the classified subject area set D Domain , Education stage set D Stage and the set of regional characteristics D Region , respectively construct knowledge graph partitions:

[0028] G i ={E i ,R i ,A i},i∈{Domain, Stage, Region};

[0029] Among them, E i Represents the entity set of partition i, including core concept entities and attribute entities, R i Represents the relationship set of partition i, which is consistent with the relationship set defined in the multi-level ontology. i Represents the attribute value set of partition i, which is used to describe the specific attribute data of the entity;

[0030] S23. In each knowledge graph partition G i In the above example, we use the semantic rule set Q of the multi-level ontology i Perform semantic reasoning and consistency checks on the data within the partition to eliminate redundant relationships and semantic conflicts within the partition;

[0031] S24. Based on the knowledge graph partition set G = {G Domain ,G Stage ,G Region Design a semantic connection strategy across partitions i Connect :

[0032]

[0033] S25. Store the constructed knowledge graph partition set G in a distributed graph database, and establish a global index and a local index mechanism for each partition. The global index is used to quickly locate the partition category, and the local index is used to query specific entities and relationships within the partition;

[0034] S26. Combined semantic connection strategy The cross-partition association relationships are dynamically and synchronously updated. When the knowledge graph data in a partition changes, the semantic connection rules of its related partitions are automatically adjusted.

[0035] Optionally, S3 specifically includes the following steps:

[0036] S31. Set a dynamic update trigger mechanism for the knowledge graph partition set G to monitor the newly added education data set D New and change the education data set D Update ,determine the update trigger conditions based on the data type, including adding new nodes, relationship updates, or attribute modifications;

[0037] S32. In each knowledge graph partition G iAccording to the semantic content of the newly added or changed educational data set, the knowledge graph structure in the partition is updated according to the following rules:

[0038] If the data involves a new entity, then in the entity set E i Add a new entity to n ;

[0039] If the data involves a relationship change, then in the relationship set R i Add or modify a relationship in u ;

[0040] If the data involves attribute updates, then in the attribute set A i Add or modify attribute value a u ;

[0041] S33. After updating the graph structure in the partition, use the semantic rule set Q in the partition i Perform consistency checks on new or modified content, check whether the new entities are semantically consistent with the existing entities, check whether the modified relationships comply with the constraints defined in the ontology, and check whether the attribute values ​​comply with the field specifications;

[0042] S34. Semantic connection strategy based on cross-partition Synchronize and adjust the updated content associated across partitions:

[0043] If the newly added entity involves a cross-partition hierarchical relationship, update The relevant connection rules;

[0044] If the relationship change involves a subordinate relationship, update Association rules of

[0045] If the attribute modification involves cross-domain semantic association, update Connection rules;

[0046] S35. After the dynamic synchronization update is completed, the index mechanism of the partitioned knowledge graph is updated, and the global index I Global Update the partition location of the newly added or changed entity or relationship in the partition, and update the local index I Local,i Update the specific node or attribute address of the newly added or changed content in the partition;

[0047] S36. Partition the updated knowledge graph into a set G new Stored back to the distributed graph database.

[0048] Optionally, the S4 specifically includes the following steps:

[0049] S41. Receive the user's natural language query request Q NL, the query request includes a query keyword or sentence input by a user;

[0050] S42. Using semantic parsing algorithm to query natural language Q NL Parse and segment the user's query request, extract the keyword set K in the query, and then use the multi-level ontology set Ontology Multi Perform semantic matching on the keyword set K, identify the concepts, attributes and relationships corresponding to the keywords, resolve the ambiguity of keywords based on the user's query context or historical behavior, and generate the query intent set I Q ;

[0051] S43. According to the query intention set I Q The concepts, attributes or relations in the knowledge graph are located in the partitions related to the query The semantic content corresponding to the matched query intent is represented as query semantics S Q , where S Q Including the query target entity set E Q , query target relation set R Q and query target attribute set A Q , combining the target partition information P and the query semantics S Q Generate structured query statement Q Structured =(P,S Q );

[0052] S44. In the generated structured query statement Q Structured In the above example, the logical consistency of the target partition information and the query semantics is verified so that the structured query statement satisfies the following conditions:

[0053] Query the target entity set E Q The entities in can be mapped to the entity set E in the partition i ;

[0054] Query target relation set R Q The relations in can match the relation set R in the partition i ;

[0055] Query target attribute set A Q The attributes in can match the attribute set A in the partition i .

[0056] Optionally, the S5 specifically includes the following steps:

[0057] S51. According to the structured query statement Q Structured Extract the knowledge graph partition set to be retrieved from the target partition set P;

[0058] S52. Based on the partition type, use the global index I Global Lock the target partition, for each target partition G i ∈P, through the local index I Local,i Determine the specific locations of the target entity set, target relationship set, and target attribute set within the partition;

[0059] S53. Based on the target partition set P and query semantics, dynamic semantic reasoning is performed in combination with the cross-partition semantic connection strategy to generate a cross-partition semantic association path. Dynamic semantic reasoning is performed in combination with the cross-partition semantic connection strategy to generate a cross-partition semantic association path Path Cross :

[0060] If the query semantics S Q The target entity set in partition G i and G j The relationship between the upper and lower levels is determined by the semantic connection rule. Reasoning about superior and subordinate paths across partitions;

[0061] If the query semantics S Q The target relationship set in involves the subordinate relationship between partitions, so the semantic connection rule Reasoning about dependency paths across partitions;

[0062] If the query semantics S Q The target attribute set in involves cross-domain semantic associations, so the semantic connection rules are used Manage the semantic association paths across partitions;

[0063] S54. Using semantic association path Path Cross Dynamically integrate the query results between target partitions to generate a cross-partition query path:

[0064] Path Query = {Path 1 ,Path 2 ,…,Path m}.

[0065] Optionally, the S6 specifically includes the following steps:

[0066] S61. Determine the target knowledge graph partition set and related query semantics corresponding to each path based on the cross-partition query path set;

[0067] S62. In the distributed computing framework, for each cross-partition query path Path k Assign computing tasks to each partition G in the path i Allocate independent computing nodes N i , for each computing node N iAssign target query semantics;

[0068] S63. At each computing node N i In the process, query operations are performed based on local knowledge graph partitions and query semantics to generate a local partition query result set R Local,i ;

[0069] S64. Query path based on cross-partition k , according to the semantic connection strategy in the path Query the local partition result set R Local,i Perform dynamic integration to generate path-level query result set R Path,k :

[0070] If the path Path k If it involves a hyponymous semantic relationship, then the Connect partition query results;

[0071] If the path Path k Involving subordinate semantic relations, Connect partition query results;

[0072] If the path Path k If it involves semantic association, then according to Connect partition query results;

[0073] S65. Summarize all path-level query result sets R Path = {R Path,1 ,R Path,2 ,…,R Path,m}, generate a preliminary query result set R Initial .

[0074] Optionally, the S7 specifically includes the following steps:

[0075] S71. Analyze the semantic information in the preliminary query result set using a multi-level ontology set;

[0076] S72. Preliminary query result set R Initial Perform deduplication processing to identify duplicate entities in the result entity set Unique identifier based on entity i ∈E Q Determine duplication with its corresponding multi-level ontology definition and identify duplicate relations in the result relation set According to the semantics of the relationship and the semantic connection rules associated in the target partition Determine duplication;

[0077] S73. Preliminary query result set R Initial, and if the same entity has ambiguity in different partitions, the query path Path Query and the context constraints defined by the multi-level ontology to resolve ambiguity. If the resulting relation set R Q If there is a conflict relationship in i Optimize conflicting relationships so that each relationship satisfies the constraints defined in the ontology;

[0078] S74. Use the semantic reasoning ability of multi-level ontology to complete the implicit related data in the initial query result set. If the query result entity set E Q Entity i If the associated attributes or relationships are not fully revealed, the semantic reasoning rule q k ∈Q i Complete the missing associated attributes and relationships. If the query result relationship set R Q If there is an indirect association relationship in i Connect Reasoning to complete implicit associations;

[0079] S75. The optimized query result set R Optimized Integrate into the final query result set:

[0080] R Final ={E Final ,R Final ,A Final}.

[0081] The beneficial effects of the present invention are:

[0082] (1) The present invention realizes semantic modeling and standardized expression of multi-source heterogeneous data in the field of education by constructing a multi-level ontology set, solves the problem of semantic inconsistency between multiple data sources in traditional methods, introduces a dynamic update mechanism, automatically adjusts the nodes, relationships and attributes in the knowledge graph according to the real-time changes of educational data, and synchronously updates the association information between partitions in combination with the cross-partition semantic connection strategy. The optimized data is checked for consistency and implicit associations are completed through a set of semantic reasoning rules, ensuring the real-time and semantic integrity of the knowledge graph, and significantly improving the uniformity of educational data and the efficiency of query associations.

[0083] (2) The present invention combines a distributed computing framework to quickly locate the target partition in cross-partition queries through global indexing and local indexing mechanisms, and uses semantic connection strategies to dynamically infer and generate cross-partition query paths. During the query process, parallel computing is used to execute local queries on each target partition, and query results are dynamically integrated according to predefined cross-partition semantic association rules to generate an efficient cross-domain query path.

[0084] (3) Based on the preliminary query results, the present invention uses the semantic reasoning capabilities of multi-level ontologies to deeply optimize the results, eliminate duplicate results, resolve semantic conflicts and complete implicit related data. The potential ambiguity and implicit association relationships in the query results are automatically parsed and completed through dynamic semantic reasoning rules, thereby ensuring the semantic consistency and integrity of the query results. The mechanism of completing implicitly associated data in multi-domain associated queries can effectively improve the comprehensiveness of the query results. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0086] Figure 1 This is a flow chart of a cross-domain query method for education system data based on knowledge graph proposed by the present invention. DETAILED DESCRIPTION

[0087] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0088] refer to Figure 1 , a cross-domain query method for education system data based on knowledge graph, comprising the following steps:

[0089] S1. Based on the specific needs of the education field, define the core concepts, attributes and their hierarchical relationships, and construct a multi-level ontology in the education field;

[0090] S2. Based on the constructed multi-level ontology, the multi-source heterogeneous data in the field of education are partitioned and stored according to subject areas, educational stages and regions. An independent knowledge graph storage structure is established for each partition, and the association relationship between the partitions is recorded through the semantic connection strategy;

[0091] S3. Set up a dynamic update mechanism for each partition in the knowledge graph, update the graph nodes and relationships in the partition in real time according to the newly added or changed education data, and use the semantic connection strategy to synchronously update the association information between partitions;

[0092] S4. Receive a natural language query request from a user, perform semantic analysis on the query request, identify the query intent and generate a structured query statement, the structured query statement includes target partition information to be retrieved and query semantics;

[0093] S5. Based on the target partition information and query semantics in the query request, the query scheduling algorithm is used to locate the relevant partitions in the distributed storage structure of the knowledge graph, and the semantic association relationship between the target partitions is dynamically inferred through the semantic connection strategy to generate a cross-partition query path;

[0094] S6. Perform parallel query operations on the located knowledge graph partitions based on the distributed computing framework, integrate the query results of each partition according to the cross-partition query path, and form a preliminary query result set;

[0095] S7. Use the semantic reasoning capabilities of multi-level ontologies to semantically optimize the preliminary query result set, eliminate duplicate results, resolve semantic conflicts, and complete implicit related data, and integrate the optimized query results into the final output results.

[0096] In this implementation, S1 specifically includes the following steps:

[0097] S11. Determine the specific needs in the field of education and define the core concept set C in the education system. The core concepts include education-related subject themes, teaching objectives, learning objects, and resource types. For each core concept, define its attribute set A (c i ), each attribute a in the attribute set j Represents concept c i Characteristics information;

[0098] S12. Construct a hierarchical relationship set between concepts by analyzing the logical relationships in the field of education:

[0099] R = {r 1 ,r 2 ,…,r k};

[0100] Among them, each relationship r k Including the hierarchical relationship between concepts Dependencies and semantic association relation r r k elated -to;

[0101] S13. According to the core concept set C and attribute set A (c i ) and the relationship set R to establish the subject domain ontology Domain ,Discipline domain ontology includes the knowledge model of a specific discipline;

[0102] S14. Divide the education system into primary school, junior high school and high school according to the characteristics of the education stage, and construct the education stage ontology Stage , the education stage ontology defines the core concepts of different stages C Stage, attribute set A Stage and the relation set R Stage ;

[0103] S15. Define regional ontology based on the characteristics of educational resources and standardization requirements in different regions Region , combining educational resources with regional characteristics, the concept set C in the regional ontology Region including education policies and regional resource distribution;

[0104] S16. Associate the subject domain ontology, education stage ontology and region ontology to form a multi-level ontology set:

[0105] Ontology Multi = {Ontology Domain ,Ontology Stage ,Ontology Region};

[0106] S17. Based on multi-level ontology collection Ontology Multi Perform semantic structured modeling on multi-source heterogeneous data in the education system, map the fields in the data with the concepts, attributes and relationships in the multi-level ontology, and form a unified ontology basic framework.

[0107] In this implementation, S2 specifically includes the following steps:

[0108] S21. Classify the multi-source heterogeneous data set D in the field of education based on a multi-level ontology set, where the data classification is based on subject areas, educational stages, and regional characteristics;

[0109] S22. For the classified subject area set D Domain , Education stage set D Stage and the set of regional characteristics D Region , respectively construct knowledge graph partitions:

[0110] G i ={E i ,R i ,A i},i∈{Domain, Stage, Region};

[0111] Among them, E i Represents the entity set of partition i, including core concept entities and attribute entities, R i Represents the relationship set of partition i, which is consistent with the relationship set defined in the multi-level ontology. i Represents the attribute value set of partition i, which is used to describe the specific attribute data of the entity;

[0112] S23. In each knowledge graph partition G i In the above example, we use the semantic rule set Q of the multi-level ontology i Perform semantic reasoning and consistency checks on the data within the partition to eliminate redundant relationships and semantic conflicts within the partition;

[0113] S24. Based on the knowledge graph partition set G = {G Domain ,G Stage ,G Region Designing semantic connection strategies across partitions

[0114]

[0115] S25. Store the constructed knowledge graph partition set G in a distributed graph database, and establish a global index and a local index mechanism for each partition. The global index is used to quickly locate the partition category, and the local index is used to query specific entities and relationships within the partition;

[0116] S26. Combined semantic connection strategy The cross-partition association relationships are dynamically and synchronously updated. When the knowledge graph data in a partition changes, the semantic connection rules of its related partitions are automatically adjusted.

[0117] In this implementation, S3 specifically includes the following steps:

[0118] S31. Set a dynamic update trigger mechanism for the knowledge graph partition set G to monitor the newly added education data set D New and change the education data set D Update ,determine the update trigger conditions based on the data type, including adding new nodes, relationship updates, or attribute modifications;

[0119] S32. In each knowledge graph partition G i According to the semantic content of the newly added or changed educational data set, the knowledge graph structure in the partition is updated according to the following rules:

[0120] If the data involves a new entity, then in the entity set E i Add a new entity to n ;

[0121] If the data involves a relationship change, then in the relationship set R i Add or modify a relationship in u ;

[0122] If the data involves attribute updates, then in the attribute set A i Add or modify attribute value a u ;

[0123] S33. After updating the graph structure in the partition, use the semantic rule set Q in the partition i Perform consistency checks on new or modified content, check whether the new entities are semantically consistent with the existing entities, check whether the modified relationships comply with the constraints defined in the ontology, and check whether the attribute values ​​comply with the field specifications;

[0124] S34. Semantic connection strategy based on cross-partition Synchronize and adjust the updated content associated across partitions:

[0125] If the newly added entity involves a cross-partition hierarchical relationship, update The relevant connection rules;

[0126] If the relationship change involves a subordinate relationship, update Association rules of

[0127] If the attribute modification involves cross-domain semantic association, update Connection rules;

[0128] S35. After the dynamic synchronization update is completed, the index mechanism of the partitioned knowledge graph is updated, and the global index I Global Update the partition location of the newly added or changed entity or relationship in the partition, and update the local index I Local,i Update the specific node or attribute address of the newly added or changed content in the partition;

[0129] S36. Partition the updated knowledge graph into a set G new Stored back to the distributed graph database.

[0130] In this implementation, S4 specifically includes the following steps:

[0131] S41. Receive the user's natural language query request Q NL ,The query request includes the query keyword or sentence input by the user;

[0132] S42. Using semantic parsing algorithm to query natural language Q NL Parse and segment the user's query request, extract the keyword set K in the query, and then use the multi-level ontology set Ontology Multi Perform semantic matching on the keyword set K, identify the concepts, attributes and relationships corresponding to the keywords, resolve the ambiguity of keywords based on the user's query context or historical behavior, and generate the query intent set I Q ;

[0133] S43. According to the query intention set I Q The concepts, attributes or relations in the knowledge graph are located in the partitions related to the query The semantic content corresponding to the matched query intent is represented as query semantics S Q , where S Q Including the query target entity set E Q , query target relation set R Q and query target attribute set A Q , combining the target partition information P and the query semantics S Q Generate structured query statement Q Structured =(P,S Q );

[0134] S44. In the generated structured query statement Q Structured In the query, verify the logical consistency of the target partition information and query semantics, so that the structured query statement meets the following conditions:

[0135] Query the target entity set E Q The entities in can be mapped to the entity set E in the partition i ;

[0136] Query target relation set R Q The relations in can match the relation set R in the partition i ;

[0137] Query target attribute set A Q The attributes in can match the attribute set A in the partition i .

[0138] In this implementation, S5 specifically includes the following steps:

[0139] S51. According to the structured query statement Q Structured Extract the knowledge graph partition set to be retrieved from the target partition set P;

[0140] S52. Based on the partition type, use the global index I Global Lock the target partition, for each target partition G i ∈P, through the local index I Local,i Determine the specific locations of the target entity set, target relationship set, and target attribute set within the partition;

[0141] S53. Based on the target partition set P and query semantics, dynamic semantic reasoning is performed in combination with the cross-partition semantic connection strategy to generate a cross-partition semantic association path. Dynamic semantic reasoning is performed in combination with the cross-partition semantic connection strategy to generate a cross-partition semantic association path Path Cross :

[0142] If the query semantics S Q The target entity set in partition G i and G jThe relationship between the upper and lower levels is determined by the semantic connection rule. Reasoning about superior and subordinate paths across partitions;

[0143] If the query semantics S Q The target relationship set in involves the subordinate relationship between partitions, so the semantic connection rule Reasoning about dependency paths across partitions;

[0144] If the query semantics S Q The target attribute set in involves cross-domain semantic associations, so the semantic connection rules are used Manage the semantic association paths across partitions;

[0145] S54. Using semantic association path Path Cross Dynamically integrate the query results between target partitions to generate a cross-partition query path:

[0146] Path Query = {Path 1 ,Path 2 ,…,Path m}.

[0147] In this implementation, S6 specifically includes the following steps:

[0148] S61. Determine the target knowledge graph partition set and related query semantics corresponding to each path based on the cross-partition query path set;

[0149] S62. In the distributed computing framework, for each cross-partition query path Path k Assign computing tasks to each partition G in the path i Allocate independent computing nodes N i , for each computing node N i Assign target query semantics;

[0150] S63. At each computing node N i In the process, query operations are performed based on local knowledge graph partitions and query semantics to generate a local partition query result set R Local,i ;

[0151] S64. Query path based on cross-partition k , according to the semantic connection strategy in the path Query the local partition result set R Local,i Perform dynamic integration to generate path-level query result set R Path,k :

[0152] If the path Path k If it involves a hyponymous semantic relationship, then the Connect partition query results;

[0153] If the path Path k Involving subordinate semantic relations, Connect partition query results;

[0154] If the path Path k If it involves semantic association, then according to Connect partition query results;

[0155] S65. Summarize all path-level query result sets R Path = {R Path,1 ,R Path,2 ,…,R Path,m}, generate a preliminary query result set R Initial .

[0156] In this implementation, S7 specifically includes the following steps:

[0157] S71. Analyze the semantic information in the preliminary query result set using a multi-level ontology set;

[0158] S72. Preliminary query result set R Initial Perform deduplication processing to identify duplicate entities in the result entity set Unique identifier based on entity i ∈E Q Determine duplication with its corresponding multi-level ontology definition and identify duplicate relations in the result relation set According to the semantics of the relationship and the semantic connection rules associated in the target partition Determine duplication;

[0159] S73. Preliminary query result set R Initial , and if the same entity has ambiguity in different partitions, the query path Path Query and the context constraints defined by the multi-level ontology to resolve ambiguity. If the resulting relation set R Q If there is a conflict relationship in i Optimize conflicting relationships so that each relationship satisfies the constraints defined in the ontology;

[0160] S74. Use the semantic reasoning ability of multi-level ontology to complete the implicit related data in the initial query result set. If the query result entity set E Q Entity i If the associated attributes or relationships are not fully revealed, the semantic reasoning rule q k ∈Q iComplete the missing associated attributes and relationships. If the query result relationship set R Q If there is an indirect association relationship, the cross-partition semantic connection strategy is used Reasoning to complete implicit associations;

[0161] S75. The optimized query result set R Optimized Integrate into the final query result set:

[0162] R Final ={E Final ,R Final ,A Final}.

[0163] Embodiment 1:

[0164] Embodiment In a basic education information platform construction project of a city education bureau, the city hopes to realize cross-regional, cross-disciplinary and cross-education stage data sharing and intelligent query functions to support comprehensive evaluation of students' academic performance and accurate resource recommendation. However, the existing education data management system is decentralized and heterogeneous, and cannot efficiently integrate data from different schools, grades and subjects. For this reason, the city education bureau decided to adopt the method of the present invention to build a cross-domain query platform for education system data based on knowledge graph to solve the problems of data dispersion, semantic inconsistency and low query efficiency.

[0165] The city's data sources include student performance data, teacher evaluation data, and educational resource data from 60 primary and secondary schools in the region. The data are distributed in different subsystems, including school management systems, online course platforms, and teaching resource libraries. The data are in the form of structured performance records, semi-structured teaching plan documents, and unstructured classroom videos and homework texts. Previous data queries only supported simple retrieval in a single dimension and were unable to associate data between subjects or educational stages. For example, a school wanted to query "a list of students with excellent math grades in elementary school and rapidly improved English grades in junior high school." The traditional method requires searching in the elementary and junior high school systems separately, and then manually comparing and analyzing them, which is time-consuming, labor-intensive, and prone to errors.

[0166] The whole process of applying the method of the present invention in this project is as follows:

[0167] First, in view of the dispersion and heterogeneity of the city's education data, a multi-level ontology modeling method was used to divide the data into three levels: subject area, education stage, and region. The ontology model defines core concepts (such as "students", "grades", and "resources"), attributes (such as "subject name", "time node"), and relationships (such as "grades belong to students"). By modeling historical data, a multi-level ontology set is formed, and the data of each school is standardized. For example, elementary school mathematics score records were originally stored in Excel files, while junior high school English score data were stored in database tables. After modeling, these data were uniformly mapped to the knowledge graph and represented as "student-subject-grade" triples.

[0168] Then, using the knowledge graph partition storage technology, the data is partitioned and stored according to subject areas, education stages, and school regions. For example, primary school mathematics data is stored in Junior high school English data is stored in At the same time, cross-partition semantic connection rules are established. For example, for the cross-stage data association of "students", the upper and lower relationships are defined. Represents the data association between a student in primary school and junior high school.

[0169] In one actual application scenario, a key middle school in a certain city wanted to query "a list of students who ranked in the top 10% in math scores in primary school and whose English scores in junior high school improved significantly in a short period of time (the last two years)". Using traditional methods, this query needs to be searched in the primary school and junior high school systems respectively, exported in table form, and then manually compared and calculated based on student IDs. This process usually takes an information management staff member about 7 hours to complete, and due to data ambiguity and inconsistent formats, the accuracy rate is less than 80%.

[0170] After adopting the method of the present invention, the user inputs the query request "Find students who rank in the top 10% in elementary school math scores and whose junior high school English scores have improved significantly" through natural language. The system first uses semantic parsing technology to segment and parse the query request, identifies the key semantics of "math scores", "English scores", "top 10%" and "improvement", and maps them to the corresponding concepts and attributes in the knowledge graph.

[0171] The system uses a distributed index mechanism to quickly lock the target partition and During the parallel query execution, the primary school mathematics data partition returns the entity set E of the top 10% of students in terms of scores. Top10 , the junior high school English data partition returns the student entity set E with significant improvement according to the change in grades within two years Improved , then the system follows the cross-partition semantic connection rules Integrate the two entity sets and generate a list of students that meet the conditions as the preliminary query result.

[0172] In the result optimization stage, the system completes and deduplicates the query results through the semantic reasoning rules of multi-level ontologies. For example, a student's name in elementary school and junior high school was mistakenly identified as two entities due to slight differences caused by the pinyin input method. After semantic reasoning, the system automatically merged them into the same entity and completed its missing attribute information. Finally, the system output a list of 25 students, while providing each student's math ranking, English score change curve and related learning resource recommendations.

[0173] In order to verify the performance of the method of the present invention, data of 15,000 students from three schools in the city were selected for comparative testing. The results showed that the traditional method took 420 minutes and the accuracy rate was 78%; the method of the present invention took only 15 minutes and the accuracy rate was increased to 96%. The specific data are shown in Table 1 below:

[0174] Table 1 Comparative data of the present invention and the traditional method

[0175] method Query time (minutes) Query accuracy Implicit data completion rate Traditional methods 420 78％ none Method of the present invention 15 96％ 92％

[0176] In addition, in the test of implicitly associated data completion capability, the method of the present invention successfully completed 92% of the missing attribute information based on semantic reasoning rules, while the traditional method completely relied on manual analysis and was unable to complete the completion. This result fully demonstrates the superiority and practical value of the method of the present invention in cross-domain query of education system data.

[0177] The present invention realizes semantic modeling and standardized expression of multi-source heterogeneous data in the field of education by constructing a multi-level ontology set, solves the problem of semantic inconsistency between multiple data sources in traditional methods, introduces a dynamic update mechanism, automatically adjusts the nodes, relationships and attributes in the knowledge graph according to the real-time changes of education data, and synchronously updates the association information between partitions in combination with the cross-partition semantic connection strategy. The optimized data is checked for consistency and implicit associations are completed through a set of semantic reasoning rules, ensuring the real-time and semantic integrity of the knowledge graph, and significantly improving the uniformity of education data and the efficiency of query association.

[0178] The present invention combines a distributed computing framework to quickly locate the target partition through global indexing and local indexing mechanisms in cross-partition queries, and uses semantic connection strategies to dynamically infer and generate cross-partition query paths. During the query process, parallel computing is used to execute local queries on each target partition, and query results are dynamically integrated according to predefined cross-partition semantic association rules to generate an efficient cross-domain query path.

[0179] Based on the preliminary query results, the present invention deeply optimizes the results through the semantic reasoning ability of multi-level ontologies, eliminates duplicate results, resolves semantic conflicts and completes implicit related data. For the potential ambiguity and implicit association relationships in the query results, dynamic semantic reasoning rules are used to automatically parse and complete them, thereby ensuring the semantic consistency and integrity of the query results. The mechanism of completing implicit associated data in multi-domain associated queries can effectively improve the comprehensiveness of the query results.

[0180] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A cross-domain query method for education system data based on knowledge graph, characterized in that: The steps include: S1. Based on the specific needs of the education field, define the core concepts, attributes and their hierarchical relationships, and construct a multi-level ontology in the education field; S2. Based on the constructed multi-level ontology, the multi-source heterogeneous data in the field of education are partitioned and stored according to subject areas, educational stages and regions. An independent knowledge graph storage structure is established for each partition, and the association relationship between the partitions is recorded through the semantic connection strategy; S3. Set up a dynamic update mechanism for each partition in the knowledge graph, update the graph nodes and relationships in the partition in real time according to the newly added or changed education data, and use the semantic connection strategy to synchronously update the association information between partitions; S4. Receive a natural language query request from a user, perform semantic analysis on the query request, identify the query intent and generate a structured query statement, the structured query statement includes target partition information to be retrieved and query semantics; S5. Based on the target partition information and query semantics in the query request, the query scheduling algorithm is used to locate the relevant partitions in the distributed storage structure of the knowledge graph, and the semantic association relationship between the target partitions is dynamically inferred through the semantic connection strategy to generate a cross-partition query path; S6. Perform parallel query operations on the located knowledge graph partitions based on the distributed computing framework, integrate the query results of each partition according to the cross-partition query path, and form a preliminary query result set; S7. Use the semantic reasoning capabilities of multi-level ontologies to semantically optimize the preliminary query result set, eliminate duplicate results, resolve semantic conflicts, and complete implicit related data, and integrate the optimized query results into the final output results.

2. According to claim 1, a cross-domain query method for education system data based on knowledge graph is characterized in that: The S1 specifically includes the following steps: S11. Determine the specific needs in the field of education and define the core concept set C in the education system. The core concepts include education-related subject themes, teaching objectives, learning objects, and resource types. For each core concept, define its attribute set A (c i ), each attribute a in the attribute set j Represents concept c i Characteristics information; S12. Construct a hierarchical relationship set between concepts by analyzing the logical relationships in the field of education: R＝{r1,r2,…,r k}； Among them, each relationship r k Including the hierarchical relationship between concepts Dependencies and semantic association S13. According to the core concept set C and attribute set A (c i ) and the relationship set R to establish the subject domain ontology Domain ,Discipline domain ontology includes the knowledge model of a specific discipline; S14. Divide the education system into primary school, junior high school and high school according to the characteristics of the education stage, and construct the education stage ontology Stage , the education stage ontology defines the core concepts of different stages C Stage , attribute set A Stage and the relation set R Stage ; S15. Define regional ontology based on the characteristics of educational resources and standardization requirements in different regions Region , combining educational resources with regional characteristics, the concept set C in the regional ontology Region including education policies and regional resource distribution; S16. Associate the subject domain ontology, education stage ontology and region ontology to form a multi-level ontology set: Ontology Multi ={Ontology Domain ,Ontology Stage ,Ontology Region }; S17. Based on multi-level ontology collection Ontology Multi Perform semantic structured modeling on multi-source heterogeneous data in the education system, map the fields in the data with the concepts, attributes and relationships in the multi-level ontology, and form a unified ontology basic framework.

3. According to claim 1, a cross-domain query method for education system data based on knowledge graph is characterized in that: The S2 specifically includes the following steps: S21. Classify the multi-source heterogeneous data set D in the field of education based on a multi-level ontology set, where the data classification is based on subject areas, educational stages, and regional characteristics; S22. For the classified subject area set D Domain , Education stage set D Stage and the set of regional characteristics D Region , respectively construct knowledge graph partitions: G i ={E i ,R i ,A i },i∈{Domain,Stage,Region}; Among them, E i Represents the entity set of partition i, including core concept entities and attribute entities, R i Represents the relationship set of partition i, which is consistent with the relationship set defined in the multi-level ontology. i Represents the attribute value set of partition i, which is used to describe the specific attribute data of the entity; S23. In each knowledge graph partition G i In the above example, we use the semantic rule set Q of the multi-level ontology i Perform semantic reasoning and consistency checks on the data within the partition to eliminate redundant relationships and semantic conflicts within the partition; S24. Based on the knowledge graph partition set G = {G Domain ,G Stage ,G Region Designing semantic connection strategies across partitions S25. Store the constructed knowledge graph partition set G in a distributed graph database, and establish a global index and a local index mechanism for each partition. The global index is used to quickly locate the partition category, and the local index is used to query specific entities and relationships within the partition; S26. Combined semantic connection strategy The cross-partition association relationships are dynamically and synchronously updated. When the knowledge graph data in a partition changes, the semantic connection rules of its related partitions are automatically adjusted.

4. According to claim 1, a cross-domain query method for education system data based on knowledge graph is characterized in that: The S3 specifically includes the following steps: S31. Set a dynamic update trigger mechanism for the knowledge graph partition set G to monitor the newly added education data set D New and change the education data set D Update ,determine the update trigger conditions based on the data type, including adding new nodes, relationship updates, or attribute modifications; S32. In each knowledge graph partition G i According to the semantic content of the newly added or changed educational data set, the knowledge graph structure in the partition is updated according to the following rules: If the data involves a new entity, then in the entity set E i Add a new entity to n ; If the data involves a relationship change, then in the relationship set R i Add or modify a relationship in u ; If the data involves attribute updates, then in the attribute set A i Add or modify attribute value a u ; S33. After updating the graph structure in the partition, use the semantic rule set Q in the partition i Perform consistency checks on new or modified content, check whether the new entities are semantically consistent with the existing entities, check whether the modified relationships comply with the constraints defined in the ontology, and check whether the attribute values ​​comply with the field specifications; S34. Semantic connection strategy based on cross-partition Synchronize and adjust the updated content associated across partitions: If the newly added entity involves a cross-partition hierarchical relationship, update The relevant connection rules; If the relationship change involves a subordinate relationship, update Association rules of If the attribute modification involves cross-domain semantic association, update Connection rules; S35. After the dynamic synchronization update is completed, the index mechanism of the partitioned knowledge graph is updated, and the global index I Global Update the partition location of the newly added or changed entity or relationship in the partition, and update the local index I Local,i Update the specific node or attribute address of the newly added or changed content in the partition; S36. Partition the updated knowledge graph into a set G new Stored back to the distributed graph database.

5. According to claim 1, a cross-domain query method for education system data based on knowledge graph is characterized in that: The S4 specifically comprises the following steps: S41. Receive the user's natural language query request Q NL , the query request includes a query keyword or sentence input by a user; S42. Using semantic parsing algorithm to query natural language Q NL Parse and segment the user's query request, extract the keyword set K in the query, and then use the multi-level ontology set Ontology Multi Perform semantic matching on the keyword set K, identify the concepts, attributes and relationships corresponding to the keywords, resolve the ambiguity of keywords based on the user's query context or historical behavior, and generate the query intent set I Q ; S43. According to the query intention set I Q The concepts, attributes or relations in the knowledge graph are located in the partitions related to the query The semantic content corresponding to the matched query intent is represented as query semantics S Q , where S Q Including the query target entity set E Q , query target relation set R Q and query target attribute set A Q , combining the target partition information P and the query semantics S Q Generate structured query statement Q Structured =(P,S Q ); S44. In the generated structured query statement Q Structured In the above example, the logical consistency of the target partition information and the query semantics is verified so that the structured query statement satisfies the following conditions: Query the target entity set E Q The entities in can be mapped to the entity set E in the partition i ; Query target relation set R Q The relations in can match the relation set R in the partition i ; Query target attribute set A Q The attributes in can match the attribute set A in the partition i .

6. According to claim 1, a cross-domain query method for education system data based on knowledge graph is characterized in that: The S5 specifically includes the following steps: S51. According to the structured query statement Q Structured Extract the knowledge graph partition set to be retrieved from the target partition set P; S52. Based on the partition type, use the global index I Global Lock the target partition, for each target partition G i ∈P, through the local index I Local,i Determine the specific locations of the target entity set, target relationship set, and target attribute set within the partition; S53. Based on the target partition set P and query semantics, dynamic semantic reasoning is performed in combination with the cross-partition semantic connection strategy to generate a cross-partition semantic association path. Dynamic semantic reasoning is performed in combination with the cross-partition semantic connection strategy to generate a cross-partition semantic association path Path Cross : If the query semantics S Q The target entity set in partition G i and G j The relationship between the upper and lower levels is determined by the semantic connection rule. Reasoning about superior and subordinate paths across partitions; If the query semantics S Q The target relationship set in involves the subordinate relationship between partitions, so the semantic connection rule Reasoning about dependency paths across partitions; If the query semantics S Q The target attribute set in involves cross-domain semantic associations, so the semantic connection rules are used Manage semantic association paths across partitions; S54. Using semantic association path Path Cross Dynamically integrate the query results between target partitions to generate a cross-partition query path: Path Query ={Path1,Path2,…,Path m }。 7. According to claim 1, a cross-domain query method for education system data based on knowledge graph is characterized in that: The S6 specifically comprises the following steps: S61. Determine the target knowledge graph partition set and related query semantics corresponding to each path based on the cross-partition query path set; S62. In the distributed computing framework, for each cross-partition query path Path k Assign computing tasks to each partition G in the path i Allocate independent computing nodes N i , for each computing node N i Assign target query semantics; S63. At each computing node N i In the process, query operations are performed based on local knowledge graph partitions and query semantics to generate a local partition query result set R Local,i ; S64. Query path based on cross-partition k , according to the semantic connection strategy in the path Query the local partition result set R Local,i Perform dynamic integration to generate path-level query result set R Path,k : If the path Path k If it involves a hyponymous semantic relationship, then the Connect partition query results; If the path Path k Involving subordinate semantic relations, Connect partition query results; If the path Path k If it involves semantic association, then according to Connect partition query results; S65. Summarize all path-level query result sets R Path = {R Path,1 ,R Path,2 ,…,R Path,m }, generate a preliminary query result set R Initial .

8. According to claim 1, a cross-domain query method for education system data based on knowledge graph is characterized in that: The S7 specifically comprises the following steps: S71. Analyze the semantic information in the preliminary query result set using a multi-level ontology set; S72. Preliminary query result set R Initial Perform deduplication processing to identify duplicate entities in the result entity set Unique identifier based on entity i ∈E Q Determine duplication with its corresponding multi-level ontology definition and identify duplicate relations in the result relation set According to the semantics of the relationship and the semantic connection rules associated in the target partition Determine duplication; S73. Preliminary query result set R Initial , and if the same entity has ambiguity in different partitions, the query path Path Query and the context constraints defined by the multi-level ontology to resolve ambiguity. If the resulting relation set R Q If there is a conflict relationship in i Optimize conflicting relationships so that each relationship satisfies the constraints defined in the ontology; S74. Use the semantic reasoning ability of multi-level ontology to complete the implicit related data in the initial query result set. If the query result entity set E Q Entity i If the associated attributes or relationships are not fully revealed, the semantic reasoning rule q k ∈Q i Complete the missing associated attributes and relationships. If the query result relationship set R Q If there is an indirect association relationship, the cross-partition semantic connection strategy is used Reasoning to complete implicit associations; S75. The optimized query result set R Optimized Integrate into the final query result set: R Final ={E Final ,R Final ,A Final }。

Images