Calculating two key indexes of the support degree suppp and the confidence degree conf, defining two key parameters minsupp and minconf, and substituting the two key parameters minsupp and minconf into the formula to obtain:

in the same way, the method can obtain,

the method for constructing the knowledge graph provided by the embodiment comprises the following steps: acquiring initial learning data, storing the initial learning data into a learning resource database, and extracting and labeling ontology terms of the initial learning data according to a subject teaching outline and teaching materials; inputting the relation and sequence between each element knowledge point, constructing a directed disciplinary knowledge map through a body editor, generating an OWL file and storing the OWL file into a learning system in a triple form; updating the learning process state data of the input user in real time, and performing multi-knowledge point modeling on the learning process state data of the user through a DKT algorithm to obtain the learning mastery state of the user; and mining the relation among the knowledge points according to the learning and mastering state of the user, and dynamically updating the subject knowledge map. According to the knowledge graph modeling method based on deep learning, DKT is tracked based on deep learning, the problem of multi-knowledge-point modeling scenes is effectively solved, the latest knowledge state of a learner is accurately mastered, the construction of the knowledge graph is completed, the internal relation of knowledge points is obtained, knowledge reasoning is carried out on the knowledge graph, and the knowledge graph is dynamically updated.

Second embodiment of the invention

Please refer to fig. 2-5.

As shown in fig. 5, an embodiment of the present invention further provides a system for constructing a knowledge graph, including:

thedata module 100 is used for collecting initial learning data, storing the initial learning data into a learning resource database, extracting body terms of the initial learning data according to a subject teaching outline and teaching materials, and labeling the extracted body terms;

specifically, for thedata module 100, initial learning data is collected, including course education targets, learning resources, course structures, teaching strategies, exercise test question libraries and the like in each field, subject experts extract subject terms from subject teaching outlines and teaching materials in advance, and the process of labeling the subject terms is divided into two modes of automatic labeling and manual labeling of the system, so that a more efficient labeling mode can be switched according to different scenes, and flexibility and work efficiency are improved.

Themap building module 200 is used for inputting the relation and sequence between each element knowledge point, building a directed discipline knowledge map through the ontology editor, generating an OWL file and storing the OWL file into the learning system in a triple form;

specifically, for theatlas construction module 200, after the association and sequence between each meta-knowledge point are entered according to the initial learning data acquired by thedata module 100, a directed disciplinary knowledge atlas is constructed by using the ontology editor Prot g e according to the extracted and labeled ontology terms and the association and sequence between the meta-knowledge points, and after the construction is completed, an OWL file is generated and stored in the learning system in the form of a triplet.

TheDKT module 300 is configured to update the learning process state data entered into the user in real time, and perform multi-knowledge point modeling on the learning process state data of the user through a DKT algorithm to obtain a learning mastery state of the user;

specifically, for theDKT module 300, a learner logs in the learning system to learn and test, and during the learning process, according to the learning and testing data of the learner, the model of the learner with multiple knowledge points is realized through the DKT algorithm, so as to obtain the current learning and mastering state of the learner, such as the mastering state, answering time, answering accuracy and other data of each knowledge point.

A knowledgepoint mining module 400, configured to mine the relation between the knowledge points according to the learning and mastering state of the user, and dynamically update the discipline knowledge graph;

specifically, for the knowledgepoint mining module 400, as shown in fig. 4, the answer correct probability of each knowledge point of all learners is obtained through the DKT model; adopting a probability association rule mining technology to identify implicit relations between relation acquisition knowledge points; and judging the mastering state of the knowledge points by the user according to the implicit relationship among the knowledge points, and dynamically updating the disciplinary knowledge map in real time.

The excavating method comprises the following specific steps: acquiring all knowledge point answer pair probabilities of all learners at present through DKT; and identifying the implicit relation between the relation acquisition knowledge points by a probability association rule mining technology. From the perspective of the prerequisite relationship, if the concept Si is a prerequisite of the concept Sj, then a learner who does not know Si is likely not to know Sj, and a learner who knows Sj is likely to know Si, resulting in

in the same way, the method can obtain,

and theknowledge inference module 500 is used for performing knowledge inference on the subject knowledge graph by adopting a TransE algorithm, adding new element knowledge points and hierarchical relations thereof, and further improving the subject knowledge graph.

Specifically, for theknowledge inference module 500, the OWL files in the form of triples stored in the learning system are translated, including subjects, relationships and objects, and the subjects and the objects are collectively referred to as entities. When a directed knowledge graph is constructed, the three are linked to form a graph, wherein each node is an entity, each edge is a relation, a subject points to an object, the visual meaning of TransE is distributed vector representation based on the entity and the relation, the relation in each triple is regarded as translation from the subject to the object, and the vector value of the subject plus the vector value of the relation is enabled to be as far as possible with the vector value of the object by continuously adjusting the subject, the relation and the object, so that the relation deduction between the subject and the object is completed, new element knowledge points and the relation thereof are increased and decreased, the subject knowledge graph is further perfected, and the prediction accuracy and the effectiveness of the knowledge graph are improved.

The training process of the TransR algorithm comprises the following steps: firstly, determining a training set, a hyper-parameter gamma and a learning rate lambda; initializing a relation vector and an entity vector, randomly taking a value in a preset value range for each dimension of each vector, and normalizing all vectors after initializing; entering a loop, adopting minipatch, accelerating the training speed of a batch of training, carrying out negative sampling on each batch of data (randomly replacing a certain entity of a triplet in a training set), and initially setting T _ batch as an empty list, and then adding a list consisting of tuple pairs (original triplets, broken triplets) to the empty list: t _ batch [ ([ h, r, T ], [ h ', r, T' ]), ([ ], [ ]), ·. ·; and taking the T _ batch, training, and adjusting parameters by adopting gradient descent.

In a preferred embodiment, as shown in fig. 3, theDKT module 300 is specifically constructed by the following steps:

implementing an LSTM training by adopting a tensorflow method;

Specifically, the DKT model structure shown in fig. 2 is developed in time sequence, and the sequence x1, x2, x3. corresponds to t1, t2, t3..₁,x′₂,x^′₃… corresponds to the external characteristics (answer times and answer time) of the input sequences x1, x2 and x3, hidden layer states h0, h1 and h2 … correspond to knowledge point mastering conditions of students at all times, the output sequences y1, y2 and y3 … of the model correspond to the probability that all exercises in an answer question library of the students at all times answer correct answers, when a learner answers the questions, the system carries out real-time online calling of the model through tensoflow serving, the knowledge point mastering state of the students is returned from the hidden layer, the output layer returns the correct probability of the next answer of the knowledge point, the learning and mastering state of the user is accurately obtained, and the subject knowledge map is dynamically updated.

The system for constructing a knowledge graph provided by the embodiment comprises: the data module is used for collecting initial learning data, storing the initial learning data into a learning resource database, extracting body terms of the initial learning data according to a subject teaching outline and teaching materials, and labeling the extracted body terms; the map building module is used for inputting the relation and the sequence between each element knowledge point, building a directed discipline knowledge map through the body editor, generating an OWL file and storing the OWL file into the learning system in a triple form; the DKT module is used for updating and inputting the learning process state data of the user in real time, and performing multi-knowledge point modeling on the learning process state data of the user through a DKT algorithm to obtain the learning mastery state of the user; the knowledge point mining module is used for mining the relation among the knowledge points according to the learning and mastering state of the user and dynamically updating the disciplinary knowledge map; and the knowledge inference module is used for carrying out knowledge inference on the subject knowledge graph by adopting a TransE algorithm, adding new element knowledge points and hierarchical relations thereof and further perfecting the subject knowledge graph. According to the knowledge graph modeling method based on deep learning, DKT is tracked based on deep learning, the problem of multi-knowledge-point modeling scenes is effectively solved, the latest knowledge state of a learner is accurately mastered, the construction of the knowledge graph is completed, the internal relation of knowledge points is obtained, knowledge reasoning is carried out on the knowledge graph, and the knowledge graph is dynamically updated.

In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the modules may be a logical division, and in actual implementation, there may be another division, for example, multiple modules or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.

The modules described as separate parts may or may not be physically separate, and parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.

In addition, functional modules in the embodiments of the present invention may be integrated into one processing module, or each of the modules may exist alone physically, or two or more modules are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode.

The foregoing is directed to the preferred embodiment of the present invention, and it is understood that various changes and modifications may be made by one skilled in the art without departing from the spirit of the invention, and it is intended that such changes and modifications be considered as within the scope of the invention.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.

Claims

1. A construction method of a knowledge graph is characterized by at least comprising the following steps:

2. The method of constructing a knowledge-graph of claim 1, further comprising:

3. The method of constructing a knowledge-graph of claim 1 wherein said initial learning data includes learner personal information, subject learning content information, subject exercise test data and learning resource data.

4. The method of constructing a knowledge-graph of claim 1 wherein said labeling comprises automatic labeling and manual labeling, wherein,

5. The method of constructing a knowledge graph according to claim 1, wherein the learning process state data comprises a knowledge state record and a knowledge level test record, wherein,

6. The method for constructing the knowledge graph according to claim 1, wherein the learning process state data of the user is updated and recorded in real time, and the learning process state data of the user is subjected to multi-knowledge-point modeling through a DKT algorithm to obtain the learning mastery state of the user, and specifically comprises the following steps:

implementing an LSTM training by adopting a tensorflow method;

7. The method for constructing a knowledge graph according to claim 1, wherein the relation between knowledge points is mined according to the learning and mastering states of the user, and the subject knowledge graph is dynamically updated, specifically:

8. A system for constructing a knowledge graph, comprising:

9. The system for constructing a knowledge graph according to claim 8, wherein the DKT module is specifically:

implementing an LSTM training by adopting a tensorflow method;

10. A computer-readable storage medium, comprising a stored computer program, wherein the computer program, when executed, controls an apparatus in which the computer-readable storage medium is located to perform the method of constructing a knowledge-graph according to any one of claims 1 to 7.