Micro-learning path generation method based on cross-media search

Abstract

The invention discloses a micro-learning path generation method based on cross-media search. The method comprises the following steps: acquiring a learning target of a learner; performing feature extraction according to the learning target to obtain a target feature vector; performing iterative solution on the similarity matrix based on the cross relation matrix according to the target feature vector to obtain a marked path; wherein the cross relation matrix and the similarity matrix are constructed by utilizing a trained cross-media search model; according to the marked path, obtaining a predecessor learning path and a postdecessor learning path so as to realize the generation of a micro-learning path; according to the method, the learning target is converted into the target feature vector, the similarity matrix and the cross relation matrix of the micro-learning resources are generated based on the cross-media search model, and the learning path is generated based on the similarity and the mutual relation of the micro-learning resource contents, so that the quality of the generated learning path is improved.

Description

Technical Field

[0001] The present invention relates to a method for generating a micro - learning learning path based on cross - media search, and belongs to the technical field of learning path generation. Background Art

[0002] In related technologies, existing cross - media search methods mostly adopt the RNN algorithm based on two media types of images and texts, and cannot meet the needs of micro - learning mainly based on videos and audios; while most existing learning path generation methods mainly adopt recommendation technologies, and give recommendations on the learning content of the next stage based on the learning achievements of learners at each stage, so as to generate a learning path. This method faces the problems of learner enthusiasm and the inherent cold - start problem of recommendation technologies. Summary of the Invention

[0003] The purpose of the present invention is to overcome the deficiencies in the prior art, and provide a method for generating a micro - learning learning path based on cross - media search. By converting the learning objective into a target feature vector, and generating a similarity matrix and a cross - relationship matrix of micro - learning resources based on a cross - media search model, a learning path is generated based on the similarity and mutual relationship of the content of micro - learning resources, thereby improving the quality of the generated learning path.

[0004] To achieve the above object, the present invention is implemented by the following technical solutions:

[0005] The present invention discloses a method for generating a micro - learning learning path based on cross - media search, including:

[0006] Obtain the learning objective of the learner; according to the learning objective, perform feature extraction to obtain a target feature vector;

[0007] According to the target feature vector, iteratively solve the similarity matrix based on the cross - relationship matrix to obtain a marked path; wherein, the cross - relationship matrix and the similarity matrix are constructed by using a trained cross - media search model;

[0008] According to the marked path, obtain a precursor learning path and a successor learning path to realize the generation of the micro - learning learning path;

[0009] Wherein, the training method of the cross - media search model includes:

[0010] Obtain micro - learning resources, and the micro - learning resources include media data sets of multiple media types;

[0011] For the media data in the media data set of any media type, perform feature extraction to obtain a media feature vector, and then construct a training set;

[0012] Based on the training set, the pre-constructed cross-media search model is trained according to the total loss function to obtain a trained cross-media search model; wherein, the total loss function includes a classification constraint loss, a center constraint loss, and a ranking constraint loss.

[0013] Further, the media types include image type, video type, audio type, and text type;

[0014] The steps of feature extraction are as follows:

[0015] For the media data in the media dataset of image type, feature extraction is performed based on a convolutional neural network to obtain an image media feature vector;

[0016] For the media data in the media dataset of video type, feature extraction is performed by extracting a preset number of evenly distributed video frames and converting them into images to obtain a video media feature vector;

[0017] For the media data in the media dataset of audio type, feature extraction is performed by obtaining a frequency spectrum through Fourier transform to obtain an audio media feature vector;

[0018] For the media data in the media dataset of text type, feature extraction is performed by generating a word vector model to obtain a text media feature vector.

[0019] Further, the expression of the total loss function is as follows:

[0020]

[0021] In the formula, represents the total loss function; represents the classification constraint loss; represents the center constraint loss; represents the ranking constraint loss;

[0022] The classification constraint loss has the following expression:

[0023]

[0024] In the formula, m represents the index of the media type, represents that the media type is image type, represents that the media type is video type, represents that the media type is audio type, represents that the media type is text type;

[0025] represents the number of media data in the media dataset of media type m; represents the nth media feature vector of media type m; The label of the nth piece of media data with media type m; Indicates the cross-entropy loss function with respect to the media feature vector and the label;

[0026] The central constraint loss has the following expression:

[0027]

[0028] In the formula, represents the number of media data in the media dataset of all media types; represents the nth media feature vector; represents the label of the nth piece of media data corresponding subcategory clustering center; represents the square of the Euclidean norm;

[0029] The expression of the ranking constraint loss is as follows:

[0030]

[0031] In the formula, represents the media feature vectors of any four different media types in the media dataset, and among the inputs, two of the media feature vectors belong to the same subcategory, while the other two media feature vectors are different from each other; represents the first balance parameter; represents the second balance parameter;

[0032] represents the media feature vector and the media feature vector the cosine distance between them;

[0033] represents the media feature vector and the media feature vector the cosine distance between them;

[0034] represents the media feature vector and the media feature vector the cosine distance between them.

[0035] Furthermore, the expression of the similarity matrix is as follows:

[0036]

[0037] In the formula, represents the similarity matrix; Indicates the first media feature vector; Indicates the nth media feature vector; Indicates the media feature vector And the media feature vector Of the cosine similarity; Indicates the media feature vector And the media feature vector Of the cosine similarity.

[0038] Furthermore, the expression of the cross-relationship matrix is as follows:

[0039]

[0040] In the formula, Indicates the cross-relationship matrix; Indicates the media feature vector And the media feature vector Of the cosine similarity; Indicates the media feature vector And the media feature vector Of the cosine similarity.

[0041] Furthermore, the target feature vector Belongs to the elements in the media feature vector set , , ;

[0042] The steps of iteratively solving the similarity matrix based on the cross-relationship matrix according to the target feature vector to obtain the marked path include:

[0043] Obtain the elements of the target row i corresponding to the target feature vector In the similarity matrix , sort them from largest to smallest to obtain the similarity queue , , where Indicates the maximum threshold of the similarity between the front and back contents; Indicates the minimum threshold of the similarity between the front and back contents; Indicates the element in the th row and th column of the similarity matrix; Indicates the element in the th row and th column of the similarity matrix; Indicates the element in the th row and th column of the similarity matrix;

[0044] According to the cross-relationship matrix , based on a preset cross-relationship judgment function Solve the similarity matrix to obtain the adjacent rows of target row i ;

[0045] Delete target row i and target column i of the said similarity matrix and replace the said target row i with the adjacent rows , repeat and go back to the step of obtaining the similarity queue for iterative loop, and finally obtain the marked path.

[0046] Furthermore, the expression of the preset cross-relationship judgment function is:

[0047]

[0048] In the formula, represents the cross-relationship judgment function; represents the element in the th row and th column of the cross-relationship matrix P; represents the element in the th row and th column of the cross-relationship matrix P;

[0049] In response to , it is marked as ;

[0050] In response to , it is marked as .

[0051] Furthermore, the expression of the said marked path is:

[0052] ;

[0053] In the formula, represents the serial number of the second learning unit in the predecessor learning path with as the target; represents the serial number of the first learning unit in the predecessor learning path with as the target; represents the serial number of the learning unit with the learning target as the starting point of path generation; represents the serial number of the first learning unit in the successor learning path with as the starting point; represents the serial number of the second learning unit in the successor learning path with as the starting point;

[0054] The expression of the said predecessor learning path is:

[0055]

[0056] In the formula, represents the predecessor learning path; represents the second learning unit in the predecessor learning path with as the target; represents the first learning unit in the predecessor learning path with

[0057] as the target;

[0058]

[0059] In the formula, represents the successor learning path; represents the learning unit as the starting point for generating the learning path; represents the first learning unit in the successor learning path with as the starting point; represents the second learning unit in the successor learning path with

[0060] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0061] For the method for generating a microlearning learning path based on cross-media search of the present invention, firstly, based on the cross-media search model, media data of different media types can be mapped to the same representation space and mutual search can be realized. Secondly, by introducing the center constraint loss and the ranking constraint loss into the cross-media search model, the fine-grained feature learning ability of the model is improved. In addition, based on the similarity matrix and the cross-relationship matrix, the sequence relationship of media data for learning can be discriminated. Finally, by converting the learning target into a target feature vector and generating the similarity matrix and the cross-relationship matrix of the microlearning resources based on the cross-media search model, a learning path is generated based on the similarity and mutual relationship of the content of the microlearning resources, improving the quality of the generated learning path. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a flowchart of the method for generating a microlearning learning path based on cross-media search provided by an embodiment of the present invention.

[0063] Figure 2 is a schematic diagram of the cross-media search model provided by an embodiment of the present invention;

[0064] Figure 3 is a schematic diagram of the principle of generating an undirected graph using the similarity matrix provided by an embodiment of the present invention;

[0065] Figure 4 It is a schematic diagram for converting an undirected graph into a directed graph and removing redundant edges by using a cross-relationship matrix provided by an embodiment of the present invention. Specific embodiments

[0066] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and cannot be used to limit the protection scope of the present invention.

[0067] This embodiment provides a method for generating a microlearning learning path based on cross-media search, as Figure 1 shown, including:

[0068] Obtain the learning objective of the learner; according to the learning objective, perform feature extraction to obtain a target feature vector;

[0069] Based on the target feature vector, iteratively solve the similarity matrix based on the cross-relationship matrix to obtain a marked path; wherein, the cross-relationship matrix and the similarity matrix are constructed by using a trained cross-media search model;

[0070] According to the marked path, obtain a precursor learning path and a successor learning path to realize the generation of the microlearning learning path;

[0071] Among them, the training method of the cross-media search model includes:

[0072] Obtain microlearning resources, and the microlearning resources include media data sets of multiple media types;

[0073] For the media data in the media data set of any media type, perform feature extraction to obtain a media feature vector, and then construct a training set;

[0074] According to the training set, train a pre-constructed cross-media search model based on the total loss function to obtain a trained cross-media search model; wherein, the total loss function includes a classification constraint loss, a center constraint loss, and a ranking constraint loss.

[0075] The technical concept of the present invention is as follows: First, based on the cross-media search model, media data of different media types can be mapped to the same representation space and mutual search can be realized. Second, the center constraint loss and the ranking constraint loss are introduced into the cross-media search model, which improves the fine-grained feature learning ability of the model. In addition, based on the similarity matrix and the cross-relationship matrix, the sequential relationship of learning media data can be discriminated. Finally, by converting the learning objective into a target feature vector, and based on the cross-media search model to generate the similarity matrix and the cross-relationship matrix of the microlearning resources, a learning path is generated based on the similarity and mutual relationship of the microlearning resource content, which improves the quality of the generated learning path.

[0076] Before generating the target learning path, it is necessary to first construct and train a cross-media search model, as well as construct a cross-relationship matrix and a similarity matrix. The specific steps are as follows:

[0077] (1) Obtain micro learning resources, which include media data sets of various media types.

[0078] For the media data in the media data set of any media type, feature extraction is performed to obtain media feature vectors, and then a training set is constructed.

[0079] The media types include image type, video type, audio type, and text type. The steps of feature extraction are as follows:

[0080] For the media data in the media data set of the image type, feature extraction is performed based on a convolutional neural network to obtain image media feature vectors;

[0081] For the media data in the media data set of the video type, feature extraction is performed by extracting a preset number of evenly distributed video frames and converting them into images to obtain video media feature vectors;

[0082] For the media data in the media data set of the audio type, feature extraction is performed by obtaining the frequency spectrum through Fourier transform to obtain audio media feature vectors;

[0083] For the media data in the media data set of the text type, feature extraction is performed by generating a word vector model to obtain text media feature vectors.

[0084] (2) According to the training set, the pre-constructed cross-media search model is trained based on the total loss function to obtain a trained cross-media search model; among them, the total loss function includes classification constraint loss, center constraint loss, and ranking constraint loss.

[0085] Specifically, the expression of the total loss function is as follows:

[0086]

[0087] In the formula, represents the total loss function; represents the classification constraint loss; represents the center constraint loss; represents the ranking constraint loss;

[0088] The classification constraint loss The expression of

[0089]

[0090] In the formula, m represents the index of the media type, Indicates that the media type is an image type, Indicates that the media type is a video type, Indicates that the media type is an audio type, Indicates that the media type is a text type;

[0091] Indicates the number of media data in the media dataset with the media type of m; Indicates the nth media feature vector with the media type of m; Indicates the label of the nth media data with the media type of m; Indicates regarding the media feature vector and the label The cross-entropy loss function is calculated as follows:

[0092]

[0093] Center constraint loss The expression is as follows:

[0094]

[0095] In the formula, Indicates the number of media data in the media dataset of all media types; Indicates the nth media feature vector; Indicates the label of the nth media data The corresponding sub-category clustering center; Indicates the square of the Euclidean norm;

[0096] The calculation expression of the Euclidean norm is as follows:

[0097]

[0098] The expression of the ranking constraint loss is as follows:

[0099]

[0100] In the formula, Indicates the media feature vectors of any four different media types in the media dataset, and two of the media feature vectors belong to the same sub-category when input, and the other two media feature vectors are different from each other; Indicates the first balance parameter; Indicates the second balance parameter;

[0101] Indicates the media feature vector and the media feature vector The cosine distance between;

[0102] Represents the media feature vector With the media feature vector The cosine distance between;

[0103] Represents the media feature vector With the media feature vector The cosine distance between.

[0104] The calculation expression of the cosine distance is as follows:

[0105]

[0106] In the formula, Represents the modulus of the vector.

[0107] The first balance parameter in this embodiment The value is 1, and the second balance parameter The value is 0.5.

[0108] During the training process, the media data of four media types are used as input, and the trained cross-media search model is obtained by optimizing through minimizing the loss function. The structure of the cross-media search model is as Figure 2 Shown, the learning rate is initialized to 0.001 and increases by 0.5 every three rounds.

[0109] After the training is completed, the output of the fully connected layer is extracted as the common representation of the four media types, and the cosine distance is used to compare the similarity of each output value with others to complete the cross-media search.

[0110] By adopting the above technical solutions, the micro learning course data of four media types, namely video, audio, picture, and text, can be mapped to the same representation space, and mutual search can be realized. The fine-grained features of the media data are extracted through three loss functions: classification constraint, center constraint, and ranking constraint.

[0111] (3)According to the trained cross-media search model, construct a similarity matrix and a cross-relationship matrix.

[0112] Specifically, the expression of the similarity matrix is as follows:

[0113]

[0114] In the formula, Represents the similarity matrix; Represents the first media feature vector; Represents the nth media feature vector; Represents the media feature vector With the media feature vector Cosine similarity; Represents the media feature vector And the media feature vector Cosine similarity of.

[0115] The calculation formula of cosine similarity is: , Represents the norm of the vector.

[0116] The expression of the cross-relationship matrix is as follows:

[0117]

[0118] In the formula, Represents the cross-relationship matrix; Represents the media feature vector And the media feature vector Cosine similarity of; Represents the media feature vector And the media feature vector Cosine similarity of.

[0119] When constructing the learning path, the matrix Can be regarded as an undirected graph, as Figure 3 Shown, by calculating the matrix The direction of each edge in the matrix Can be obtained, that is, when , regard b as the pre-node of a, and thus the directed graph Can be obtained, where the weight of each edge is the cosine similarity between two nodes, as Figure 4 Shown. Given the target feature vector As the learning goal, learning path generation is to generate a minimum spanning tree in the directed graph With As the starting point and end point respectively.

[0120] By adopting the above technical solution, the cross-media search model can be used to establish a similarity and cross-relationship matrix for media data mapped to the same representation space, so as to judge the sequence relationship of learning for each media data.

[0121] The specific steps of the micro-learning learning path generation method based on cross-media search are as follows:

[0122] Step 1: Obtain the learning goal of the learner; according to the learning goal, perform feature extraction to obtain the target feature vector;

[0123] Target feature vector Belongs to the set of media feature vectors Elements in, , 。

[0124] Step 2: Based on the target feature vector, iteratively solve the similarity matrix based on the cross-relationship matrix to obtain the marked path.

[0125] 2.1. Obtain the similarity matrix The elements of the target row i corresponding to the target feature vector in are sorted from largest to smallest to obtain the similarity queue , , where represents the maximum threshold of the similarity between the front and back contents; represents the minimum threshold of the similarity between the front and back contents; represents the element in the th row and th column of the similarity matrix; represents the element in the th row and th column of the similarity matrix; represents the element in the th row and th column of the similarity matrix;

[0126] By setting the minimum threshold and the maximum threshold of the similarity between two adjacent micro-learning resources, the calculation accuracy can be improved.

[0127] 2.2. Based on the cross-relationship matrix , solve the similarity matrix using a preset cross-relationship judgment function . The expression of the preset cross-relationship judgment function is:

[0128]

[0129] In the formula, represents the cross-relationship judgment function; represents the element in the th row and th column of the cross-relationship matrix P; represents the element in the th row and th column of the cross-relationship matrix P;

[0130] In response to , it is marked as ;

[0131] In response to , it is marked as .

[0132] Based on Solve the matrix to obtain the adjacent rows of the target row i ;

[0133] 2.3. Delete the target row i and target column i of the similarity matrix and replace the target row i with the adjacent row , and repeat to return to the step of obtaining the similarity queue in 2.1 for iterative loop until the marked path is obtained, ;

[0134] In the formula, represents the serial number of the second learning unit in the predecessor learning path with as the target; represents the serial number of the first learning unit in the predecessor learning path with as the target; represents the serial number of the learning unit with the learning target as the starting point of path generation; represents the serial number of the first learning unit in the successor learning path with as the starting point; represents the serial number of the second learning unit in the successor learning path with as the starting point.

[0135] Step 3. Obtain the predecessor learning path and the successor learning path according to the marked path to generate the micro - learning learning path;

[0136] Specifically, the expression of the predecessor learning path is:

[0137]

[0138] In the formula, represents the predecessor learning path; represents the second learning unit in the predecessor learning path with as the target; represents the first learning unit in the predecessor learning path with as the target; represents the learning unit as the starting point of learning path generation.

[0139] The expression of the successor learning path is:

[0140]

[0141] In the formula, represents the successor learning path; represents the learning unit as the starting point of learning path generation; represents with The 1st learning unit in the successor learning path starting from Indicates The 2nd learning unit in the successor learning path starting from

[0142] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0143] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 One process or multiple processes and / or blocks Figure 1 One block or multiple blocks.

[0144] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implement the functions specified in Figure 1 One process or multiple processes and / or blocks Figure 1 One block or multiple blocks.

[0145] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 One process or multiple processes and / or blocks Figure 1 One block or multiple blocks.

[0146] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A micro-learning learning path generation method based on cross-media search, characterized by: include: Obtain learners’ learning goals; According to the learning objective, feature extraction is performed to obtain a target feature vector; According to the target feature vector, the similarity matrix is ​​iteratively solved based on the cross-relationship matrix to obtain a labeling path; wherein the cross-relationship matrix and the similarity matrix are constructed using a trained cross-media search model; According to the marked path, a predecessor learning path and a successor learning path are obtained to realize the generation of a micro-learning learning path; The training method of the cross-media search model includes: Acquire micro-learning resources, wherein the micro-learning resources include media data sets of multiple media types; For the media data in the media data set of any media type, feature extraction is performed to obtain the media feature vector, and then a training set is constructed; According to the training set, the pre-constructed cross-media search model is trained based on the total loss function to obtain a trained cross-media search model; wherein the total loss function includes classification constraint loss, center constraint loss and ranking constraint loss.

2. The method for generating a micro-learning learning path based on cross-media search according to claim 1 is characterized in that: The media types include image type, video type, audio type and text type; The steps of feature extraction are as follows: For the media data of the image type media dataset, feature extraction is performed based on the convolutional neural network to obtain the image media feature vector; For the media data of the video type media data set, a uniform preset number of video frames are extracted and converted into images for feature extraction to obtain a video media feature vector; For the media data of the audio type media data set, the frequency spectrum is obtained by Fourier transform to perform feature extraction and obtain the audio media feature vector; For the media data of the text type media data set, feature extraction is performed by generating a word vector model to obtain a text media feature vector.

3. The method for generating a micro-learning learning path based on cross-media search according to claim 1 is characterized in that: The expression of the total loss function is as follows: ； In the formula, represents the total loss function; represents the classification constraint loss; represents the center constraint loss; represents the ranking constraint loss; The classification constraint loss The expression is as follows: ； Where m is the index of the media type. Indicates that the media type is image type. Indicates that the media type is video type. Indicates that the media type is audio type. Indicates that the media type is text type; Indicates the number of media data in the media data set with media type m; The nth media feature vector representing the media type m; The label indicating the nth media data with media type m; Represents the media feature vector and tags The cross entropy loss function of The center constraint loss The expression is as follows: ； In the formula, The number of media data in the media data set representing all media types; represents the nth media feature vector; The tag indicating the nth piece of media data The corresponding subcategory cluster center; represents the square of the Euclidean norm; The expression of the ranking constraint loss is as follows: ； In the formula, A media feature vector in a media dataset representing any four different media types, where two of the media feature vectors belong to the same subcategory and the other two are different; represents the first equilibrium parameter; represents the second equilibrium parameter; Represents a media feature vector With media feature vector The cosine distance between Represents a media feature vector With media feature vector The cosine distance between Represents a media feature vector With media feature vector The cosine distance between .

4. The method for generating a micro-learning learning path based on cross-media search according to claim 1 is characterized in that: The expression of the similarity matrix is ​​as follows: ； In the formula, represents the similarity matrix; Represents the first media feature vector; represents the nth media feature vector; Represents a media feature vector With media feature vector The cosine similarity of Represents a media feature vector With media feature vector The cosine similarity of .

5. The method for generating a micro-learning learning path based on cross-media search according to claim 4 is characterized in that: The expression of the cross-relation matrix is ​​as follows: ； In the formula, represents the cross-relation matrix; Represents a media feature vector With media feature vector The cosine similarity of Represents a media feature vector With media feature vector The cosine similarity of .

6. The method for generating a micro-learning learning path based on cross-media search according to claim 5 is characterized in that: The target feature vector Belongs to the media feature vector set The elements in , ; According to the target feature vector, the similarity matrix is ​​iteratively solved based on the cross-relation matrix to obtain the step of marking the path, which includes: Get the similarity matrix The target feature vector The elements of the corresponding target row i are sorted from large to small to obtain the similarity queue , ,in, Indicates the maximum threshold of the similarity between the previous and next contents; Indicates the minimum threshold of similarity between previous and next content; Represents the similarity matrix Line Elements of a column; Represents the similarity matrix Line Elements of a column; Represents the similarity matrix Line Elements of a column; According to the cross-relationship matrix , based on the preset cross-relationship judgment function Similarity Matrix Solve and get the adjacent rows of the target row i ; Delete the similarity matrix The target row i and target column i are replaced by the adjacent row , repeat the steps of getting the similarity queue and iterate, and finally get the marked path.

7. The method for generating a micro-learning learning path based on cross-media search according to claim 6 is characterized in that: The expression of the preset cross-relationship judgment function is: ； In the formula, represents the cross-relationship judgment function; Indicates the cross-relation matrix P Line Elements of a column; Indicates the cross-relation matrix P Line Elements of a column; In response to , then marked as ; In response to , then marked as .

8. The method for generating a micro-learning learning path based on cross-media search according to claim 6 is characterized in that: The marked path The expression is: ； In the formula, Indicates The number of the second learning unit in the predecessor learning path of the target; Indicates The number of the first learning unit in the predecessor learning path of the target; Indicates the sequence number of the learning unit that uses the learning goal as the starting point for path generation; Indicates The number of the first learning unit in the subsequent learning path as the starting point; Indicates The number of the second learning unit in the subsequent learning path as the starting point; The expression of the predecessor learning path is: ； In the formula, represents the predecessor learning path; Indicates The second learning unit in the predecessor learning path of the target; Indicates The first learning unit in the predecessor learning path of the target; Represents the learning unit that is the starting point for generating the learning path; The expression of the subsequent learning path is: ； In the formula, represents the subsequent learning path; Represents the learning unit that is the starting point for generating the learning path; Indicates The first learning unit in the subsequent learning path as the starting point; Indicates The second learning unit in the subsequent learning path as the starting point.

Images