A teaching resource recommendation method and system based on natural language processing

Abstract

The application provides a teaching resource recommendation method and system based on natural language processing, relates to the technical field of natural language processing, and performs structural segmentation on text features of teaching text data to obtain a plurality of teaching segments; determines a semantic dependency tree of each teaching segment, determines an explicit dependency degree of teaching logic between the teaching segments according to all semantic dependency trees and semantic correlations between different teaching segments; performs implicit correlation analysis on teaching semantics between the teaching segments to obtain an implicit correlation degree, and determines implicit constraint conditions of the teaching logic between the teaching segments from all implicit correlation degrees; constructs a collaborative relationship graph between the teaching segments in a teaching resource allocation process based on the explicit dependency degree and the implicit constraint conditions of the teaching logic between the teaching segments, and adaptively allocates teaching resources according to the collaborative relationship graph. The above scheme can realize adaptive allocation of teaching resources based on the explicit dependency and the implicit constraint of the teaching logic between the teaching segments.

Description

Technical Field

[0001] This application relates to the field of natural language processing technology, and more specifically, to a teaching resource recommendation method and system based on natural language processing. Background Technology

[0002] With the development of online education and intelligent teaching, the recommendation of teaching resources has become an important part of improving learning efficiency and personalized teaching experience. Natural language processing technology can automatically understand the semantics and extract features from teaching texts, thereby helping the system to identify the logical relationships and knowledge structures between teaching content, providing more intelligent support for resource allocation and recommendation. The introduction of this technology not only improves the precision of resource scheduling, but also provides a new way to realize the efficient use of teaching resources and individualized instruction.

[0003] However, most existing teaching resource recommendation technologies rely on keyword-based or surface-level semantic matching methods, neglecting the deep logical dependencies between different teaching segments. In this case, the recommendation results often only reflect the superficial relevance of the content, failing to effectively reflect the progression between knowledge points and the logical chain of teaching. For example, when the system needs to allocate resources to a segment with higher difficulty, if the logical dependencies and implicit knowledge transfer relationships with preceding segments are not considered, it may lead to a mismatch between resource allocation and the learning path, thereby reducing the continuity of teaching and learning effectiveness. Therefore, how to achieve adaptive allocation of teaching resources based on the explicit dependencies and implicit constraints of the teaching logic between teaching segments has become a challenge for the industry. Summary of the Invention

[0004] This application provides a teaching resource recommendation method and system based on natural language processing, which can achieve adaptive allocation of teaching resources based on explicit dependencies and implicit constraints of teaching logic between teaching segments.

[0005] Firstly, this application provides a teaching resource recommendation method based on natural language processing, including:

[0006] Obtain teaching text data from teaching equipment for the allocation of teaching resources;

[0007] The teaching text data is structured and segmented based on text features using a pre-trained natural language processing model to obtain multiple teaching segments.

[0008] Determine the semantic dependency tree of each teaching segment, and then determine the explicit dependency degree of the teaching logic between each teaching segment based on the semantic dependency trees of all teaching segments and the semantic relevance between different teaching segments.

[0009] Based on the teaching intent of the teaching text data, implicit association analysis is performed on the teaching semantics between each teaching segment to obtain the implicit association degree of the teaching semantics between each teaching segment. Then, the implicit constraints of the teaching logic between each teaching segment are determined through all the implicit association degrees.

[0010] Based on the explicit dependency and implicit constraints of the teaching logic between each teaching segment, a collaborative relationship graph is constructed between each teaching segment during the allocation of teaching resources. Then, the teaching resources in the teaching equipment are adaptively allocated according to the collaborative relationship graph.

[0011] In some embodiments, the teaching text data refers to electronic textbook data involved in the teaching process.

[0012] In some embodiments, the teaching text data is structurally segmented based on text features using a pre-trained natural language processing model to obtain multiple teaching segments, specifically including:

[0013] Based on a pre-trained natural language processing model, semantic features are extracted from the sentences in the teaching text data to obtain the semantic features of each sentence.

[0014] Based on the semantic features of each statement and the text structure of the teaching text data, the teaching text data is semantically segmented to obtain multiple teaching segments.

[0015] In some embodiments, determining the semantic dependency tree for each teaching segment specifically includes;

[0016] For each teaching segment, determine the part of speech of each word in the segment;

[0017] Analyze the dependency relationships between words based on their parts of speech and the contextual information of the teaching segment;

[0018] Based on the dependency relationships between words, a tree structure is constructed to connect the words, resulting in a semantic dependency tree for each teaching segment, and thus a semantic dependency tree for each teaching segment.

[0019] In some embodiments, determining the explicit dependency degree of instructional logic between instructional segments based on the semantic dependency tree of all instructional segments and the semantic relevance between different instructional segments specifically includes:

[0020] For every two teaching segments, extract the semantic feature vectors of the two teaching segments;

[0021] The semantic relevance between two teaching segments is determined by semantic feature vectors;

[0022] The syntactic and semantic relationships of the teaching logic between two teaching segments are determined based on the matching degree of the semantic dependency tree between the two teaching segments;

[0023] The explicit dependency of the teaching logic between two teaching segments is determined by the semantic relevance and the syntactic semantic relationship, thereby obtaining the explicit dependency of the teaching logic between each pair of teaching segments.

[0024] In some embodiments, implicit association analysis of the teaching semantics between teaching segments is performed based on the teaching intent of the teaching text data to obtain the implicit association degree of the teaching semantics between teaching segments, specifically including:

[0025] The teaching semantics of each teaching segment are annotated to obtain the teaching semantics of each segment.

[0026] Based on the teaching intent of the teaching text data, the teaching semantics of each teaching segment are matched and scored to obtain the matching score of the teaching intent of each teaching segment.

[0027] By using all the matching scores, the implicit association of the teaching semantics between each teaching segment is obtained, thus obtaining the implicit association degree of the teaching semantics between each teaching segment.

[0028] In some embodiments, determining the implicit constraints of the teaching logic between teaching segments through all implicit correlations specifically includes:

[0029] The teaching difficulty of each teaching segment is assessed, and a difficulty coefficient matrix of the teaching segment is generated.

[0030] For any two teaching segments, the association attention of the teaching logic between the two teaching segments is determined based on the implicit correlation of their teaching semantics and the difficulty coefficient matrix.

[0031] Construct and initialize a machine learning model for modeling implicit logical relationships between teaching segments;

[0032] Set the associated attention as the connection weight in the forward propagation phase of the machine learning model;

[0033] The implicit logical dependencies between any two teaching segments are extracted using the machine learning model, and implicit constraints between the two teaching segments are generated, thereby obtaining a set of implicit constraints between all pairs of teaching segments.

[0034] Secondly, this application provides a teaching resource recommendation system based on natural language processing, including:

[0035] The acquisition module is used to acquire teaching text data for the allocation of teaching resources from teaching equipment;

[0036] The feature processing module is used to perform structured segmentation of the teaching text data based on pre-trained natural language processing models to obtain multiple teaching segments.

[0037] The feature processing module is also used to determine the semantic dependency tree of each teaching segment, and then determine the explicit dependency degree of the teaching logic between each teaching segment based on the semantic dependency trees of all teaching segments and the semantic relevance between different teaching segments.

[0038] The feature processing module is also used to perform implicit association analysis on the teaching semantics between each teaching segment according to the teaching intention of the teaching text data, to obtain the implicit association degree of the teaching semantics between each teaching segment, and then determine the implicit constraints of the teaching logic between each teaching segment through all the implicit association degrees.

[0039] The allocation module is used to construct a collaborative relationship graph between teaching segments during the allocation process based on the explicit dependency and implicit constraints of the teaching logic between each teaching segment, and then adaptively allocate teaching resources in the teaching equipment according to the collaborative relationship graph.

[0040] Thirdly, this application provides a computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described teaching resource recommendation method based on natural language processing.

[0041] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described teaching resource recommendation method based on natural language processing.

[0042] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0043] In this embodiment of the application, teaching text data to be allocated teaching resources is obtained from the teaching equipment;

[0044] The teaching text data is structured and segmented based on pre-trained natural language processing models to obtain multiple teaching segments. A semantic dependency tree for each teaching segment is determined, and then the explicit dependency degree of the teaching logic between each teaching segment is determined based on the semantic dependency trees of all teaching segments and the semantic relevance between different teaching segments. Implicit association analysis is performed on the teaching semantics between each teaching segment based on the teaching intent of the teaching text data to obtain the implicit association degree of the teaching semantics between each teaching segment. Then, implicit constraints of the teaching logic between each teaching segment are determined through all implicit association degrees. Based on the explicit dependency degree and implicit constraints of the teaching logic between each teaching segment, a collaborative relationship graph of each teaching segment is constructed during the allocation of teaching resources. Then, teaching resources in the teaching equipment are adaptively allocated based on the collaborative relationship graph.

[0045] Therefore, this application constructs a collaborative relationship graph between teaching segments during the allocation of teaching resources based on the explicit dependency and implicit constraints of the teaching logic between each teaching segment. Then, it adaptively allocates teaching resources in teaching equipment according to the collaborative relationship graph. First, by structurally segmenting the teaching text and extracting semantic dependency trees, it achieves refined semantic modeling of teaching segments, thus laying a data foundation for the subsequent construction of logical relationships and avoiding the shortcomings of fragmented teaching content and lack of logical continuity in existing technologies. Second, in the explicit dependency modeling process, semantic dependency tree matching and semantic relevance calculation clarify the direct sequential relationship and pre-dependency links between teaching segments, ensuring that resource allocation conforms to the natural progression logic of knowledge points and helps improve the rationality of the teaching path. Third, in the implicit constraint modeling, by capturing the potential semantic connections and knowledge transfer relationships between segments, it can compensate for the potential logical links not covered by explicit dependencies, making resource allocation more efficient. The allocation not only stays at the level of explicit logic but also reflects the implicit connections and transitions in the teaching process, thereby enhancing the system's ability to perceive complex teaching relationships. Then, by merging explicit dependency path graphs and implicit constraint path graphs to form collaborative constraint relationships, a comprehensive characterization of the dependency strength, order, and resource sharing conditions between teaching segments is achieved, providing more comprehensive and dynamic logical constraints for resource allocation. Finally, adaptive allocation is implemented based on the collaborative relationship graph, dynamically adjusting resource allocation strategies based on explicit and implicit logical relationships, ensuring a high degree of match between resource delivery and the teaching path. This guarantees both gradual learning and promotes the transfer and absorption of implicit knowledge. In summary, this application's solution, through the fusion modeling of explicit dependencies and implicit constraints, not only solves the technical problems of fragmented recommendation results and lack of logic in existing technologies but also achieves dual optimization of teaching resources in terms of logical progression and latent semantic continuity, significantly improving teaching continuity, personalized adaptability, and learning effectiveness. Attached Figure Description

[0046] Figure 1 This is an exemplary flowchart of a teaching resource recommendation method based on natural language processing, according to some embodiments of this application;

[0047] Figure 2 This is a flowchart illustrating the process of determining a semantic dependency tree according to some embodiments of this application;

[0048] Figure 3 This is a flowchart illustrating the process of determining implicit correlation degree according to some embodiments of this application;

[0049] Figure 4 This is a schematic diagram of the structure of a teaching resource recommendation system based on natural language processing, according to some embodiments of this application;

[0050] Figure 5 This is a schematic diagram of the structure of a computer device that implements a teaching resource recommendation method based on natural language processing, according to some embodiments of this application. Detailed Implementation

[0051] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] refer to Figure 1 The figure is an exemplary flowchart of a teaching resource recommendation method based on natural language processing, according to some embodiments of this application. This teaching resource recommendation method based on natural language processing mainly includes the following steps:

[0053] In step 101, the teaching text data to be allocated for teaching resources is obtained from the teaching equipment.

[0054] It should be noted that the teaching text data in this application refers to the electronic textbook data involved in the teaching process. In specific implementation, the teaching system or terminal can be connected through device interface calls to access the teaching content resource library stored locally or in the cloud, and data requests and transmissions can be made using communication protocols. During the transmission process, the teaching text data is parsed in accordance with data format specifications to extract valid text fields. Then, preliminary cleaning operations (such as removing format symbols and correcting encoding) are performed to ensure the integrity of the teaching text data, and the cleaned text data is used as the teaching text data for resource allocation.

[0055] In step 102, the teaching text data is segmented into structured segments based on text features using a pre-trained natural language processing model to obtain multiple teaching segments.

[0056] In some embodiments, the teaching text data is structurally segmented based on text features using a pre-trained natural language processing model to obtain multiple teaching segments, specifically including:

[0057] Based on a pre-trained natural language processing model, semantic features are extracted from the sentences in the teaching text data to obtain the semantic features of each sentence.

[0058] Based on the semantic features of each statement and the text structure of the teaching text data, the teaching text data is semantically segmented to obtain multiple teaching segments.

[0059] It should be noted that the natural language processing model in this application refers to a machine learning model that can understand, encode, and generate semantic features of natural language text through deep learning methods. Its function is to transform raw text data into vector representations with high-dimensional semantic expression. In this embodiment, the Bidirectional Encoder Representations from Transformers (BERT) pre-trained model can be used. This pre-trained model is based on the Transformer architecture and captures the global dependencies between different words in a sentence through a multi-layer self-attention mechanism to form context-related semantic embedding vectors. Its technical principle is to perform unsupervised pre-training on a large-scale text corpus to learn the semantic rules and structural patterns within the language. Then, a sentence is input in a specific task. The natural language processing model in this embodiment can extract feature representations containing rich semantic information through layer-by-layer encoding. The semantic features in this application refer to vectorized representations used to characterize deep semantic information in text content.

[0060] In practical implementation, firstly, the teaching text data can be preliminarily segmented according to natural sentences (based on periods, question marks, and exclamation marks as the segmentation criteria) to obtain a complete set of sentences. Then, a pre-trained natural language processing model, such as the BERT model, is used to input each sentence into the natural language processing model. After encoding by the embedding layer of the natural language processing model, word-level embeddings are generated. Then, the contextual relationships within the sentences are modeled through a multi-layer Transformer structure, and the overall semantic feature vector (i.e., semantic features) of each sentence is output. Usually, the vector corresponding to the CLS marker is taken as the unified representation of the sentence. To ensure feature stability, standardization processing (such as L2 regularization) can be performed after vector extraction. Then, all the extracted semantic feature vectors (i.e., semantic features) can be arranged according to the sentence order, and adjacent words can be calculated. The semantic similarity between sentences is calculated using the cosine similarity formula in this application. When the similarity between adjacent sentences is lower than a set threshold (pre-set to 0.7 in this application), it is determined that there is a semantic breakpoint. At the same time, the original structural information of the text, such as paragraph line breaks, chapter titles, and numbering marks (such as "1"), is used for auxiliary determination. Natural structural changes are prioritized as potential segmentation points. Combining semantic breakpoints and structural changes, the start and end boundaries of each teaching segment are determined through heuristic rules to ensure semantic coherence within the teaching segment and the independence of themes between segments. Finally, the structured segmentation of the teaching text data is completed, resulting in multiple teaching segments. It should be noted that the heuristic rules in this application are used to combine semantic similarity changes and text structure changes to quickly determine the segmentation boundaries of teaching segments in the teaching text, ensuring that the segmentation is both reasonable and efficient.

[0061] In step 103, the semantic dependency tree of each teaching segment is determined, and then the explicit dependency degree of the teaching logic between each teaching segment is determined based on the semantic dependency trees of all teaching segments and the semantic relevance between different teaching segments.

[0062] In some embodiments, reference Figure 2 As shown, this figure is a flowchart illustrating the process of determining the semantic dependency tree in some embodiments of this application. In this embodiment, determining the semantic dependency tree for each teaching segment specifically includes:

[0063] In step 1031, for each teaching segment, the part-of-speech tag for each word in the teaching segment is determined;

[0064] In step 1032, the dependency relationships between words are analyzed based on the tagged parts of speech of the words and the contextual information of the teaching segment;

[0065] In step 1033, the words are connected in a tree structure according to the dependency relationship between them to obtain the semantic dependency tree of the teaching segment, and then the semantic dependency tree of each teaching segment is obtained.

[0066] It should be noted that the semantic dependency tree in this application refers to a tree structure constructed by analyzing the grammatical dependency relationships between words in a sentence.

[0067] In specific implementation, for each teaching segment, firstly, a pre-trained part-of-speech tagging model (such as a part-of-speech tagging model based on the BiLSTM-CRF architecture) is called to tag each word in the text of the teaching segment, outputting the part-of-speech tag (such as noun, verb, adjective) corresponding to each word; then, the part-of-speech tags of each word are arranged according to the order of the corresponding words in the teaching segment, and the resulting sequence is used as the part-of-speech sequence; then, the part-of-speech sequence and the teaching segment are input into a pre-trained syntactic analysis model (in this application, a transition-based syntactic analysis model is used), and the syntactic analysis model synthesizes... Considering the part-of-speech, relative position, and contextual features of words in the teaching segment, the dependency relationship type (such as subject-verb, verb-object, and modifier relationships) between each pair of words is predicted using the learned parameters. This yields the dependency relationships between each word in the teaching segment, which are then stored in a list. Finally, based on the dependency relationships between words, with words as nodes and dependency relationships as directed edges, the system expands downwards from the core word (usually the predicate verb or headword of the sentence), connecting each subordinate word layer by layer to form a rooted, hierarchical tree structure, which is the semantic dependency tree of the teaching segment. This process is repeated for each teaching segment.

[0068] In some embodiments, determining the explicit dependency degree of instructional logic between instructional segments based on the semantic dependency tree of all instructional segments and the semantic relevance between different instructional segments specifically includes:

[0069] For every two teaching segments, extract the semantic feature vectors of the two teaching segments;

[0070] The semantic relevance between two teaching segments is determined by semantic feature vectors;

[0071] The syntactic and semantic relationships of the teaching logic between two teaching segments are determined based on the matching degree of the semantic dependency tree between the two teaching segments;

[0072] The explicit dependency of the teaching logic between two teaching segments is determined by the semantic relevance and the syntactic semantic relationship, thereby obtaining the explicit dependency of the teaching logic between each pair of teaching segments.

[0073] It should be noted that the semantic relevance in this application measures the similarity of two teaching segments in terms of thematic semantics; the explicit dependency in this application is an indicator that measures the strength of the direct logical order and syntactic semantic dependence between teaching segments.

[0074] In practical implementation, for every two teaching segments, firstly, a pre-trained natural language processing model (such as the BERT model) can be used to extract the semantic feature vectors of the two teaching segments; secondly, the Euclidean distance between the semantic feature vectors of the two teaching segments can be used as a measure of the semantic relevance between the two teaching segments, which can measure the similarity between the two teaching segments in terms of topic, content, etc.; then, the semantic dependency trees of the two teaching segments are compared, and tree matching algorithms (such as tree edit distance or node matching-based similarity calculation) are used to evaluate their structural similarity. The more similar the syntactic dependency tree structure, the more similar the two teaching segments are in terms of syntactic structure. The closer the segments are, the more likely they are to form a dependency relationship. Therefore, the structural similarity of the semantic dependency tree between two teaching segments can be used as the syntactic and semantic relationship of the teaching logic between the teaching segments. Finally, based on the degree of semantic relevance and the degree of matching of syntactic structure, the dependency strength between the two segments (e.g., sequential relationship, causal relationship, parallel relationship) can be obtained by weighting or combining. The explicit dependency degree can be identified by setting rules to identify the teaching logic relationship between segments (e.g., "premise-consequence" "sequence relationship"), and the output teaching logic relationship can be used as the explicit dependency degree of the teaching logic between the two teaching segments, thus obtaining the explicit dependency degree of the teaching logic between each pair of teaching segments.

[0075] In step 104, implicit association analysis is performed on the teaching semantics between each teaching segment based on the teaching intention of the teaching text data to obtain the implicit association degree of the teaching semantics between each teaching segment, and then the implicit constraints of the teaching logic between each teaching segment are determined through all the implicit association degrees.

[0076] In some embodiments, reference Figure 3 As shown in the figure, this is a flowchart illustrating the process of determining implicit association degree in some embodiments of this application. In this embodiment, implicit association analysis of the teaching semantics between each teaching segment is performed based on the teaching intent of the teaching text data to obtain the implicit association degree of teaching semantics between each teaching segment. Specifically, this includes:

[0077] In step 1041, the teaching semantics of each teaching segment are annotated to obtain the teaching semantics of each teaching segment;

[0078] In step 1042, the teaching semantics of each teaching segment are scored based on the teaching intent of the teaching text data to obtain the matching score of the teaching intent of each teaching segment.

[0079] In step 1043, implicit associations of the teaching semantics between each teaching segment are performed using all the matching scores to obtain the implicit association degree of the teaching semantics between each teaching segment.

[0080] It should be noted that the teaching intention in this application refers to the teaching objectives that the teaching activities are expected to achieve; the teaching semantics in this application is an indicator that quantifies the teaching purpose information in the teaching segment; the fit score in this application refers to the degree of matching between the teaching segment and the predetermined teaching intention; and the implicit correlation in this application is an indicator that measures the potential semantic correlation between teaching segments.

[0081] In practice, firstly, for each teaching segment, an existing pre-trained language model (such as BERT) can be used to process the segment, converting it into word embeddings and capturing its deep semantic information. Then, using semantic-based labeling techniques (such as label-based sequence labeling), appropriate semantic labels (such as concepts, topics, and learning objectives) are assigned to the teaching segment. These semantic labels clearly define the semantic content of the teaching segment. Next, the teaching intent of the teaching text data is input into a pre-trained intent recognition model (such as a Long Short-Term Memory network model). The intent recognition model then interprets the teaching intent of each segment. Regression analysis is performed on the teaching intention, and the regression value between the teaching semantics and the teaching intention of each teaching segment is used as the matching score between the corresponding teaching segment and the teaching intention. The matching score can reflect the degree of matching of the teaching segment in achieving the teaching objectives. The higher the matching score, the more consistent the segment is with the teaching intention. Finally, an association matrix between teaching segments is constructed based on all the matching scores, where each element represents the association value of the matching degree between two teaching segments. In this application, the Euclidean distance between matching scores is used as the association value of the matching degree, and the natural exponential function value of the negative number of the association value can be used as the implicit association degree of the teaching semantics between the corresponding two teaching segments.

[0082] In some embodiments, determining the implicit constraints of the teaching logic between teaching segments through all implicit correlations specifically includes:

[0083] The teaching difficulty of each teaching segment is assessed, and a difficulty coefficient matrix of the teaching segment is generated.

[0084] For any two teaching segments, the association attention of the teaching logic between the two teaching segments is determined based on the implicit correlation of their teaching semantics and the difficulty coefficient matrix.

[0085] Construct and initialize a machine learning model for modeling implicit logical relationships between teaching segments;

[0086] Set the associated attention as the connection weight in the forward propagation phase of the machine learning model;

[0087] The implicit logical dependencies between any two teaching segments are extracted using the machine learning model, and implicit constraints between the two teaching segments are generated, thereby obtaining a set of implicit constraints between all pairs of teaching segments.

[0088] It should be noted that the difficulty coefficient matrix in this application is a matrix that quantifies the relative difficulty of each teaching segment and the degree of difference between them; the associative attention in this application is an indicator that measures the strength of the association between two teaching segments in the implicit logical relationship; and the implicit constraints in this application are indicators that reflect the strength of the potential semantic association and knowledge transfer relationship between teaching segments.

[0089] In specific implementation, firstly, the teaching difficulty of each teaching segment in the teaching path is assessed. Existing deep learning models, such as the BERT model or other pre-trained language models suitable for text semantic understanding, can be used. Through semantic analysis of the teaching text content, combined with teaching difficulty standards in the education field (such as the Ministry of Education or textbook difficulty level classification standards), each segment is scored, generating a quantitative index representing the relative difficulty and differences of each teaching segment, i.e., a difficulty coefficient. The difficulty coefficients of all teaching segments are arranged into a matrix according to the teaching order to form the difficulty coefficient matrix. This matrix can be used to reflect the relative teaching difficulty and difficulty differences of each segment. Secondly, for any two teaching segments, the corresponding difficulty coefficients are obtained from the difficulty coefficient matrix, and the difficulty difference between the two teaching segments (i.e., the absolute difference between the two difficulty coefficients) is calculated. At the same time, the implicit semantic correlation between the two segments is calculated through semantic feature vectors. The product of implicit semantic association and difficulty difference is normalized to obtain the teaching logical association attention between two teaching segments. This association attention is used to measure the strength of the potential dependency between the two teaching segments in the implicit logical relationship. Furthermore, the network input structure can be designed based on the implicit association and teaching difficulty matrix. The semantic features, implicit association, and difficulty features of each pair of teaching segments are used as input tensors. A model framework with good relation modeling capabilities, such as a Transformer-based relation encoder, is selected, and its initial parameters are set, including the number of multi-head attention layers, the size of hidden layers, and the forward propagation network structure. Further, the association attention of each pair of teaching segments is used as the initial connection weights and embedded in the model's self-attention mechanism to guide the model during training. The model prioritizes segments with strong correlations and defines cross-entropy or mean squared error as the training loss function. Weights are then optimized using subsequent data input and gradient descent to enable the model to accurately capture implicit logical dependencies between segments, thus initializing the machine learning model. Next, the initialized model is input into training data for forward propagation and inference. The model optimizes the implicit logical dependencies between segments based on the connection weights, and the implicit dependencies captured by the model are mapped to implicit constraints. These implicit constraints represent the potential dependencies between teaching segments in the teaching path. By repeating the above steps, corresponding implicit constraints are generated for every two teaching segments in the teaching path, ultimately obtaining a complete set of implicit constraints between all pairs of teaching segments.

[0090] In step 105, a collaborative relationship graph between teaching segments is constructed based on the explicit dependency and implicit constraints of the teaching logic between each teaching segment during the allocation of teaching resources. Then, the teaching resources in the teaching equipment are adaptively allocated according to the collaborative relationship graph.

[0091] In some embodiments, constructing a collaborative relationship graph among teaching segments during the allocation of teaching resources based on the explicit dependencies and implicit constraints of the teaching logic among the teaching segments specifically includes:

[0092] A teaching path map with explicit dependencies is established based on the explicit dependencies between each teaching segment, which is used to represent the direct logical order and prerequisite relationships between teaching segments;

[0093] Based on the implicit constraints between each teaching segment, potential semantic associations and knowledge transfer relationships are extracted, and then a teaching path map with implicit constraints is constructed to supplement the potential logical links not covered by explicit dependencies.

[0094] By performing a fusion analysis on the explicitly dependent teaching path graph and the implicitly constrained teaching path graph, the collaborative constraint relationship in the teaching resource allocation process is determined.

[0095] Based on the collaborative constraints in the allocation of teaching resources, a comprehensive model is built on the dependence intensity, sequence, and resource sharing conditions among the teaching segments, thereby generating a collaborative relationship map among the teaching segments in the allocation of teaching resources.

[0096] It should be noted that the explicit dependency teaching path graph in this application refers to a directed graph that reflects the direct logical order and prerequisite relationships between teaching segments; the implicit constraint teaching path graph in this application refers to a weighted graph that reflects the potential semantic associations and knowledge transfer relationships between teaching segments; the collaborative constraint relationship in this application is a feature that reflects the intensity of resource allocation dependency and coordination characteristics between teaching segments; and the collaborative relationship graph in this application refers to a structured dependency graph used to optimize the allocation of teaching resources and coordinate the connection between segments.

[0097] In practical implementation, firstly, the explicit dependency degree of each teaching segment is calculated and modeled. Existing natural language processing technologies, including dependency parsing, semantic role labeling, and text ranking algorithms, can be used to parse the text content of each teaching segment. Dependency parsing extracts the core concepts, prerequisite knowledge points, and logical order of each segment, and semantic role labeling identifies knowledge transfer relationships. The teaching segments are mapped to nodes in a graph, and the direct logical dependencies between segments are mapped to directed edges. To ensure the quantifiability and operability of explicit dependency degree, each directed edge can be assigned a weight. The weight can be quantified based on the importance of the logical order between segments or the progression of teaching objectives, thereby generating... The first step is to construct an explicit dependency teaching path graph that reflects the direct logical order and prior relationships of teaching segments. Secondly, for the implicit constraints between teaching segments, it is necessary to extract latent semantic associations and knowledge transfer relationships. This can be achieved by using sentence vectors or segment vectors based on pre-trained language models (such as BERT, RoBERTa, ERNIE, etc.) to obtain the semantic features of each teaching segment, and calculating the semantic similarity or latent topic relevance between any two segments as an implicit association index. Simultaneously, the teaching difficulty, knowledge point coverage, and other features of each segment are normalized to generate an association attention matrix reflecting the strength of latent dependencies. Furthermore, this association attention matrix is ​​used to construct an implicitly constrained teaching path graph. The latent semantic or knowledge transfer relationships between segments are mapped as graph edge weights to supplement the logical links not covered by explicit dependencies, thereby capturing the hidden teaching logic connections between teaching segments. Then, the explicit dependency graph and the implicit constraint graph are fused. Graph merging algorithms or graph neural networks can be used for node matching, edge weighting integration, and handling of repeated relationships. For example, nodes can be unified by using the same name or semantic similarity. The explicit dependency weights and implicit constraint weights can be weighted and averaged according to priority or learning objectives. Through fusion analysis, a comprehensive collaborative relationship graph can be obtained. The dependency relationships of each teaching segment in the graph retain both the explicit logical order and integrate implicit latent constraints, while also considering the dependency strength and sequence between segments. The order and resource sharing conditions are labeled. Finally, teaching resource allocation is modeled based on the collaborative relationship graph. Specifically, graph analysis algorithms (such as topological sorting, shortest path, or maximum flow algorithms) combined with constraint solvers can be used to calculate and optimize the order, priority, and potential conflicts of teaching resource allocation. For example, weighted topological sorting can be used to determine the order of resource allocation, edge weights can be used to determine the priority allocation of resources for key segments, and graph constraint solvers can be used to coordinate resource sharing between different segments to minimize resource conflicts between segments. The resulting collaborative relationship graph not only reflects the direct logical dependencies between teaching segments but also embodies potential knowledge transfer and resource sharing constraints, thus providing a scientific and operable basis for teaching resource allocation.

[0098] It should be noted that this application's solution utilizes the explicit dependencies and implicit constraints of teaching segments to construct a collaborative relationship graph among teaching segments during the allocation of teaching resources. This effectively solves two major technical problems existing in the resource allocation and logical connection of teaching in the current technology: First, existing methods usually only consider the explicit order or preconditions of segments, making it difficult to reflect potential semantic associations and knowledge transfer, which may lead to resource allocation ignoring the inherent logic of teaching content and learning coherence; second, traditional resource scheduling is mostly based on static rules or single-segment optimization, which cannot simultaneously consider the dependency strength, sequence, and resource sharing needs between segments, easily leading to resource conflicts or unreasonable teaching sequences. This solution captures the direct logical sequence by establishing a teaching path graph with explicit dependencies, supplements the potential logical links with an implicit constraint graph, and obtains collaborative constraint relationships through fusion analysis. This achieves comprehensive modeling of the dependency strength, sequence, and resource sharing conditions among segments, thereby generating a collaborative relationship graph. Thus, this solution can achieve overall optimization of teaching resource allocation, making the logical connection between teaching segments more reasonable, resource utilization more efficient, and improving the coherence of teaching paths and learning outcomes.

[0099] It should also be noted that the adaptive allocation of teaching resources in the teaching equipment based on the aforementioned collaborative relationship graph in this application refers to dynamically adjusting the resource configuration of each teaching segment to optimize teaching effectiveness by utilizing the dependency relationships and resource coordination characteristics reflected in the graph. Specifically, firstly, the dependency strength, sequence, and resource sharing conditions of each teaching segment in the collaborative relationship graph are used as input features to form a constraint matrix for teaching resource allocation. Secondly, a resource scheduling module can be established in the teaching equipment. This module analyzes resource conflicts and priorities between segments based on the constraint matrix and uses known scheduling algorithms, such as weighted topological sorting, priority queues, or simple greedy scheduling strategies, to determine the resource allocation order and amount for each teaching segment on the equipment. Then, the teaching equipment dynamically allocates computational resources, display resources, or interactive resources to the corresponding teaching segments based on the scheduling results. Simultaneously, it monitors resource usage in real time during execution and makes fine adjustments based on the segment completion status and resource occupancy to achieve adaptive allocation. In this way, teaching resources can be rationally allocated according to the dependency and constraint characteristics in the collaborative relationship graph, ensuring that key segments receive necessary resources first, while reducing resource conflicts and improving overall teaching efficiency and learning continuity.

[0100] On the other hand, in some embodiments, this application provides a teaching resource recommendation system based on natural language processing, with reference to... Figure 4 The figure is a schematic diagram of the structure of a teaching resource recommendation system based on natural language processing according to some embodiments of this application. The teaching resource recommendation system based on natural language processing includes: an acquisition module 401, a feature processing module 402, and an allocation module 403, which are described below:

[0101] The acquisition module 401 is used to acquire teaching text data to be allocated from the teaching equipment;

[0102] The feature processing module 402 is used to perform structured segmentation of the teaching text data based on pre-trained natural language processing models to obtain multiple teaching segments.

[0103] The feature processing module 402 is also used to determine the semantic dependency tree of each teaching segment, and then determine the explicit dependency degree of the teaching logic between each teaching segment based on the semantic dependency trees of all teaching segments and the semantic relevance between different teaching segments.

[0104] The feature processing module 402 is also used to perform implicit association analysis on the teaching semantics between each teaching segment according to the teaching intention of the teaching text data, to obtain the implicit association degree of the teaching semantics between each teaching segment, and then determine the implicit constraints of the teaching logic between each teaching segment through all the implicit association degrees.

[0105] The allocation module 403 is used to construct a collaborative relationship graph between teaching segments during the allocation of teaching resources based on the explicit dependency and implicit constraints of the teaching logic between each teaching segment, and then adaptively allocate teaching resources in the teaching equipment according to the collaborative relationship graph.

[0106] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described teaching resource recommendation method based on natural language processing.

[0107] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device implementing a natural language processing-based teaching resource recommendation method according to some embodiments of this application. The natural language processing-based teaching resource recommendation method in the above embodiments can... Figure 5 The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0108] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0109] The communication bus 502 can be used to transmit information between the aforementioned components.

[0110] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0111] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. In the above embodiments, the teaching resource recommendation method based on natural language processing can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0112] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0113] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0114] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0115] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described teaching resource recommendation method based on natural language processing.

[0116] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0117] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A teaching resource recommendation method based on natural language processing, characterized in that, include: Obtain teaching text data from teaching equipment for the allocation of teaching resources; The teaching text data is structured and segmented based on text features using a pre-trained natural language processing model to obtain multiple teaching segments. The semantic dependency tree of each teaching segment is determined. For every two teaching segments, the semantic feature vectors of the two teaching segments are extracted. The semantic relevance between the two teaching segments is determined by the semantic feature vectors. The syntactic and semantic relations of the teaching logic between the two teaching segments are determined based on the matching degree of the semantic dependency trees between the two teaching segments. The explicit dependency degree of the teaching logic between the two teaching segments is determined by the semantic relevance and the syntactic and semantic relations. Thus, the explicit dependency degree of the teaching logic between every two teaching segments is obtained. The explicit dependency degree is an indicator that measures the strength of the direct logical order and syntactic and semantic dependence between teaching segments. Each teaching segment is labeled with teaching semantics to obtain the teaching semantics of each teaching segment. Based on the teaching intention of the teaching text data, the teaching semantics of each teaching segment are matched and scored to obtain the matching score of the teaching intention of each teaching segment. The teaching semantics between each teaching segment are implicitly associated through all matching scores to obtain the implicit association degree of teaching semantics between each teaching segment. The implicit association degree is an indicator that measures the potential association between teaching segments at the semantic level. Then, the implicit constraints of the teaching logic between each teaching segment are determined through all implicit association degrees. Based on the explicit dependency and implicit constraints of the teaching logic between each teaching segment, a collaborative relationship graph is constructed between each teaching segment during the allocation of teaching resources. Then, the teaching resources in the teaching equipment are adaptively allocated according to the collaborative relationship graph.

2. The method as described in claim 1, characterized in that, The teaching text data refers to the electronic textbook data involved in the teaching process.

3. The method as described in claim 1, characterized in that, Based on a pre-trained natural language processing model, the teaching text data is structurally segmented using text features to obtain multiple teaching segments, specifically including: Based on a pre-trained natural language processing model, semantic features are extracted from the sentences in the teaching text data to obtain the semantic features of each sentence. Based on the semantic features of each statement and the text structure of the teaching text data, the teaching text data is semantically segmented to obtain multiple teaching segments.

4. The method as described in claim 1, characterized in that, Determining the semantic dependency tree for each teaching segment specifically includes: For each teaching segment, determine the part of speech of each word in the segment; Analyze the dependency relationships between words based on their parts of speech and the contextual information of the teaching segment; Based on the dependency relationships between words, a tree structure is constructed to connect the words, resulting in a semantic dependency tree for each teaching segment, and thus a semantic dependency tree for each teaching segment.

5. The method as described in claim 1, characterized in that, The implicit constraints on the teaching logic between each teaching segment, determined by all implicit correlations, specifically include: The teaching difficulty of each teaching segment is assessed, and a difficulty coefficient matrix of the teaching segment is generated. For any two teaching segments, the association attention of the teaching logic between the two teaching segments is determined based on the implicit correlation of their teaching semantics and the difficulty coefficient matrix. Construct and initialize a machine learning model for modeling implicit logical relationships between teaching segments; Set the associated attention as the connection weight in the forward propagation phase of the machine learning model; The implicit logical dependencies between any two teaching segments are extracted using the machine learning model, and implicit constraints between the two teaching segments are generated, thereby obtaining a set of implicit constraints between all pairs of teaching segments.

6. A teaching resource recommendation system based on natural language processing, used to perform the method as described in any one of claims 1 to 5, characterized in that, include: The acquisition module is used to acquire teaching text data for the allocation of teaching resources from teaching equipment; The feature processing module is used to perform structured segmentation of the teaching text data based on pre-trained natural language processing models to obtain multiple teaching segments. The feature processing module is also used to determine the semantic dependency tree of each teaching segment, and then determine the explicit dependency degree of the teaching logic between each teaching segment based on the semantic dependency trees of all teaching segments and the semantic relevance between different teaching segments. The feature processing module is also used to perform implicit association analysis on the teaching semantics between each teaching segment according to the teaching intention of the teaching text data, to obtain the implicit association degree of the teaching semantics between each teaching segment, and then determine the implicit constraints of the teaching logic between each teaching segment through all the implicit association degrees. The allocation module is used to construct a collaborative relationship graph between teaching segments during the allocation process based on the explicit dependency and implicit constraints of the teaching logic between each teaching segment, and then adaptively allocate teaching resources in the teaching equipment according to the collaborative relationship graph.

7. A computer device comprising a memory and a processor, the memory storing code, characterized in that, The processor is configured to acquire the code and execute the teaching resource recommendation method based on natural language processing as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the teaching resource recommendation method based on natural language processing as described in any one of claims 1 to 5.

Images